back to indexJohn Carmack: Doom, Quake, VR, AGI, Programming, Video Games, and Rockets | Lex Fridman Podcast #309
link |
I remember the reaction where he had drawn these characters
link |
and he was slowly moving around
link |
and like people had no experience with 3D navigation.
link |
It was all still keyboard.
link |
We didn't even have mice set up at that time,
link |
but slowly moving, going up, picked up a key, go to a wall.
link |
The wall disappears in a little animation
link |
and there's a monster like right there.
link |
And he practically fell out of his chair.
link |
It was just like, ah, and games just didn't do that.
link |
You know, the games were the God's eye view.
link |
You were a little invested in your little guy.
link |
You can be like, you know, happy or sad when things happen,
link |
but you just did not get that kind of startled reaction.
link |
You weren't inside the game.
link |
Something in the back of your brain,
link |
some reptile brain thing is just going,
link |
oh shit, something just happened.
link |
And that was one of those early points where it's like,
link |
yeah, this is going to make a difference.
link |
This is going to be powerful and it's going to matter.
link |
The following is a conversation with John Carmack,
link |
widely considered to be one of the greatest programmers
link |
He was the cofounder of id Software
link |
and the lead programmer on several games
link |
that revolutionized the technology, the experience,
link |
and the role of gaming in our society,
link |
including Commander Keen, Wolfenstein 3D, Doom, and Quake.
link |
He spent many years as the CTO of Oculus VR,
link |
helping to create portals into virtual worlds
link |
and to define the technological path
link |
to the metaverse and meta.
link |
And now he has been shifting some of his attention
link |
to the problem of artificial general intelligence.
link |
This was the longest conversation on this podcast
link |
at over five hours.
link |
And still, I could talk to John many, many more times,
link |
and we hope to do just that.
link |
This is the Lux Readman podcast.
link |
To support it, please check out our sponsors
link |
in the description.
link |
And now, dear friends, here's John Carmack.
link |
What was the first program you've ever written?
link |
So I remember being in a radio shack
link |
going up to the TRS 80 computers
link |
and learning just enough to be able to do
link |
10 print John Carmack.
link |
And it's kind of interesting how, of course,
link |
Carnegie and Ritchie kind of standardized Hello World
link |
as the first thing that you do
link |
in every computer programming language in every computer,
link |
but not having any interaction
link |
with the cultures of Unix or any other standardized things.
link |
It was just like, well, what am I gonna say?
link |
I'm gonna say my name.
link |
And then you learn how to do go to 10
link |
and have it scroll all off the screen.
link |
And that was definitely the first thing
link |
that I wound up doing on a computer.
link |
Can I ask you a programming advice?
link |
I was always told in the beginning
link |
that you're not allowed to use go to statements.
link |
That's really bad programming.
link |
Is this correct or not?
link |
Jumping around code, can we look at the philosophy
link |
and the technical aspects of the go to statement
link |
that seems so convenient,
link |
but it's supposed to be bad programming?
link |
Well, so certainly back in the day
link |
in basic programming languages,
link |
you didn't have proper loops.
link |
You didn't have four whiles and repeats.
link |
That was the land of Pascal for people
link |
that kind of generally had access to it back then.
link |
So you had no choice but to use go tos.
link |
And as you made what were big programs back then,
link |
which were a thousand line basic program
link |
is a really big program.
link |
They did tend to sort of degenerate into madness.
link |
You didn't have good editors or code exploration tools.
link |
So you would wind up fixing things in one place,
link |
add a little patch.
link |
And there's reasons why structured programming
link |
generally helps understanding,
link |
but go tos aren't poisonous.
link |
Sometimes they're the right thing to do.
link |
Usually it's because there's a language feature missing
link |
like nested breaks or something
link |
where it can sometimes be better to do a go to cleanup
link |
or go to error rather than having multiple flags,
link |
multiple if statements littered throughout things.
link |
I mean, if you gripped through all of my code right now,
link |
I don't think any of my current code bases
link |
would actually have a go to,
link |
but deep within sort of the technical underpinnings
link |
of a major game engine,
link |
you're gonna have some go tos in a couple of places probably.
link |
Yeah, the infrastructure on top of,
link |
like the closer you get to machine code,
link |
the more go tos you're gonna see,
link |
the more of these like hacks you're going to see
link |
because the set of features available to you
link |
in low level programming languages is not, is limited.
link |
So print John Carmack,
link |
when is the first time,
link |
if we could talk about love,
link |
that you fell in love with programming?
link |
You said like, this is really something special.
link |
It really was something
link |
that was one of those love at first sight things
link |
where just really from the time
link |
that I understood what a computer was even,
link |
I mean, I remember looking through old encyclopedias
link |
of the black and white photos of the IBM mainframes
link |
at the reel to reel tape decks.
link |
And for people nowadays,
link |
it can be a little hard to understand
link |
what the world was like then from information gathering
link |
where I would go to the libraries
link |
and there would be a couple books on the shelf
link |
about computers and they would be very out of date
link |
even at that point, just not a lot of information,
link |
but I would grab everything that I could find
link |
and devour everything.
link |
Whenever Time or Newsweek had some article about computers,
link |
I would like cut it out with scissors and put it somewhere.
link |
It just, it felt like this magical thing to me,
link |
this idea that the computer would just do exactly
link |
what you told it to.
link |
I mean, and there's a little bit of the genie monkey's paw
link |
sort of issues there where you'd better be really,
link |
really careful with what you're telling it to do,
link |
but it wasn't gonna back talk you.
link |
It wasn't gonna have a different point of view.
link |
It was gonna carry out what you told it to do.
link |
And if you had the right commands,
link |
you could make it do these pretty magical things.
link |
And so what kind of programs did you write at first?
link |
So beyond the print, John Carmack.
link |
So I can remember as going through the learning process
link |
where you find at the start,
link |
you're just learning how to do
link |
the most basic possible things.
link |
And I can remember stuff like a Superman comic
link |
that RadioShack commissioned to have,
link |
it's like Superman had lost some of his super brain
link |
and kids had to use RadioShack TRS 80 computers
link |
to do calculations for it to help him
link |
kind of complete his heroics.
link |
And I'd find little things like that
link |
and then get a few basic books
link |
to be able to kind of work my way up.
link |
And again, it was so precious back then.
link |
I had a couple of books
link |
that would teach me important things about it.
link |
I had one book that I could start to learn
link |
a little bit of assembly language from,
link |
and I'd have a few books on basic
link |
and some things that I could get from the libraries.
link |
But my goals in the early days
link |
was almost always making games of various kinds.
link |
I loved the arcade games and the early Atari 2600 games.
link |
And being able to do some of those things myself
link |
on the computers was very much what I aspired to.
link |
And it was a whole journey where
link |
if you learn normal basic,
link |
you can't do any kind of an action game.
link |
You can write an adventure game.
link |
You can write things where you say, what do you do here?
link |
I get sword attack troll, that type of thing.
link |
And that can be done in the context of basic.
link |
But to do things that had moving graphics,
link |
there were only the most limited things
link |
you could possibly do.
link |
You could maybe do breakout or pong
link |
or that sort of thing in low resolution graphics.
link |
And in fact, one of my first sort of major technical hacks
link |
that I was kind of fond of was on the Apple II computers,
link |
they had a mode called low resolution graphics
link |
where of course all graphics were low resolution back then,
link |
but regular low resolution graphics,
link |
it was a grid of 40 by 40 pixels normally,
link |
but they could have 16 different colors.
link |
And I wanted to make a game kind of like
link |
the arcade game Vanguard, just a scrolling game.
link |
And I wanted to just kind of have it scroll vertically up.
link |
And I could move a little ship around.
link |
You could manage to do that in basic,
link |
but there's no way you could redraw the whole screen.
link |
And I remember at the time just coming up
link |
with what felt like a brainstorm to me
link |
where I knew enough about the way the hardware was controlled
link |
where the text screen and the low resolution graphics screen
link |
were basically the same thing.
link |
And all those computers could scroll their text screen
link |
You could do a listing and it would scroll things up.
link |
And I figured out that I could kind of tweak just a couple
link |
things that I barely understood to put it into a graphics mode
link |
and I could draw graphics and then I could just
link |
do a line feed at the very bottom of the screen.
link |
And then the system would scroll it all up
link |
using an assembly language routine
link |
that I didn't know how to write back then.
link |
So that was like this first great hack
link |
that sort of had analogs later on in my career
link |
for a lot of different things.
link |
So I found out that I could draw a screen.
link |
I could do a line feed at the bottom.
link |
It would scroll it up once.
link |
I could draw a couple more lines of stuff at the bottom.
link |
And that was my first way to kind of scroll
link |
the screen, which was interesting in that that played
link |
a big part later on in the id Software days as well.
link |
So do efficient drawing where you
link |
don't have to draw the whole screen,
link |
but you draw from the bottom using
link |
the thing that was designed in the hardware for text output.
link |
Where so much of, until recently,
link |
game design was limited by what you could actually
link |
get the computer to do.
link |
Where it's easy to say, OK, I want to scroll the screen.
link |
You just redraw the entire screen at a slight offset.
link |
And nowadays, that works just fine.
link |
Computers are ludicrously fast.
link |
But up until a decade ago or so, there
link |
were all these things everybody wanted to do.
link |
But if they knew enough programming
link |
to be able to make it happen, it would happen too slow
link |
to be a good experience.
link |
Either just ridiculously slow or just slow enough
link |
that it wasn't fun to experience it like that.
link |
So much of the first couple of decades of the programming work
link |
that I did was largely figuring out
link |
how to do something that everybody
link |
knows how they want it to happen.
link |
It just has to happen 2 to 10 times faster
link |
than the straightforward way of doing things
link |
would make it happen.
link |
And it's different now because at this point, lots of things
link |
you can just do in the most naive possible way,
link |
and it still works out.
link |
You don't have nearly the creative limitations
link |
or the incentives for optimizing on that level.
link |
And there's a lot of pros and cons to that.
link |
But I do generally, I'm not going
link |
to do the angry old man shaking my fist at the clouds bit
link |
where back in my day, programmers
link |
had to do real programming.
link |
It's amazing that you can just pick an idea
link |
and go do it right now.
link |
And you don't have to be some assembly language wizard
link |
or deep GPU arcanist to be able to figure out
link |
how to make your wishes happen.
link |
Well, there's still, see, that's true.
link |
But let me put on my old man with a fist hat
link |
and say that probably the thing that will define the future
link |
still requires you to operate at the limits
link |
of the current system.
link |
So we'll probably talk about this.
link |
But if you talk about building the metaverse
link |
and building a VR experience that's compelling,
link |
it probably requires you to not, to go to assembly
link |
or maybe not literally, but sort of spiritually
link |
to go to the limits of what the system is capable of.
link |
And that really was why virtual reality
link |
was specifically interesting to me
link |
where it had all the ties to,
link |
you could say that even back in the early days,
link |
I have some old magazine articles
link |
that's talking about Doom as a virtual reality experience
link |
back when just seeing anything in 3D.
link |
So you could say that we've been trying to build
link |
those virtual experiences from the very beginning.
link |
And in the modern era of virtual reality,
link |
especially on the mobile side of things,
link |
when it's standalone and you're basically using
link |
a cell phone chip to be able to produce
link |
these very immersive experiences,
link |
it does require work.
link |
It's not at the level of what an old school
link |
console game programmer would have operated at
link |
where you're looking at hardware registers
link |
and you're scheduling all the DMA accesses,
link |
but it is still definitely a different level
link |
than what a web developer or even a PC Steam game developer
link |
usually has to work at.
link |
And again, it's great.
link |
There's opportunities for people that wanna operate
link |
at either end of that spectrum there
link |
and still provide a lot of value to the world.
link |
Let me ask you sort of a big question about preference.
link |
What would you say is the best programming language?
link |
Your favorite, but also the best.
link |
You've seen throughout your career,
link |
you're considered by many to be
link |
the greatest programmer ever.
link |
I mean, it's so difficult to place that label on anyone,
link |
but if you put it on anyone, it's you.
link |
So let me ask you these kind of ridiculous questions
link |
of what's the best band of all time,
link |
but in your case, what's the best programming language?
link |
Everything has all the caveats about it.
link |
But so what I use, so nowadays I do program
link |
a reasonable amount of Python for AI, ML sorts of work.
link |
I'm not a native Python programmer.
link |
It's something I came to very late in my career.
link |
I understand what it's good for.
link |
But you don't dream in Python.
link |
And it has some of those things
link |
where there's some amazing stats when you say,
link |
if you just start, if you make a loop,
link |
a triply nested loop and start doing operations in Python,
link |
you can be thousands to potentially a million times slower
link |
than a proper GPU tensor operation.
link |
And these are staggering numbers.
link |
You can be as much slower as we've almost gotten faster
link |
in our pace of progress and all this other miraculous stuff.
link |
So your intuitions about inefficiencies
link |
within the Python sort of...
link |
It keeps hitting me upside the face
link |
where it's gotten to the point now I understand.
link |
It's like, okay, you just can't do a loop
link |
if you care about performance in Python.
link |
You have to figure out how you can reformat this
link |
into some big vector operation
link |
or something that's going to be done completely
link |
within a C++ library.
link |
But the other hand is it's amazingly convenient
link |
and you just see stuff that people are able
link |
to cobble together by you just import
link |
a few different things and you can do stuff
link |
that nobody on earth could do 10 years ago.
link |
And you can do it in a little cookbook thing
link |
that you copy paste it out of a website.
link |
So that is really great.
link |
When I'm sitting down to do what I consider
link |
kind of serious programming, it's still in C++.
link |
And it's really kind of a C flavored C++ at that
link |
where I'm not big into the modern template
link |
metaprogramming sorts of things.
link |
I see a lot of train wrecks coming from
link |
some of that over abstraction.
link |
I spent a few years really going kind of deep
link |
into the kind of the historical Lisp work
link |
and Haskell and some of the functional programming
link |
And there is a lot of value there
link |
in the way you think about things.
link |
And I changed a lot of the way I write my C and C++ code
link |
based on what I learned about the value
link |
that comes out of not having this random mutable state
link |
that you kind of lose track of.
link |
Because something that many people don't really appreciate
link |
till they've been at it for a long time
link |
is that it's not the writing of the program initially,
link |
it's the whole lifespan of the program.
link |
And that's when it's not necessarily just
link |
how fast you wrote it or how fast it operates,
link |
but it's how can it bend and adapt
link |
as situations change.
link |
And then the thing that I've really been learning
link |
in my time at Meta with the Oculus and VR work
link |
is it's also how well it hands off
link |
between the continuous kind of revolving door of programmers
link |
taking over maintenance and different things
link |
and how you get people up to speed in different areas.
link |
And there's all these other different aspects of it.
link |
So C++ is a good language for handover between engineers.
link |
Probably not the best.
link |
And there's some really interesting aspects to this
link |
where in some cases languages
link |
that are not generally thought well of for many reasons,
link |
like C is derided pretty broadly that yes,
link |
obviously all of these security flaws
link |
that happen with the memory and unsafeness
link |
and buffer overruns and the things that you've got there,
link |
but there is this underappreciated aspect
link |
to the language is so simple, anyone can go.
link |
And if you know C, you can generally jump in someplace
link |
and not have to learn what paradigms they're using
link |
because there just aren't that many available.
link |
I think there's some really, really well written C code.
link |
Like I find it great that if I'm messing around
link |
with something in open BSD say,
link |
I mean, I can be walking around in the kernel
link |
and I'm like, I understand everything that's going on here.
link |
It's not hard for me to figure out what I need to do
link |
to make whatever change that I need to
link |
while you can have more significant languages.
link |
Like it's a downside of Lisp
link |
where I don't regret the time that I spent with Lisp.
link |
I think that it did help my thinking
link |
about programming in some ways,
link |
but the people that are the biggest defenders of Lisp
link |
are saying how malleable of a language it is
link |
that if you write a huge Lisp program,
link |
you've basically invented your own kind of language
link |
and structure because it's not the primitives
link |
of the language you're using very much.
link |
It's all of the things you've built on top of that.
link |
And then a language like Racket,
link |
kind of one of the more modern Lisp versions,
link |
it's essentially touted as a language
link |
for building other languages.
link |
And I understand the value of that
link |
for a tiny little project,
link |
but the idea of that for one of these longterm
link |
supported by lots of people kind of horrifies me
link |
where all of those abstractions that you're like,
link |
okay, you can't touch this code till you educate yourself
link |
on all of these things that we've built on top of that.
link |
And it was interesting to see how when Google made Go,
link |
a lot of the criticisms of that are it's like,
link |
wow, this is not a state of the art language.
link |
This language is just so simple and almost crude.
link |
And you could see the programming language people
link |
just looking down at it.
link |
But it does seem to be quite popular as basically saying,
link |
this is the good things about C,
link |
everybody can just jump right in and use it.
link |
You don't need to restructure your brain
link |
to write good code in it.
link |
So I wish that I had more opportunity
link |
for doing some work in Go.
link |
Rust is the other modern language that everybody talks about
link |
that I'm not fit to pass judgment on.
link |
I've done a little bit beyond Hello World.
link |
I wrote some like video decompression work in Rust
link |
just as an exercise, but that was a few years ago
link |
and I haven't really used it since.
link |
The best programming language is the one that works generally
link |
that you're currently using.
link |
Because that's another trap is in almost every case
link |
I've seen when people mixed languages on a project,
link |
I would rather stay just in one language
link |
so that everybody can work across the entire thing.
link |
And we have, I get meta, we have a lot of projects
link |
that use kind of React frameworks.
link |
So you've got JavaScript here
link |
and then you have C++ for real work.
link |
And then you may have Java interfacing
link |
with some other part of the Android system.
link |
And those are all kind of horrible things.
link |
And that was one thing that I remember talking
link |
with Boz at Facebook about it where like,
link |
man, I wish we could have just said,
link |
we're only hiring C++ programmers.
link |
And he just thought from the Facebook meta perspective,
link |
well, we just wouldn't be able to find enough.
link |
With the thousands of programmers they've got there,
link |
it is not necessarily a dying breed,
link |
but you can sure find a lot more Java
link |
or JavaScript programmers.
link |
And I kind of mentioned that to Elon one time
link |
and he was kind of flabbergasted about that.
link |
It's like, well, you just,
link |
you go out and you find those programmers
link |
and you don't hire the other programmers
link |
that don't do the languages that you wanna use.
link |
But right now, I guess, yeah,
link |
they're using JavaScript on a bunch of the SpaceX work
link |
for the UI side of things.
link |
When you go find UI programmers,
link |
they're JavaScript programmers.
link |
I wonder if that's because there's a lot
link |
of JavaScript programmers.
link |
Because I do think that great programmers are rare.
link |
That it's not, you know, if you just look at statistics
link |
of how many people are using different programming languages,
link |
that doesn't tell you the story
link |
of what the great programmers are using.
link |
And so you have to really look at what you were speaking to,
link |
which is the fundamentals of a language.
link |
What does it encourage?
link |
How does it encourage you to think?
link |
What kind of systems does it encourage you to build?
link |
There is something about C++
link |
that has elements of creativity,
link |
but forces you to be an adult about your programming.
link |
It expects you to be an adult.
link |
It does not force you to.
link |
And so it brings out people that are willing to be creative
link |
in terms of building large systems
link |
and coming up with interesting solutions,
link |
but at the same time have the sort of the good
link |
software engineering practices that amend themselves
link |
to real world systems.
link |
Let me ask you about this other language, JavaScript.
link |
So if we, you know, aliens visit in thousands of years,
link |
and humans are long gone, something tells me
link |
that most of the systems they find will be run by humans.
link |
Will be running JavaScript.
link |
I kind of think that if we're living in a simulation,
link |
it's written in JavaScript.
link |
You know, for the longest time, even still,
link |
JavaScript didn't get any respect,
link |
and yet it runs so much of the world,
link |
and an increasing number of the world.
link |
Is it possible that everything will be written
link |
in JavaScript one day?
link |
So the engineering under JavaScript
link |
is really pretty phenomenal.
link |
The systems that make JavaScript run as fast
link |
as it does right now are kind of miracles
link |
of modern engineering in many ways.
link |
It does feel like it is not an optimal language
link |
for all the things that it's being used for,
link |
or an optimal distribution system to build huge apps
link |
in something like this without type systems and so on.
link |
But I think for a lot of people,
link |
it does reasonably the necessary things.
link |
It's still a C flavored language.
link |
It's still a braces and semicolon language.
link |
It's not hard for people to be trained in JavaScript
link |
and then understand the roots of where it came from.
link |
I think garbage collection is unequivocally a good thing
link |
for most programs to be written in.
link |
It's funny that I still, just this morning,
link |
I was seeing a Twitter thread
link |
of a bunch of really senior game dev people
link |
arguing about the virtues and costs of garbage collection.
link |
You will run into some people that are top notch programmers
link |
that just say, no, this is literally not a good thing.
link |
Oh, because it makes you lazy?
link |
Yes, that it makes you not think about things.
link |
And I do disagree.
link |
I think that there is so much objective data
link |
on the vulnerabilities that have happened
link |
in C and C++ programs,
link |
sometimes written by the best programmers in the world.
link |
It's like nobody is good enough to avoid
link |
ever shooting themselves in the foot with that.
link |
You write enough C code,
link |
you're going to shoot yourself in the foot.
link |
And garbage collection is a very great thing
link |
for the vast majority of programs.
link |
It's only when you get into the tightest of real time things
link |
that you start saying, it's like, no,
link |
the garbage collection has more costs
link |
than it has benefits for me there,
link |
but that's not 99 plus percent
link |
of all the software in the world.
link |
So JavaScript is not terrible in those ways.
link |
And so much of programming is not the language itself.
link |
It's the infrastructure around that surrounds it.
link |
All the libraries that you can get
link |
and the different stuff that you can,
link |
ways you can deploy it,
link |
the portability that it gives you.
link |
And JavaScript is really strong on a lot of those things
link |
where for a long time, and it still does if I look at it,
link |
the web stack about everything that has to go
link |
when you do something really trivial in JavaScript
link |
and it shows up on a web browser
link |
to kind of X ray through that
link |
and see everything that has to happen
link |
for your one little JavaScript statement
link |
to turn into something visible in your web browser,
link |
it's very, very disquieting,
link |
just the depth of that stack
link |
and the fact that so few people can even comprehend
link |
all of the levels that are going on there.
link |
But it's again, I have to caution myself
link |
to not be the in the good old days old man about it
link |
because clearly there's enormous value here.
link |
The world does run on JavaScript
link |
to a pretty good approximation there
link |
and it's not falling apart.
link |
There's a bunch of scary stuff
link |
where you look at console logs
link |
and you just see all of these bad things that are happening
link |
but it's still kind of limping along
link |
and nobody really notices.
link |
But so much of my systems design
link |
and systems analysis goes around.
link |
You should understand what the speed of light is,
link |
like what would be the best you could possibly do here.
link |
And it sounds horrible, but in a lot of cases,
link |
you can be a thousand times off your speed of light,
link |
velocity for something and it's still be okay.
link |
And in fact, it can even sometimes still be
link |
the optimal thing in a larger system standpoint
link |
where there's a lot of things
link |
that you don't wanna have to parachute in someone like me
link |
to go in and say, make this webpage run
link |
a thousand times faster,
link |
make this web app into a hardcore native application
link |
that starts up in 37 milliseconds
link |
and everything responds in less than one frame latency.
link |
That's just not necessary.
link |
And if somebody wants to go pay me millions of dollars
link |
to do software like that,
link |
when they can take somebody right out of a bootcamp
link |
and say, spin up an application for this.
link |
Often being efficient is not really the best metric.
link |
And it's like, that applies in a lot of areas
link |
where it's kind of interesting how a lot of our appliances
link |
and everything are all built around energy efficiency,
link |
sometimes at the expense of robustness in some other ways
link |
or higher costs in other ways
link |
where there's interesting things
link |
where energy or electricity could become much cheaper
link |
in a future world.
link |
And that could change our engineering trade offs
link |
for the way we build certain things
link |
where you could throw away efficiency
link |
and actually get more benefits that actually matter.
link |
I mean, that's one of the directions
link |
I was considering swerving into was nuclear energy.
link |
When I was kind of like, what do I wanna do next?
link |
It was either gonna be cost effective nuclear fission
link |
or artificial general intelligence.
link |
And one of my pet ideas there is like,
link |
people don't understand how cheap nuclear fuel is.
link |
And there would be ways that,
link |
you could be a quarter of the efficiency or less,
link |
but if it wound up making your plant 10 times cheaper,
link |
that could be a radical innovation in something like that.
link |
So there's like some of these thoughts
link |
around like direct fission energy conversion,
link |
fission fragment conversion,
link |
that maybe you build something
link |
that doesn't require all the steam turbines and everything,
link |
even if it winds up being less efficient.
link |
So that applies a lot in programming
link |
where it's always good to know what you could do.
link |
If you really sat down and took it far,
link |
because sometimes there's discontinuities
link |
like around user reaction times,
link |
there are some points where the difference between
link |
operating in one second and 750 milliseconds,
link |
You'll see it in webpage statistics,
link |
but most of the usability stuff, not that great.
link |
But if you get down to 50 milliseconds,
link |
then all of a sudden this just feels amazing.
link |
It's just like doing your bidding instantly
link |
rather than you're giving it a command,
link |
twiddling your thumbs, waiting for it to respond.
link |
So sometimes it's important to really crunch hard
link |
to get over some threshold,
link |
but there are broad basins in the value metric
link |
for lots of work where it just doesn't pay
link |
to even go that extra mile.
link |
And there are craftsmen that just don't wanna buy that
link |
and more power to them.
link |
If somebody just wants to say,
link |
no, I'm going to be, my pride is in my work,
link |
I'm never going to do something
link |
that's not as good as I could possibly make it,
link |
I respect that and sometimes I am that person,
link |
but I try to focus more on the larger value picture
link |
and you do pick your battles
link |
and you deploy your resources in the play
link |
that's going to give you the best user value in the end.
link |
Well, if you look at the evolution of life on Earth
link |
as a kind of programming effort,
link |
it seems like efficiency isn't the thing
link |
that's being optimized for.
link |
Like natural selection is very inefficient,
link |
but it kind of adapts and through the process
link |
of adaptations building more and more complex systems
link |
that are more and more intelligent,
link |
the final result is kind of pretty interesting.
link |
And so I think of JavaScript the same way.
link |
It's like this giant mess
link |
that things naturally die off if they don't work
link |
and if they become useful to people,
link |
they kind of naturally live
link |
and then you build this community,
link |
large community of people that are generating code
link |
and some code is sticky, some is not
link |
and nobody knows the inefficiencies or the efficiencies
link |
or the breaking points, like how reliable this code is
link |
and you kind of just run it, assume it works
link |
and then get unpleasantly surprised
link |
and then that's very kind of the evolutionary process.
link |
So that's a really good analogy
link |
and we can go a lot of places with that
link |
where in the earliest days of programming,
link |
when you had finite, you could count the bytes
link |
that you had to work on this,
link |
you had all the kind of hackers playing code golf
link |
to be one less instruction
link |
than the other person's multiply routine
link |
to kind of get through and it was so perfectly crafted.
link |
It was a crystal piece of artwork when you had a program
link |
because there just were not that many,
link |
you couldn't afford to be lazy in different ways
link |
and in many ways I see that as akin to the symbolic AI work
link |
where again, if you did not have the resources to just say,
link |
well, we're gonna do billions and billions
link |
of programmable weights here,
link |
you have to turn it down into something
link |
that is symbolic and crafted like that
link |
but that's definitely not the way DNA and life
link |
and biological evolution and things work.
link |
On the one hand, it's almost humbling
link |
how little programming code is in our bodies.
link |
We've got a couple of billion base pairs
link |
and it's like this all fits on a thumb drive for years now
link |
and then our brains are even a smaller section of that.
link |
You've got maybe 50 megabytes
link |
and this is not like Shannon Limit perfectly
link |
information dense conveyances here.
link |
It's like these are messy codes,
link |
they're broken up into amino acids.
link |
A lot of them don't do important things
link |
or they do things in very awkward ways
link |
but it is this process of just accumulation
link |
on top of things and you need scale,
link |
both you need scale for the population for that to work out
link |
and in the early days in the 50s and 60s,
link |
the kind of ancient era of computers
link |
where you could count when they say like
link |
when the internet started even in the 70s,
link |
there were like 18 hosts or something on it.
link |
It was this small finite number
link |
and you were still optimizing everything
link |
to be as good as you possibly could be
link |
but now it's billions and billions of devices
link |
and everything going on and you can have
link |
this very much natural evolution going on
link |
where lots of things are tried,
link |
lots of things are blowing up,
link |
venture capitalists lose their money
link |
when a startup invested in the wrong tech stack
link |
and things completely failed or failed the scale
link |
but good things do come out of it
link |
and it's interesting to see the mimetic evolution
link |
of the way different things happen
link |
like mentioning hello world at the beginning.
link |
It's funny how some little thing like that
link |
where everybody, every programmer knows hello world now
link |
and that was a completely arbitrary sort of decision
link |
that just came out of the dominance of Unix and C
link |
and early examples of things like that.
link |
So millions of experiments are going on all the time
link |
but some things do kind of rise to the top
link |
and win the fitness war for whether it's mind space
link |
or programming techniques or anything.
link |
Like there's a site on Stack Exchange called CodeGolf
link |
where people compete to write the shortest possible program
link |
for a particular task in all the different kinds
link |
of languages and it's really interesting to see
link |
folks kind of, they're masters of their craft,
link |
really play with the limits of programming languages.
link |
It's really beautiful to see
link |
and across all the different programming languages,
link |
you get to see some of these weird programming languages
link |
and mainstream ones, difference between Python 2 and 3,
link |
you get to see the difference between C and C++ and Java
link |
and you get to see JavaScript, all of that
link |
and it's kind of inspiring to see how much depth
link |
of possibility there is within programming languages
link |
that CodeGolf kind of tasks reveal.
link |
Most of us, if you do any kind of programming,
link |
you kind of do boring kind of very vanilla type of code.
link |
That's the way to build large systems
link |
but it's nice to see that the possibility
link |
of creative genius is still within those languages.
link |
It's laden within those languages.
link |
So given that you are once again,
link |
one of the greatest programmers ever,
link |
what do you think makes a good programmer?
link |
Maybe a good modern programmer.
link |
So I just gave a long rant slash lecture at Meta
link |
to the TPM organization and my biggest point was
link |
everything that we're doing really should flow
link |
from user value, all the good things that we're doing.
link |
It's like, we're not technical people.
link |
It's like, you shouldn't be taking pride
link |
just in the specific thing.
link |
Like CodeGolf is the sort of thing, it's a fun puzzle game
link |
but that really should not be a major motivator for you.
link |
It's like, we're solving problems for people
link |
or we're providing entertainment to people.
link |
We're doing something of value to people
link |
that's displacing something else in their life.
link |
So we wanna be providing a net value
link |
over what they could be doing
link |
but instead they're choosing to use our products.
link |
And that's where, I mean, it sounds trite or corny
link |
but I fundamentally do think that's how you make
link |
the world a better place.
link |
If you have given more value to people
link |
than it took you and your team to create,
link |
then the world's a better place.
link |
People have, they've gone from something of lesser value,
link |
chosen to use your product
link |
and their life feels better for that.
link |
And if you've produced that economically,
link |
that's a really good thing.
link |
On the other hand, if you spent ridiculous amounts of money
link |
you've just kind of shoveled a lot of cash
link |
into a wood chipper there
link |
and you should maybe not feel so good
link |
about what you're doing.
link |
So being proud about like a specific architecture
link |
or a specific technology or a specific code sequence
link |
that you've done, it's great to get a little smile
link |
like a tiny little dopamine hit for that
link |
but the top level metrics should be
link |
that you're building things of value.
link |
Now you can get into the argument about how you,
link |
what is user value?
link |
How do you actually quantify that?
link |
And there can be big arguments about that
link |
but it's easy to be able to say,
link |
okay, this pissed off user there is not getting value
link |
from what you're doing.
link |
This user over there with the big smile on their face,
link |
the moment of delight when something happened,
link |
there's a value that's happened there.
link |
I mean, you have to at least accept
link |
that there is a concept of user value
link |
even if you have trouble exactly quantifying it,
link |
you can usually make relative arguments about it.
link |
Well, this was better than this, we've improved things.
link |
So being a servant to the user is your job
link |
when you're a developer, you wanna be producing something
link |
that other people are gonna find valuable.
link |
And if you are technically inclined
link |
then finding the right levers to be able to pull
link |
to be able to make a design that's going to produce
link |
the most value for the least amount of effort.
link |
And it always has to be kind of divided,
link |
there's a ratio there where you,
link |
it's a problem at the big tech companies,
link |
whether it's Meta Google, Apple, Microsoft, Amazon,
link |
companies that have almost infinite money.
link |
I mean, I know their CFO will complain
link |
that it's not infinite money
link |
but from most developers standpoints
link |
it really does feel like it.
link |
And it's almost counterintuitive
link |
that if you're working hard as a developer on something,
link |
there's always this thought,
link |
if only I had more resources, more people, more RAM,
link |
more megahertz, then my product will be better.
link |
And that sense that at certain points, it's certainly true
link |
that if you are really hamstrung by this,
link |
removing an obstacle will make a better product,
link |
But if you're not making your core design decisions
link |
in this fiercely competitive way
link |
where you're saying feature A or feature B,
link |
you can't just say, let's do both.
link |
Because then you're not making a value judgment about them.
link |
You're just saying, well, they both seem good.
link |
I don't wanna necessarily have to pick out
link |
which one is better or how much better
link |
and tell team B that, sorry, we're not gonna do this
link |
because A is more important.
link |
But that notion of always having to really critically value
link |
what you're doing, your time, the resources you expend,
link |
even the opportunity cost of doing something else,
link |
that's super important.
link |
Well, let me ask you about the big debates
link |
that you're mentioning of how to measure value.
link |
Is it possible to measure it kind of numerically
link |
or can you do the sort of Johnny Ive,
link |
the designer route of imagining
link |
sort of somebody using a thing
link |
and imagining a smile on their face,
link |
imagining the experience of love and joy
link |
that you have when you use the thing.
link |
That's from a design perspective
link |
or if you're building more like a lower level thing
link |
for like Linux, you imagine a developer
link |
that might come across this and use it
link |
and become happy and better off because of it.
link |
So where do you land on those things?
link |
So I imagine like Meta and Google
link |
will probably try to measure the thing.
link |
They'll try to, it's like you try to optimize engagement
link |
or something, let's measure engagement.
link |
And then I think there is a kind of,
link |
I mean, I admire the designer ethic of like,
link |
think of a future that's immeasurable
link |
and you try to make somebody in that future
link |
that's different from today happy.
link |
So I do usually favor,
link |
if you can get any kind of a metric that's good,
link |
by all means, listen to the data.
link |
But you can go too far there where we've had problems
link |
where it's like, hey, we had a performance regression
link |
because our fancy new telemetry system
link |
is doing a bazillion file writes
link |
to kind of archive this stuff
link |
because we needed to collect information
link |
to determine if we were doing,
link |
if our plans were good.
link |
So when information is available,
link |
you should never ignore it.
link |
I mean, all of it.
link |
For actual users using the thing,
link |
human beings using the thing,
link |
large number of human beings,
link |
and you get to see sort of a lot of large numbers.
link |
So there's the zero to one problem
link |
of when you're doing something really new,
link |
you do kind of have to make a guess.
link |
But one of the points that I've been making at Meta
link |
is we have more than enough users now
link |
that anything somebody wants to try in VR,
link |
we have users that will be interested in that.
link |
You do not get to make a completely
link |
greenfield blue sky pitch and say,
link |
I'm going to do this because I think it might be interesting.
link |
I challenge everyone.
link |
There are going to be people,
link |
whether it's working in VR on your,
link |
like on your desktop replacement
link |
or communicating with people in different ways
link |
or playing the games.
link |
There are going to be probably millions of people,
link |
or at least if you pick some tiny niche
link |
that we're not in right now,
link |
there's still going to be thousands of people out there
link |
that have the headsets that would be your target market.
link |
And I tell people, pay attention to them.
link |
Don't invent fictional users.
link |
Don't make an Alice Bob Charlie
link |
that fits whatever matrix of tendencies
link |
that you want to break the market down to,
link |
because it's a mistake to think about imaginary users
link |
when you've got real users that you could be working with.
link |
But on the other hand,
link |
there is value to having a kind of wholeness
link |
of vision for a product.
link |
And companies like Meta have,
link |
they understand the trade offs where you can have a company
link |
like SpaceX or Apple in the Steve Jobs era,
link |
where you have a very powerful leading personality
link |
that can micromanage at a very low level
link |
and can say it's like, no, that handle needs to be different
link |
or that icon needs to change the tint there.
link |
And they clearly get a lot of value out of it.
link |
They also burn through a lot of employees
link |
that have horror stories to tell
link |
about working there afterwards.
link |
My position is that you're at your best
link |
when you've got a leader that is at their limit
link |
of what they can kind of comprehend
link |
of everything below them.
link |
And they can have an informed opinion
link |
about everything that's going on.
link |
And you take somebody, you've got to believe
link |
that somebody that has 30, 40 years of experience,
link |
you would hope that they've got wisdom
link |
that the just out of bootcamp person contributing
link |
And that if they're like, well, that's wrong there,
link |
you probably shouldn't do it that way
link |
or even just don't do it that way, do it another way.
link |
So there's value there,
link |
but it can't go beyond a certain level.
link |
I mean, I have Steve Jobs stories of him saying things
link |
that are just wrong right in front of me
link |
about technical things because he was not operating
link |
But when it does work and you do get
link |
that kind of passionate leader
link |
that's thinking about the entire product
link |
and just really deeply cares
link |
about not letting anything slip through the cracks,
link |
I think that's got a lot of value.
link |
But the other side of that is the people saying that,
link |
well, we wanna have these independent teams
link |
that are bubbling up the ideas
link |
because it's almost it's anti capitalist
link |
or anti free market to say,
link |
it's like I want my grand, my great leader
link |
to go ahead and dictate all these points there
link |
where clearly free markets bring up things
link |
that you don't expect.
link |
Like in VR, we saw a bunch of things
link |
like it didn't turn out at all
link |
the way the early people thought
link |
were gonna be the key applications
link |
and things that would not have been approved
link |
by the dark cabal making the decisions
link |
about what gets into the store turn out
link |
to in some cases be extremely successful.
link |
So yeah, I definitely kind of wanted to be,
link |
there was a point where I did make a pitch.
link |
It's like, hey, make me VR dictator
link |
and I'll go in and get shit done.
link |
I am, and that's just, it's not in the culture at Meta.
link |
You know, and they understand the trade offs they,
link |
and that's just not the way,
link |
that's not the company that they want,
link |
the team that they wanna do.
link |
It's fascinating because VR and we'll talk about it more.
link |
It's still unclear to me
link |
in what way VR will change the world
link |
because it does seem clear that VR
link |
will somehow fundamentally transform this world
link |
and it's unclear to me how yet.
link |
Let me know when you wanna get into that.
link |
Well, hold on a second.
link |
So in the stick to the you being the best programmer ever.
link |
Okay, in the early days when you didn't have,
link |
when you didn't have adult responsibilities
link |
of leading teams and all that kind of stuff
link |
and you can focus on just being a programmer,
link |
what did the productive day in the life
link |
of John Carmack look like?
link |
How many hours is the keyboard?
link |
What was the source of calories that fueled the brain?
link |
What times did you wake up?
link |
So I was able to be remarkably consistent
link |
about what was good working conditions for me
link |
for a very long time.
link |
I was never one of the programmers
link |
that would do all nighters going through work
link |
for 20 hours straight.
link |
It's like my brain generally starts turning to mush
link |
after 12 hours or so.
link |
But the hard work is really important
link |
and I would work for decades.
link |
I would work 60 hours a week.
link |
I would work a 10 hour day, six days a week
link |
and try to be productive at that.
link |
Now my schedule shifted around a fair amount
link |
when I was young without any kids
link |
and any other responsibilities.
link |
I was on one of those cycling schedules
link |
where I'd kind of get in an hour later each day
link |
and roll around through the entire time
link |
and I'd wind up kind of pulling in at two or three
link |
in the afternoon sometimes and then working again
link |
past midnight or two in the morning.
link |
And that was, when it was just me trying
link |
to make things happen and I was usually isolated off
link |
in my office, people generally didn't bother me much at in
link |
and I could get a lot of programming work done that way.
link |
I did settle into a more normal schedule
link |
when I was taking kids to school and things like that.
link |
So kids were the forcing function that got you to wake up
link |
at the same time each day.
link |
It's not clear to me that there was much of a difference
link |
in the productivity with that where I kind of feel,
link |
if I just get up when I feel like it,
link |
it's usually a little later each day
link |
but I just recently made the focusing decision
link |
to try to push my schedule back a little bit earlier
link |
to getting up at eight in the morning
link |
and trying to shift things around.
link |
Like I'm often doing experiments with myself
link |
about what should I be doing to be more productive
link |
and one of the things that I did realize was happening
link |
in recent months where I would go for a walk or a run.
link |
I cover like four miles a day and I would usually do that
link |
just as the sun's going down here in Texas now
link |
and it's still really damn hot but I'd go out
link |
at 8.30 or something and cover the time there
link |
and then the showering and it was putting a hold in my day
link |
where I would have still a couple hours after that
link |
and sometimes my best hours were at night
link |
when nobody else is around, nobody's bothering me
link |
but that hole in the day was a problem
link |
so just a couple weeks ago, I made the change
link |
to go ahead and say, all right,
link |
I'm gonna get up a little earlier,
link |
I'm gonna do a walk or get out there first
link |
so I can have more uninterrupted time.
link |
So I'm still playing with factors like this
link |
as I kind of optimize my work efforts
link |
but it's always been, it was 60 hours a week
link |
for a very long time.
link |
To some degree, I had a little thing in the back of my head
link |
where I was almost jealous of some of the programmers
link |
that would do these marathon sessions
link |
and I had like Dave Taylor, one of the guys that he had,
link |
he would be one of those people
link |
that would fall asleep under his desk sometimes
link |
and all the kind of classic hacker tropes about things
link |
and a part of me was like always a little bothered
link |
that that wasn't me, that I wouldn't go program
link |
20 hours straight because I'm falling apart
link |
and not being very effective after 12 hours.
link |
I mean, yeah, 12 hour programming,
link |
that's fine when you're doing that
link |
but it never, you're not doing smart work much after,
link |
at least I'm not, but there's a range of people.
link |
I mean, that's something that a lot of people
link |
don't really get in their gut
link |
where there are people that work on four hours of sleep
link |
and are smart and can continue to do good work
link |
but then there's a lot of people that just fall apart.
link |
So I do tell people that I always try
link |
to get eight hours of sleep.
link |
It's not this, you know, push yourself harder,
link |
get up earlier, I just do worse work
link |
where, you know, you can work a hundred hours a week
link |
and still get eight hours of sleep
link |
if you just kind of prioritize things correctly
link |
but I do believe in working hard, working a lot.
link |
There was a comment that a game dev made
link |
that I know there's a backlash against really hard work
link |
in a lot of cases and I get into online arguments
link |
about this all the time but he was basically saying,
link |
yeah, 40 hours a week, that's kind of a part time job
link |
and if you were really in it,
link |
you're doing what you think is important,
link |
what you're passionate about, working more gets more done
link |
and it's just really not possible to argue with that
link |
if you've been around the people
link |
that work with that level of intensity
link |
and just say, it's like, no, they should just stop.
link |
And we had, I kind of came back around to that
link |
a couple of years ago where I was using
link |
the fictional example of, all right, some people say,
link |
they'll say with a straight face, they think, no,
link |
you are less productive if you work
link |
more than 40 hours a week.
link |
And they're generally misinterpreting things
link |
where your marginal productivity for an hour
link |
after eight hours is less than in one of your peak hours
link |
but you're not literally getting less done.
link |
There is a point where you start breaking things
link |
and getting worse behavior and everything out of it
link |
where you're literally going backwards
link |
but it's not at eight or 10 or 12 hours.
link |
And the fictional example I would use was,
link |
imagine there's an asteroid coming to impact,
link |
to crash into Earth, destroy all of human life.
link |
Do you want Elon Musk or the people working at SpaceX
link |
that are building the interceptor
link |
that's going to deflect the asteroid,
link |
do you want them to clock out at five
link |
because damn it, they're just gonna go do worse work
link |
if they work another couple hours.
link |
And it seems absurd.
link |
And that's a hypothetical though
link |
and everyone can dismiss that.
link |
But then when coronavirus was hitting
link |
and you have all of these medical personnel
link |
that are clearly pushing themselves really, really hard
link |
and I'd say, it's like, okay,
link |
do you want all of these scientists working on treatments
link |
and vaccines and caring for all of these people?
link |
Are they really screwing everything up
link |
by working more than eight hours a day?
link |
And of course people say, I'm just an asshole
link |
to say something like that, but it's the truth.
link |
Working longer gets more done.
link |
Well, so that's kind of the layer one.
link |
But I'd like to also say that,
link |
at least I believe, depending on the person,
link |
depending on the task, working more and harder
link |
will make you better for the next week
link |
in those peak hours.
link |
So there's something about a deep dedication to a thing
link |
that kind of gets deep in you.
link |
So the hard work isn't just about the raw hours
link |
of productivity, it's the thing it does to you
link |
in the weeks and months after too.
link |
You're tempering yourself in some ways.
link |
And I think, it's like Jiro dreams of sushi.
link |
If you really dedicate yourself completely
link |
to making the sushi, to really putting in the long hours,
link |
day after day after day, you become a true craftsman
link |
of the thing you're doing.
link |
Now there's, of course, discussions about
link |
are you sacrificing a lot of personal relationships?
link |
Are you sacrificing a lot of other possible things
link |
you could do with that time?
link |
But if you're talking about purely being a master
link |
or a craftsman of your art,
link |
that more hours isn't just about doing more,
link |
it's about becoming better at the thing you're doing.
link |
Yeah, and I don't gainsay anybody
link |
that wants to work the minimum amount.
link |
They've got other priorities in their life.
link |
My only argument that I'm making,
link |
it's not that everybody should work hard,
link |
it's that if you want to accomplish something,
link |
working longer and harder
link |
is the path to getting it accomplished.
link |
Well, let me ask you about this then,
link |
the mythical work life balance.
link |
For an engineer, it seems like
link |
that's one of the professions for a programmer
link |
where working hard does lead to greater productivity in it.
link |
But it also raises the question of
link |
sort of personal relationships and all that kind of stuff,
link |
family, and how are you able to find work life balance?
link |
Is there advice you can give,
link |
maybe even outside of yourself?
link |
Have you been able to arrive at any wisdom
link |
on this part in your years of life?
link |
I do think that there's a wide range of people
link |
where different people have different needs.
link |
It's not a one size fits all.
link |
I am certainly what works for me.
link |
I can tell enough that I'm different
link |
than a typical average person
link |
in the way things impact me,
link |
the things that I want to do,
link |
my goals are different,
link |
and sort of the levers to impact things are different,
link |
where I have literally never felt burnout,
link |
and I know there's lots of brilliant, smart people
link |
that do world leading work that get burned out,
link |
and it's never hit me.
link |
I've never been at a point where I'm like,
link |
I just don't care about this.
link |
I don't want to do this anymore.
link |
But I've always had the flexibility
link |
to work on lots of interesting things.
link |
I can always just turn my gaze to something else
link |
and have a great time working on that.
link |
And so much of the ability to actually work hard
link |
is the ability to have multiple things to choose from
link |
and to use your time on the most appropriate thing.
link |
There are time periods where it's the best time
link |
for me to read a new research paper
link |
that I need to really be thinking hard about it.
link |
Then there's a time that maybe I should just scan
link |
and organize my old notes
link |
because I'm just not on top of things.
link |
Then there's the time that,
link |
all right, let's go bang out a few hundred lines of code
link |
So switching between them has been real valuable.
link |
So you'll always have kind of joy in your heart
link |
for all the things you're doing,
link |
and that is a kind of work life balance
link |
as a first sort of step.
link |
Yeah, I do. So you're always happy.
link |
Yeah, I mean, that's like a lot of people would say
link |
that often I look like kind of a grim person,
link |
you know, with just sitting there with a neutral expression
link |
or even like knitted brows and a frown on my face
link |
as I'm staring at something.
link |
That's what happiness looks like for you.
link |
Yeah, it's kind of true where that it's like,
link |
okay, I'm pushing through this, I'm making progress here.
link |
I know that doesn't work for everyone.
link |
I know it doesn't work for most people.
link |
But what I am always trying to do in those cases
link |
is I don't want to let somebody
link |
that might be a person like that be told by someone else
link |
that no, don't even try that out as an option
link |
where work life balance versus kind of your life's work
link |
where there's a small subset of the people
link |
that can be very happy being obsessive about things.
link |
And, you know, obsession can often get things done
link |
that just practical, prudent, pedestrian work won't
link |
or at least won't for a very long time.
link |
There's legends of your nutritional intake
link |
in the early days.
link |
What can you say about sort of as a, you know,
link |
being a programmer as a kind of athlete?
link |
So what was the nutrition that fueled?
link |
I have never been that great on really paying attention
link |
to it where I'm good enough that I don't eat a lot.
link |
You know, I've never been like a big heavy guy,
link |
but it was interesting where one of the things
link |
that I can remember being an unhappy teenager,
link |
not having enough money and like,
link |
one of the things that bothered me
link |
about not having enough money
link |
is I couldn't buy pizza whenever I wanted to.
link |
So I got rich and then I bought a whole lot of pizza.
link |
So that was defining, like, that's what being rich felt like.
link |
A lot of the little things, like I could buy all the pizza
link |
and comic books and video games that I wanted to.
link |
And it really didn't take that much,
link |
but the pizza was one of those things
link |
that it's absolutely true that for a long time
link |
at id Software, I had a pizza delivered every single day.
link |
You know, the delivery guy knew me by name
link |
and I didn't find out until years later
link |
that apparently I was such a good customer
link |
that they just never raised the price on me.
link |
And I was using this six year old price for the pizzas
link |
that they were still kind of sending my way every day.
link |
So you were doing eating once a day or were you?
link |
It would be spread out.
link |
You know, you have a few pieces of pizza,
link |
you have some more later on
link |
and I'd maybe have something at home.
link |
It was one of the nice things that Facebook Meta
link |
is they feed you quite well.
link |
You get a different, I guess now it's DoorDash
link |
sorts of things delivered,
link |
but they take care of making sure
link |
that everybody does get well fed.
link |
And I probably had better food those six years
link |
that I was working in the Meta office there
link |
than I used to before.
link |
But it's worked out okay for me.
link |
My health has always been good.
link |
I get a pretty good amount of exercise
link |
and I don't eat to excess
link |
and I avoid a lot of other kind of not so good
link |
So I'm still doing quite well at my age.
link |
Did you have a kind of, I don't know,
link |
spiritual experience with food or coffee
link |
or any of that kind of stuff?
link |
I mean, you know, the programming experience,
link |
you know, with music or like I listen to brown noise
link |
on a program or like creating an environment
link |
and the things you take into your body,
link |
just everything you construct
link |
can become a kind of ritual
link |
that empowers the whole process, the programming.
link |
Did you have that relationship with pizza or?
link |
It would really be with Diet Coke.
link |
I mean, there still is that sense of, you know,
link |
drop the can down, crack open the can of Diet Coke.
link |
All right, now I mean business.
link |
We're getting to work here.
link |
Still to this day?
link |
Is Diet Coke still a part of it?
link |
Yeah, probably eight or nine a day.
link |
Okay, what about your setup?
link |
What kind of keyboard?
link |
Is there something interesting?
link |
What kind of IDE, Emacs Vim or something modern?
link |
Linux, what operating system laptop
link |
or any interesting thing that brings you joy?
link |
So I kind of migrated cultures
link |
where early on through all of game dev,
link |
there was sort of one culture there,
link |
which was really quite distinct
link |
from the more the Silicon Valley venture,
link |
you know, culture for things.
link |
They're different groups
link |
and they have pretty different mores
link |
and the way they think about things where,
link |
and I still do think a lot of the big companies
link |
can learn things from the hardcore game development
link |
side of things where it still boggles my mind
link |
how hostile to debuggers and IDEs that so much of them,
link |
the kind of big money, get billions of dollars,
link |
Silicon Valley venture backed funds are.
link |
Oh, that's interesting.
link |
Sorry, so you're saying like big companies
link |
like Google and Meta are hostile to?
link |
They are not big on debuggers and IDEs.
link |
Like so much of it is like Emacs, Vim for things.
link |
And we just assume that debuggers don't work
link |
most of the time for the systems.
link |
And a lot of this comes from a sort of Linux bias
link |
on a lot of things where I did come up
link |
through the personal computers and then the DOS
link |
and then Windows and it was Borland tools
link |
and then Visual Studio and.
link |
Do you appreciate the buggers?
link |
I mean, a debugger is how you get a view into a system
link |
that's too complicated to understand.
link |
I mean, anybody that thinks just read the code
link |
and think about it.
link |
That's an insane statement in the,
link |
you can't even read all the code on a big system.
link |
You have to do experiments on the system
link |
and doing that by adding log statements,
link |
recompiling and rerunning it
link |
is an incredibly inefficient way of doing it.
link |
I mean, yes, you can always get things done
link |
even if you're working with stone knives and bear skins.
link |
That is the mark of a good programmer
link |
is that given any tools,
link |
you will figure out a way to get it done,
link |
but it's amazing what you can do
link |
with sometimes much, much better tools
link |
where instead of just going through this iterative
link |
compile run debug cycle,
link |
you have the old Lisp direction of like,
link |
you've got a REPL and you're working interactively
link |
and doing amazing things there.
link |
But in many cases, a debugger
link |
as a very powerful user interface that can stop,
link |
examine all the different things in your program,
link |
set all of these different breakpoints.
link |
And of course you can do that with GDB or whatever there,
link |
but this is one of the user interface
link |
fundamental principles
link |
where when something is complicated to do,
link |
you won't use it very often.
link |
There's people that will break out GDB
link |
when they're at their wits end
link |
and they just have beat their head
link |
against a problem for so long.
link |
But for somebody that kind of grew up in game dev,
link |
it's like they were running into the debugger anyways
link |
before they even knew there was a problem.
link |
And you would just stop and see what was happening.
link |
And sometimes you could fix things
link |
even before you did one compile cycle.
link |
You could be in the debugger and you would say,
link |
well, I'm just going to change this right here.
link |
And yep, that did the job and fix it and go on.
link |
And for people who don't know,
link |
GDB is a sort of popular, I guess, Linux debugger
link |
primarily for C++.
link |
They handle most of the languages,
link |
but it's based on C as the original kind of Unix heritage.
link |
And it's kind of like command line.
link |
It's not user friendly.
link |
It doesn't allow for clean visualizations
link |
and you're exactly right.
link |
So that you're using this kind of debugger
link |
usually when you're at what's end
link |
and there's a problem that you can't figure out why
link |
by just looking at the codes, you have to find it.
link |
That's how I guess normal programmers use it.
link |
But you're saying there should be tools
link |
that kind of visualize and help you
link |
as part of the programming process,
link |
just the normal programming process
link |
to understand the code deeper.
link |
Yeah, when I'm working on like my C, C++ code,
link |
I'm always running it from the debugger.
link |
Just I type in the code, I run it many times.
link |
The first thing I do after writing code
link |
is set a break point and step through the function.
link |
Now other people will say, it's like,
link |
oh, I do that in my head.
link |
Well, your head is a faulty interpreter
link |
of all those things there.
link |
And I've written brand new code.
link |
I wanna step in there
link |
and I'm gonna single step through that,
link |
examine lots of things
link |
and see if it's actually doing what I expected it to.
link |
It is a kind of companion, the debugger.
link |
Like you're now coding in an interactive way
link |
with another being.
link |
The debugger is a kind of dumb being,
link |
but it's a reliable being.
link |
That is an interesting question
link |
of what role does AI play in that kind of,
link |
with codecs and these kind of ability to generate code,
link |
that might be, you might start having tools
link |
that understand the code in interesting deep ways
link |
that can work with you.
link |
There's a whole spectrum there
link |
from static code analyzers and various kind of dynamic tools
link |
there up to AI that can conceivably grok these programs
link |
that literally no human can understand.
link |
They're too big, too intertwined and too interconnected,
link |
but it's not beyond the possibility of understanding.
link |
It's just beyond what we can hold in our heads
link |
as kind of mutable state while we're working on things.
link |
And I'm a big proponent, again,
link |
of things like static analyzers and some of that stuff
link |
where you'll find some people that don't like being scolded
link |
by a program for how they've written something
link |
where it's like, oh, I know better.
link |
And sometimes you do, but that was something
link |
that it was very, very valuable for me
link |
when, and not too many people get an opportunity
link |
like this to have.
link |
This is almost one of those spiritual experiences
link |
as a programmer, an awakening to,
link |
the id software code bases
link |
were a couple of million lines of code.
link |
And at one point I had used a few
link |
of the different analysis tools,
link |
but I made a point to really go through
link |
and scrub the code base using every tool that I could find.
link |
And it was eyeopening where we had a reputation
link |
for having some of the most robust, strongest code,
link |
where there were some great things
link |
that I remember hearing from Microsoft
link |
telling us about crashes on Xbox.
link |
And we had this tiny number that they said
link |
were probably literally hardware errors.
link |
And then you have other significant titles
link |
that just have millions of faults
link |
that are getting recorded all the time.
link |
So I was proud of our code on a lot of levels,
link |
but when I took this code analysis squeegee
link |
through everything, it was shocking
link |
how many errors there were in there.
link |
Things that you can say, okay, this was a copy paste,
link |
not changing something right here.
link |
Lots of things that were, the most common problem
link |
was something in a printf format string
link |
that was the wrong data type that could cause crashes there.
link |
And you really want the warnings for things like that.
link |
Then the next most common was missing a check for null
link |
that could actually happen, that could blow things up.
link |
And those are obviously like top C, C++ things.
link |
Everybody has those problems.
link |
But the long tail of all of the different little things
link |
that could go wrong there, and we had good programmers
link |
and my own code, stuff that I'd be looking at,
link |
it's like, oh, I wrote that code, that's definitely wrong.
link |
We've been using this for a year
link |
and it's this submarine, this mine sitting there
link |
waiting for us to step on.
link |
And it was humbling.
link |
It was, and I reached the conclusion
link |
that anything that can be syntactically allowed
link |
in your language, if it's gonna show up eventually
link |
in a large enough code base,
link |
you're not gonna, good intentions
link |
aren't going to keep it from happening.
link |
You need automated tools and guardrails for things.
link |
And those start with things like static types
link |
and or even type hints in the more dynamic languages.
link |
But the people that rebel against that,
link |
that basically say that slows me down doing that.
link |
There's something to that, I get that.
link |
I've written, I've cobbled things together in a notebook.
link |
I'm like, wow, this is great that it just happened,
link |
but yeah, that's kind of sketchy,
link |
but it's working fine, I don't care.
link |
It does come back to that value analysis
link |
where sometimes it's right to not care.
link |
But when you do care, if it's going to be something
link |
that's going to live for years
link |
and it's gonna have other people working on it,
link |
and it's gonna be deployed to millions of people,
link |
then you wanna use all of these tools.
link |
You wanna be told it's like, no,
link |
you've screwed up here, here and here.
link |
And that does require kind of an ego check about things
link |
where you have to be open to the fact
link |
that everything that you're doing
link |
is just littered with flaws.
link |
It's not that, oh, you occasionally have a bad day.
link |
It's just whatever stream of code you output,
link |
there is going to be a statistical regularity of things
link |
that you just make mistakes on.
link |
And I do think there's the whole argument
link |
about test driven design and unit testing
link |
versus kind of analysis and different things.
link |
I am more in favor of the analysis and the stuff
link |
that just like you can't run your program
link |
until you fix this rather than you can run it
link |
and hopefully a unit test will catch it in some way.
link |
Yeah, in my private code, I have asserts everywhere.
link |
There's something pleasant to me, pleasurable to me
link |
about sort of the dictatorial rule of like,
link |
this should be true at this point.
link |
And too many times I've made mistakes
link |
that shouldn't have been made.
link |
And I would assume I wouldn't be the kind of person
link |
that would make that mistake,
link |
but I keep making that mistake.
link |
Therefore, an assert really catches me,
link |
really helps all the time.
link |
So my code, I would say like 10 to 20% of my private code
link |
just for personal use is probably asserts.
link |
And they're active comments.
link |
That's one of those things that in theory,
link |
they don't make any difference to the program.
link |
And if it was all operating the way you expected
link |
it would be, then they will never fire.
link |
But even if you have it right
link |
and you wrote the code right initially,
link |
then circumstances change.
link |
The world outside your program changes.
link |
And in fact, that's one of the things
link |
where I'm kind of fond in a lot of cases
link |
of static array size declarations,
link |
where I went through this period where it's like,
link |
okay, now we have general collection classes.
link |
We should just make everything variable
link |
because I had this history of in the early days,
link |
you get doom, which had some fixed limits on it.
link |
Then everybody started making crazier and crazier things.
link |
And they kept bumping up the different limits,
link |
this many lines, this many sectors.
link |
And it seemed like a good idea,
link |
well, we should just make this completely generic.
link |
It can go kind of go up to whatever.
link |
And there's cases where that's the right thing to do,
link |
but it also, the other aspect of the world
link |
changing around you is it's good to be informed
link |
when the world has changed more than you thought it would.
link |
And if you've got a continuously growing collection,
link |
you're never gonna find out.
link |
You might have this quadratic slow down on something
link |
where you thought, oh, I'm only ever gonna have
link |
a handful of these, but something changes
link |
and there's a new design style.
link |
And all of a sudden you've got 10,000 of them.
link |
So I kind of like in many cases, picking a number,
link |
some nice brown power of two number
link |
and setting it up in there and having an assert saying,
link |
it's like, hey, you hit this limit.
link |
You should probably think are the choices
link |
that you've made around all of this still relevant
link |
if somebody's using 10 times more
link |
than you thought they would.
link |
Yeah, this code was originally written
link |
with this kind of worldview,
link |
with this kind of set of constraints.
link |
You were thinking of the world in this way.
link |
If something breaks, that means you gotta rethink
link |
the initial stuff.
link |
And it's nice for it to do that.
link |
Is there any stuff like a keyboard or monitors?
link |
I'm a fairly pedestrian on a lot of that
link |
where I did move to triple monitors
link |
like in the last several years ago.
link |
I had been dual monitor for a very long time.
link |
And it was one of those things where probably years later
link |
than I should have, I'm just like, well,
link |
the video cards now generally have three output ports.
link |
I should just put the third monitor up there.
link |
That's been a pure win.
link |
I've been very happy with that.
link |
But no, I don't have fancy keyboard or mouse
link |
or anything really going on.
link |
The key things is an IDE that has helpful debuggers,
link |
has helpful tools.
link |
So it's not the Emacs Vim route and then Dicode.
link |
Yeah, so I did spend,
link |
I spent one of my week long retreats where I'm like,
link |
okay, I'm gonna make myself use,
link |
it was actually classic VI,
link |
which I know people will say you should never have done that.
link |
You should have just used Vim directly.
link |
But I gave it the good try.
link |
It's like, okay, I'm being in kind of classic
link |
Unix developer mode here.
link |
And I worked for a week on it.
link |
I used Anki to like teach myself
link |
the different little key combinations
link |
for things like that.
link |
And in the end, it was just like, all right,
link |
this was kind of like my civil war reenactment phase.
link |
It's like, I'm going out there doing it
link |
like they used to in the old days.
link |
And it was kind of fun in that regard.
link |
So many people right now.
link |
They're screaming as they're listening to this.
link |
So again, the out is that this was not modern Vim,
link |
but still, yes, I was very happy to get back
link |
to my visual studio at the end.
link |
Yeah, I'm actually, I struggle with this a lot
link |
because I use a Kinesis keyboard
link |
and I use Emacs primarily.
link |
And I feel like I can, exactly as you said,
link |
I can understand the code.
link |
I can navigate the code.
link |
There's a lot of stuff you could build with an Emacs
link |
You can customize a lot of things for yourself
link |
to help you introspect the code,
link |
like to help you understand the code
link |
and visualize different aspects of the code.
link |
You can even run debuggers,
link |
but it's work and the world moves past you
link |
and the better and better ideas are constantly being built.
link |
And that puts a kind of,
link |
I need to take the same kind of retreat
link |
as you're talking about,
link |
but now I'm still fighting the civil war.
link |
I need to kind of move into the 21st century.
link |
And it does seem like the world is,
link |
or a large chunk of the world is moving
link |
towards visual studio code,
link |
which is kind of interesting to me.
link |
Again, it's the JavaScript ecosystem on the one hand
link |
and IDs are one of those things
link |
that you want to be infinitely fast.
link |
You want them to just kind of immediately respond.
link |
And like, I mean, heck, I've got,
link |
there's someone I know,
link |
I'm an old school game dev guy
link |
that still uses visual studio six.
link |
And on a modern computer,
link |
everything is just absolutely instant
link |
on something like that,
link |
because it was made to work on a computer
link |
that's 10,000 or a hundred thousand times slower.
link |
So just everything happens immediately.
link |
And all the modern systems just feel,
link |
they feel so crufty when it's like,
link |
oh, why is this refreshing the screen
link |
and moving around and updating over here?
link |
And something blinks down there
link |
and you should update this.
link |
And there are things that we've lost
link |
with that incredible flexibility,
link |
but lots of people get tons of value from it.
link |
And I am super happy that that seems to be winning
link |
over even a lot of the old Vim and Emacs people
link |
that they're kind of like,
link |
hey, visual studio codes, maybe not so bad.
link |
I am, that may be the final peacekeeping solution
link |
where everybody is reasonably happy
link |
with something like that.
link |
So can you explain what a dot plan file is
link |
and what role that played in your life?
link |
Does it still continue to play a role?
link |
Back in the early, early days of id Software,
link |
one of our big things that was unique with what we did
link |
is I had adopted next stations
link |
or kind of next step systems
link |
from Steve Jobs's out in the woods away from Apple company.
link |
And they were basically, it was kind of interesting
link |
because I did not really have a background
link |
with the Unix system.
link |
So many of the people they get immersed in that in college
link |
and that sets a lot of cultural expectations for them.
link |
And I didn't have any of that,
link |
but I knew that my background was,
link |
I was a huge Apple II fan boy.
link |
I was always a little suspicious of the Mac.
link |
I was not really what kind of I wanted to go with.
link |
But when Steve Jobs left Apple and started Next,
link |
this computer did just seem like one of those amazing things
link |
from the future where it had all of this cool stuff in it.
link |
And we were still back in those days working on DOS,
link |
everything blew up.
link |
You had reset buttons
link |
because your computer would just freeze.
link |
If you're doing development work,
link |
literally dozens of times a day,
link |
your computer was just rebooting constantly.
link |
And so this idea of, yes, any of the Unix workstations
link |
would have given a stable development platform
link |
where you don't crash and reboot all the time.
link |
But Next also had this really amazing graphical interface
link |
and it was great for building tools.
link |
And it used Objective C as kind of an interesting dead end
link |
for things like that.
link |
So Next was Unix based, it said Objective C.
link |
So it has a lot of the elements.
link |
I mean, the kind of reverse acquisition of Apple by Next
link |
where that took over
link |
and became what the modern Mac system is.
link |
And to find some of the developer,
link |
like the tools and the whole community.
link |
Yeah, you've still got,
link |
if you're programming on Apple stuff now,
link |
there's still all these NS somethings,
link |
which was originally Next step objects
link |
of different kinds of things.
link |
But one of the aspects of those Unix systems
link |
was they had this notion of a.plan file
link |
where a.file is an invisible file.
link |
Usually in your home directory or something.
link |
And there was a trivial server running
link |
on most Unix systems at the time
link |
that when somebody ran a trivial little command
link |
called finger, you could do finger
link |
and then somebody's address,
link |
it could be anywhere on the internet
link |
if you were connected correctly.
link |
Then all that server would do was read the.plan file
link |
in that user's home directory
link |
and then just spit it out to you.
link |
And originally the idea was that could be
link |
whether you're on vacation, what your current project was.
link |
It's supposed to be like the plan of what you're doing.
link |
And people would use it for various purposes,
link |
but all it did was dump that file over
link |
to the terminal of whoever issued the finger command.
link |
And at one point I started just keeping a list
link |
of what I was doing in there,
link |
which would be what I was working on in the day.
link |
And I would have this little syntax.
link |
I kind of got to myself about,
link |
here's something that I'm working on.
link |
I put a star when I finish it.
link |
I could have a few other little bits of punctuation.
link |
And at the time it was,
link |
it started off as being just like my to do list.
link |
And it would be these trivial, obscure little things
link |
like I fixed something with collision detection code,
link |
made Fireball do something different
link |
and just little one liners that people
link |
that were following the games could kind of decipher.
link |
But I did wind up starting to write much more in depth things.
link |
I would have little notes of thoughts and insights
link |
and then I would eventually start having little essays.
link |
I would sometimes dump into the dot plan files
link |
interspersed with the work logs of things that I was doing.
link |
So in some ways it was like a super early proto blog
link |
where I was just kind of dumping out what I was working on.
link |
But it was interesting enough that there were a lot
link |
of people that were interested in this.
link |
So most of the people didn't have Unix workstation.
link |
So there were the websites back in the day
link |
that would follow the Doom and Quake development
link |
that would basically make a little service
link |
that would go grab all the changes
link |
and then people could just get it with a web browser.
link |
And there was a period where like all of the little
link |
kind of Dallas gaming diaspora of people
link |
that were at all in that orbit,
link |
there were a couple dozen plan files going on,
link |
which was, and this was some years before blogging
link |
really became kind of a thing.
link |
And it was kind of a premonition
link |
of sort of the way things would go.
link |
And there was, it's all been collected.
link |
It's available online in different places.
link |
And it's kind of fun to go back and look through
link |
what I was thinking, what I was doing in the different areas.
link |
Have you had a chance to look back?
link |
Is there some interesting, very low level specific
link |
to do items, maybe things you've never completed,
link |
all that kind of stuff and high level philosophical essay
link |
type of stuff that stands out?
link |
Yeah, there's some good stuff on both
link |
where a lot of it was low level nitpicky details
link |
about game dev and I've learned enough things
link |
where there's no project that I worked on
link |
that I couldn't go back and do a better job on now.
link |
I mean, you just, you learn things,
link |
hopefully if you're doing it right,
link |
you learn things as you get older
link |
and you should be able to do a better job
link |
at all of the early things.
link |
And there's stuff in Wolfenstein Doomquake that's like,
link |
oh, clearly I could go back and do a better job at this,
link |
whether it's something in the rendering engine side
link |
or how I implemented the monster behaviors
link |
or managed resources or anything like that.
link |
Do you see the flaws in your thinking now?
link |
Like looking back?
link |
I mean, sometimes I'll get the,
link |
I'll look at it and say, yeah, I had a pretty clear view
link |
of I was doing good work there.
link |
And I haven't really hit the point
link |
where there was another programmer, Graham Devine,
link |
who was, he had worked at ID and Seventh Guest
link |
and he made some comment one time
link |
where he said he looked back at some of his old notes
link |
and he was like, wow, I was really smart back then.
link |
And I don't hit that so much where,
link |
I mean, I look at it and I always know that,
link |
yeah, there's all the, you know, with aging,
link |
you get certain changes in how you're able to work problems,
link |
but all of the problems that I've worked,
link |
I'm sure that I could do a better job on all of them.
link |
So you can still step right in.
link |
If you could travel back in time and talk to that guy,
link |
you would teach him a few things.
link |
What about the high level philosophical stuff?
link |
Is there some insights that stand out that you remember?
link |
There's things that I was understanding about development
link |
and, you know, in the industry and so on
link |
that were in a more primitive stage
link |
where I definitely learned a lot more in the later years
link |
about business and organization and team structure.
link |
There were, I mean, there were definitely things
link |
that I was not the best person
link |
or even a very good person about managing,
link |
like how a team should operate internally,
link |
how people should work together.
link |
I was just, you know, just get out of my way
link |
and let me work on the code and do this.
link |
And more and more, I've learned how,
link |
in the larger scheme of things,
link |
how sometimes relatively unimportant
link |
some of those things are,
link |
where it is this user value generation
link |
that's the overarching importance for all of that.
link |
And I didn't necessarily have my eye on that ball correctly
link |
through a lot of my, you know, my earlier years.
link |
And there's things that, you know,
link |
I could have gotten more out of people
link |
handling things in different ways.
link |
I am, you know, I could have made, you know,
link |
in some ways more successful products
link |
by following things in different ways.
link |
There's mistakes that we made
link |
that we couldn't really have known
link |
how things would have worked out,
link |
but it was interesting to see in later years companies
link |
like Activision showing that,
link |
hey, you really can just do the same game,
link |
make it better every year.
link |
And you can look at that from a negative standpoint
link |
and say, it's like, oh, that's just being derivative
link |
But if you step back again and say, it's like, no,
link |
are the people buying it still enjoying it?
link |
Are they enjoying it more
link |
than what they might have bought otherwise?
link |
And you can say, no, that's actually
link |
a great value creation engine to do that
link |
if you're in a position where you can.
link |
I, you know, don't be forced into reinventing everything
link |
just because you think that you need to.
link |
You know, lots of things about business and team stuff
link |
that could be done better, but the technical work,
link |
the kind of technical visionary type stuff
link |
that I laid out, I still feel pretty good about.
link |
There are some classic old ones about my defending
link |
of OpenGL versus D3D,
link |
which turned out to be one of the more
link |
probably important momentous things there
link |
where it never, it was always a rear guard action on Windows
link |
where Microsoft was just not gonna let that win.
link |
But when I look back on it now,
link |
that fight to keep OpenGL relevant
link |
for a number of years there
link |
meant that OpenGL was there when mobile started happening
link |
and OpenGL ES was the thing that drove
link |
all of the acceleration of the mobile industry.
link |
And it's really only in the last few years
link |
as Apple's moved to Metal
link |
and some of the other companies have moved to Vulkan
link |
that that's moved away.
link |
But really stepping back and looking at it,
link |
it's like, yeah, I sold tens of millions of games
link |
for different things.
link |
But billions and billions of devices wound up
link |
with an appropriate capable graphics API
link |
due in no small part to me thinking
link |
that that was really important
link |
that we not just give up and use Microsoft's
link |
at that time really terrible API.
link |
The thing about Microsoft is the APIs don't stay terrible.
link |
They were terrible at the start, but a few versions on,
link |
they were actually quite good.
link |
And there was a completely fair argument to be made
link |
that by the time DX9 was out,
link |
it was probably a better programming environment
link |
than OpenGL, but it was still a wonderful good thing
link |
that we had an open standard
link |
that could show up on Linux and Android and iOS
link |
and eventually WebGL still to this day.
link |
So that was one, that would be on my greatest hits list
link |
of things that I kind of pushed with.
link |
The impact it had on billions of devices, yes.
link |
So let's talk about it.
link |
Can you tell the origin story of id Software?
link |
Again, one of the greatest game developer companies ever.
link |
It created Wolfenstein 3D games
link |
that defined my life also in many ways
link |
as a thing that made me realize
link |
what computers are capable of in terms of graphics,
link |
in terms of performance, it just unlocked something deep
link |
in me and understanding what these machines are all about
link |
as games can do that.
link |
So Wolfenstein 3D, Doom, Quake,
link |
and just all the incredible engineering innovation
link |
that went into that.
link |
So how did it all start?
link |
So I'll caveat upfront
link |
that I usually don't consider myself the historian
link |
of the software side of things.
link |
I usually do kind of point people at John Romero
link |
for stories about the early days where I've never been,
link |
like I've commented that I'm a remarkably
link |
unsentimental person in some ways
link |
where I don't really spend a lot of time
link |
unless I'm explicitly prodded to go back
link |
and think about the early days of things.
link |
And I didn't necessarily make the effort
link |
to archive everything exactly in my brain.
link |
And the more that I work on machine learning and AI
link |
and the aspects of memory
link |
and how when you go back and polish certain things,
link |
it's not necessarily exactly the way it happened.
link |
But having said all of that, from my view,
link |
the way everything happened that led up to that was
link |
after I was an adult
link |
and kind of taking a few college classes
link |
and deciding to drop out, I was doing,
link |
I was hardscrabble contract programming work,
link |
really struggling to kind of keep groceries
link |
and pay my rent and things.
link |
And the company that I was doing the most work for
link |
was a company called Softdisk Publishing,
link |
which had the sounds bizarre now business model
link |
of monthly subscription software.
link |
Before there was an internet that people could connect to
link |
and get software, you would pay a certain amount
link |
and every month they would send you a disc
link |
that had some random software on it.
link |
And people that were into computers
link |
thought this was kind of cool.
link |
And they had different ones for the Apple II,
link |
the 2GS, the PC, the Mac, the Amiga,
link |
lots of different things here.
link |
So quirky little business,
link |
but I was doing a lot of contract programming for them
link |
where I'd write tiny little games
link |
and sell them for 300, $500.
link |
And one of the things that I was doing again
link |
to keep my head above water here was
link |
I decided that I could make one program
link |
and I could port it to multiple systems.
link |
So I would write a game like Dark Designs or Catacombs
link |
and I would develop it on the Apple II, the 2GS
link |
and the IBM PC, which apparently was the thing
link |
that really kind of piqued the attention
link |
of the people working down there.
link |
Like Jay Wilbur was my primary editor
link |
and Tom Hall was a secondary editor.
link |
And they kept asking me, it's like,
link |
hey, you should come down and work for us here.
link |
And I pushed it off a couple of times
link |
because I was really enjoying my freedom
link |
of kind of being off on my own.
link |
Even if I was barely getting by, I loved it.
link |
I was doing nothing but programming all day,
link |
but I did have enough close scrapes with like,
link |
damn, I'm just really out of money
link |
that maybe I should get an actual job
link |
rather than contracting these kind of one at a time things.
link |
And Jay Wilbur was great.
link |
He was like FedExing me the checks when I would need them.
link |
I had to kind of get over whatever hump I was at.
link |
So I took the, I finally took them up on their offer
link |
to come down to Shreveport, Louisiana.
link |
I was in Kansas city at the time,
link |
drove down to through the Ozarks and everything
link |
down to Louisiana and saw the soft disk offices,
link |
went through, talked to a bunch of people,
link |
met the people I had been working with remotely
link |
But the most important thing for me
link |
was I met two programmers there,
link |
John Romero and Lane Roth that for the first time ever,
link |
I had met programmers that knew more cool stuff than I did
link |
where the world was just different back then.
link |
I was in Kansas city.
link |
It was one of those smartest kid in the school
link |
does all the computer stuff.
link |
The teachers don't have anything to teach him.
link |
But all I had to learn from was these few books
link |
It was not much at all.
link |
And there were some aspects of programming
link |
that were kind of black magic to me.
link |
Like, it's like, oh, he knows how to format a track on,
link |
I am on a low level drive programming interface.
link |
And this was, I was still not at all sure
link |
I was gonna take the job,
link |
but I met these awesome programmers
link |
that were doing cool stuff.
link |
And Romero had worked at Origin Systems
link |
and he had done like, so many different games ahead of time
link |
that I did kind of quickly decide,
link |
it's like, yeah, I'll go take the job down there.
link |
And I settled down there, moved in
link |
and started working on more little projects.
link |
And the first kind of big change that happened down there
link |
was the company wanted to make a gaming focused,
link |
a PC gaming focused subscription.
link |
Just like all their other,
link |
it was the same formula that they used for everything.
link |
Pay a monthly fee and you'll get a disc
link |
with one or two games just every month
link |
and no choice in what you get, but we think it'll be fun.
link |
And that was the model they were comfortable with
link |
and they said, all right,
link |
we're gonna start this gamers edge department.
link |
And all of us that were interested in that,
link |
like me, Romero, Tom Hall was kind of helping us
link |
from his side of things, Jay would peek in
link |
and we had a few other programmers working with us
link |
And we were going to just start making games,
link |
just the same model.
link |
And we dived in and it was fantastic.
link |
Do you have to make new games?
link |
And this, in retrospect, looking back at it,
link |
that sense that I had done all this contract programming
link |
and John Romero had done like far more of this
link |
where he had done one of his teaching himself efforts
link |
was he made a game for every letter of the alphabet.
link |
It's that sense of like,
link |
I'm just gonna go make 26 different games,
link |
give them a different theme.
link |
And you learn so much when you go through
link |
and you crank these things out
link |
like on a biweekly, monthly basis,
link |
something like that. From start to finish.
link |
It's not like just an idea.
link |
It's not just from the very beginning to the very end.
link |
It has to be done.
link |
There's no delaying.
link |
And you've got deadlines.
link |
And that kind of rapid iteration pressure cooker environment
link |
was super important for all of us developing the skills
link |
that brought us to where we eventually went to.
link |
I mean, people would say like,
link |
like in the history of the Beatles,
link |
like it wasn't them being the Beatles.
link |
It was them playing all of these other early works
link |
that that opportunity to craft all of their skills
link |
before they were famous,
link |
that was very critical to their later successes.
link |
And I think there's a lot of that here
link |
where we did these games that nobody remembers,
link |
lots of little things that contributed
link |
to building up the skillset for the things
link |
that eventually did make us famous.
link |
Yeah, Dostoevsky wrote The Gambler.
link |
I had to write it in a month, just to make money.
link |
And nobody remembers that probably
link |
because he had to figure out,
link |
because it's literally,
link |
he didn't have enough time to write it fast enough.
link |
So he had to come up with hacks.
link |
it again comes down to that point
link |
where pressure and limitation of resources
link |
is surprisingly important.
link |
And it's counterintuitive in a lot of ways
link |
where you just think that if you've got all the time
link |
in the world and you've got all the resources in the world,
link |
of course, you're gonna get something better.
link |
But sometimes it really does work out
link |
that the innovations, mother necessity,
link |
where you can, or resource constraints,
link |
and you have to do things when you don't have a choice.
link |
It's surprising what you can do.
link |
Is there any good games written in that time?
link |
Some of them are still fun to go back and play
link |
where you get the,
link |
they were all about kind of,
link |
the more modern term is game feel,
link |
about how just the exact feel that things,
link |
it's not the grand strategy of the design,
link |
but how running and jumping and shooting
link |
and those things I feel in the moment.
link |
And some of those are still,
link |
if you sat down on them, you kind of go,
link |
it's a little bit different.
link |
It doesn't have the same movement feel,
link |
but you move over and you're like, bang, jump, bang.
link |
It's like, hey, that's kind of cool still.
link |
So you can get lost in the rhythm of the game.
link |
Is that what you mean by feel?
link |
Just like there's something about it that pulls you in?
link |
Nowadays, again, people talk about compulsion loops
link |
and things where it's that sense of exactly
link |
what you're doing, what your fingers are doing
link |
on the keyboard, what your eyes are seeing,
link |
and there are gonna be these sequences of things.
link |
Grab the loot, shoot the monster,
link |
jump over the obstacle, get to the end of the level.
link |
These are eternal aspects of game design in a lot of ways,
link |
but there are better and worse ways to do all of them.
link |
And we did so many of these games
link |
that we got a lot of practice with it.
link |
So one of the kind of weird things
link |
that was happening at this time
link |
is John Romero was getting some strange fan mail.
link |
And back in the days, this was before email.
link |
So we literally got letters sometimes and telling him,
link |
it's like, oh, I wanna talk to you about your games.
link |
I wanna reach out, different things.
link |
And eventually it turned out
link |
that these were all coming from Scott Miller
link |
at Apogee Software.
link |
And he was reaching out through,
link |
he didn't think he could contact John directly,
link |
that he would get intercepted.
link |
So he was trying to get him to contact him
link |
through like back channel fan mail,
link |
because he basically was saying,
link |
hey, I'm making all this money on shareware games.
link |
I want you to make shareware games,
link |
because he had seen some of the games
link |
that Romero had done.
link |
And we looked at Scott Miller's games
link |
and we didn't think they were very good.
link |
We're like, that can't be making the kind of money
link |
that he's saying he's making 10 grand or something
link |
And we really thought that he was full of shit,
link |
that it was a lie trying to get him into this.
link |
But so that was kind of going on at one level.
link |
And it was funny the moment when Romero realized
link |
that he had some of these letters pinned up on his wall,
link |
like all of his fans.
link |
And then we noticed that they all had the same return
link |
address with different names on them,
link |
which was a little bit of a two edged sword there.
link |
Trying to figure out the puzzle laid out before him.
link |
Yeah, what happened after I kind of coincident with that
link |
was I was working on a lot of the new technologies
link |
where I was now full on the IBM PC for the first time,
link |
where I was really a long hold out on Apple II forever.
link |
And I loved my Apple II.
link |
It was the computer I always wished I had
link |
when I was growing up.
link |
And when I finally did have one,
link |
I was kind of clinging on to that well past it
link |
sort of good use by day.
link |
Is it the best computer ever made you would you say?
link |
I wouldn't make judgments like that about it,
link |
but it was positioned in such a way,
link |
especially in the school systems that it impacted
link |
a whole lot of American programmers at least,
link |
where there was programs that the Apple IIs
link |
got into the schools and they had enough capability
link |
that lots of interesting things happened with them.
link |
In Europe, it was different.
link |
You had your Amigas and Ataris and Acorns in the UK
link |
and things that had different things.
link |
But in the United States,
link |
it was probably the Apple II made the most impact
link |
for a lot of programmers of my generation.
link |
But so I was really digging into the IBM
link |
and this was even more so with the total focus
link |
because I had moved to another city
link |
where I didn't know anybody that I wasn't working with.
link |
I had a little apartment and then at Softdisk,
link |
again, the things that drew me to it,
link |
I had a couple of programmers that knew more than I did
link |
and they had a library.
link |
They had a set of books and a set of magazines.
link |
They had a couple of years of magazines,
link |
the old Dr. Dobbs journal and all of these magazines
link |
that had information about things.
link |
And so I was just in total immersion mode.
link |
It was eat, breathe, sleep, computer programming,
link |
particularly the IBM for everything that I was doing.
link |
And I was digging into a lot of these
link |
low level hardware details
link |
that people weren't usually paying attention to,
link |
the way the IBM EGA cards worked, which was fun for me.
link |
I hadn't had experience with things at that level.
link |
And back then you could get hardware documentation
link |
just down at the register levels.
link |
This is where the CRTC register is.
link |
This is how the color registers work
link |
and how the different things are applied.
link |
And they were designed for a certain reason.
link |
They were designed for an application.
link |
They had an intended use in mind,
link |
but I was starting to look at other ways
link |
that they could perhaps be exploited
link |
that they weren't initially intended for.
link |
Could you comment on like, first of all, what operating system was there?
link |
What instruction set was it?
link |
Like, what are we talking about?
link |
So this was DOS and x86.
link |
The 286s were there and 386s existed.
link |
We had a couple for our development systems,
link |
but we were still targeting the more broad.
link |
It was all DOS 16 bit.
link |
None of this was kind of DOS extenders and things.
link |
How different is it from the systems of today?
link |
Is it kind of a precursor that's similar?
link |
If you open up command.exe or com on Windows,
link |
you see some of the remnants of all of that,
link |
but it was a different world.
link |
It was the 640K is enough world, and nothing was protected.
link |
It crashed all the time.
link |
You had TSRs or terminate and stay resident hacks
link |
on top of things that would cause configuration problems.
link |
All the hardware was manually configured in your auto exec.
link |
So it was a very different world.
link |
But the code is still the same, similar.
link |
You could still write it.
link |
My earliest code there was written in Pascal.
link |
That was what I had learned at an earlier point.
link |
So between BASIC and C++, there was Pascal.
link |
So when BASIC assembly language, some of my intermediate stuff
link |
was, well, you had to for performance.
link |
BASIC was just too slow.
link |
So most of the work that I was doing
link |
as a contract programmer in my teenage years
link |
was assembly language.
link |
Wait, you wrote games in assembly?
link |
Yeah, complete games in assembly language.
link |
And it's thousands and thousands of lines
link |
of three letter acronyms for the instructions.
link |
You don't earn the once again greatest programmer ever label
link |
without being able to write a game in assembly.
link |
Everybody serious wrote their games in assembly language.
link |
Everybody serious.
link |
Everybody serious.
link |
It was an outlier to use Pascal a little bit,
link |
where there was one famous program called wizardry.
link |
It was one of the great early role playing games
link |
that was written in Pascal.
link |
But it was almost nothing used Pascal there.
link |
But I did learn Pascal.
link |
And I remember doing all of my.
link |
To this day, I sketch in data structures.
link |
When I'm thinking about something, I'll open up a file
link |
and I'll start writing struct definitions for how
link |
data is going to be laid out.
link |
And Pascal was kind of formative to that
link |
because I remember designing my RPGs in Pascal record
link |
structures and things like that.
link |
And so I had gotten the Pascal compiler for the Apple 2GS
link |
that I could work on.
link |
And the first IBM game that I developed, I did in Pascal.
link |
And that's actually kind of an interesting story,
link |
again, talking about the constraints and resources
link |
where I had an Apple 2GS.
link |
I didn't have an IBM PC.
link |
I wanted to port my applications to IBM
link |
because I thought I could make more money on it.
link |
So what I wound up doing is I rented a PC for a week
link |
and bought a copy of Turbo Pascal.
link |
And so I had a hard one week and this was cutting into
link |
what minimal profit margin I had there.
link |
But I had this computer for a week.
link |
I had to get my program ported before I had to return
link |
the PC and that was kind of what the first thing
link |
that I had done on the IBM PC and what led me
link |
to taking the job at Softdisk.
link |
And Turbo Pascal, how's that different from regular Pascal?
link |
Is it a different compiler or something like that?
link |
It was a product of Borland, which before Microsoft
link |
kind of killed them, they were the hot stuff
link |
developer tools company.
link |
You had Borland, Turbo Pascal and Turbo C
link |
and Prologue, I mean, all the different things.
link |
But what they did was they took a supremely pragmatic
link |
approach of making something useful.
link |
It was one of these great examples where Pascal
link |
was an academic language and you had things like
link |
the UCSDP system that Wizardry was actually written in
link |
that they did manage to make a game with that,
link |
but it was not a super practical system.
link |
While Turbo Pascal was, it was called Turbo
link |
because it was blazingly fast to compile.
link |
I mean, really ridiculously 10 to 20 times faster
link |
than most other compilers at the time.
link |
But it also had very pragmatic access to look,
link |
you can just poke at the hardware in these different ways
link |
and we have libraries that let you do things.
link |
And it was a pretty good, it was a perfectly good way
link |
to write games and this is one of those things
link |
where people have talked about different paths
link |
that computer development could have taken
link |
where C took over the world for reasons
link |
that came out of Unix and eventually Linux.
link |
And that was not a foregone conclusion at all.
link |
And people can make real reasoned rational arguments
link |
that the world might've been better
link |
if it had gone a Pascal route.
link |
I'm somewhat agnostic on that where I do know
link |
from experience, it was perfectly good enough to do that.
link |
And it had some fundamental improvements
link |
like it had range checked arrays as an option there,
link |
which could avoid many of C's real hazards
link |
that happened in a security space.
link |
But C1, they were basically offering it
link |
about the same level of abstraction.
link |
It was a systems programming language, but.
link |
But you said Pascal had a more emphasis on data structures.
link |
I actually, in the tree of languages,
link |
did Pascal come before C?
link |
They were pretty contemporaneous.
link |
So Pascal's lineage went to Modula 2 and eventually Oberon,
link |
which was another Nicholas word, kind of experimental language.
link |
But they were all good enough at that level.
link |
Now, some of the classic academic oriented Pascals
link |
were just missing fundamental things like, oh, you can't
link |
access this core system thing because we're just
link |
using it to teach students.
link |
But Turbo Pascal showed that only modest changes to it
link |
really did make it a completely capable language.
link |
And it had some reasons why you could implement it
link |
as a single pass compiler so it could be way, way faster,
link |
although less scope for optimizations
link |
if you do it that way.
link |
And it did have some range checking options.
link |
It had a little bit better typing capability.
link |
You'd have properly typed enums, sorts of things,
link |
and other stuff that C lacked.
link |
But C was also clearly good enough.
link |
And it wound up with a huge inertia
link |
from the Unix ecosystem and everything that came with that.
link |
Garbage collection?
link |
No, it was not garbage collected.
link |
It's just the same kind of thing as C.
link |
So you could still have your use after freeze
link |
and all those other problems.
link |
But just getting rid of array overruns,
link |
at least if you were compiled with that debugging option,
link |
certainly would have avoided a lot of problems
link |
and could have a lot of benefits.
link |
But so anyways, that was the next thing.
link |
I had to learn C because C was where it seemed like most
link |
of the things were going.
link |
So I abandoned Pascal and I started working in C.
link |
I started hacking on these hardware things dealing
link |
with the graphics controllers and the EGA systems.
link |
And what we most wanted to do, so at that time
link |
we were sitting in our darkened office
link |
playing all the different console video games.
link |
And we were figuring out what games
link |
do we want to make for our Gamer's Edge product there.
link |
And so we had one of the first Super Nintendos sitting there.
link |
And we had an older Nintendo looking at all those games.
link |
And the core thing that those consoles did that you just
link |
didn't get on the PC games was this ability
link |
to have a massive scrolling world where
link |
most of the games that you would make on the PC
link |
and earlier personal computers would be a static screen.
link |
You move little things around on it and you interact like that.
link |
Maybe you go to additional screens as you move.
link |
But arcade games and consoles had this wonderful ability
link |
to just have a big world that you're slowly
link |
moving your window through.
link |
And that was for those types of games,
link |
that kind of action exploration adventure games,
link |
that was a super, super important thing.
link |
And PC games just didn't do that.
link |
And what I had come across was a couple
link |
of different techniques for implementing that on the PC.
link |
And they're not hard, complicated things.
link |
When I explain them now, they're pretty straightforward.
link |
But just nobody was doing it.
link |
You sound like Einstein describing his five papers
link |
is pretty straightforward.
link |
But they're nevertheless revolutionary.
link |
So side scrolling is a game changer.
link |
Yeah, and scrolling is a genius invention.
link |
It's either vertical.
link |
And some of the consoles had different limitations
link |
about you could do one but not the other.
link |
And there were similar things going on as advancements,
link |
even in the console space, where you'd have the original Mario
link |
game was just horizontal scrolling.
link |
And then later, Mario games added vertical aspects to it
link |
and different things that you were
link |
doing to explore, kind of expand the capabilities there.
link |
And so much of the early game design for decades
link |
was removing limitations, letting you do things
link |
that you envisioned as a designer,
link |
you wanted the player to experience,
link |
but the hardware just couldn't really
link |
or you didn't know how to make it happen.
link |
It felt impossible.
link |
You can imagine that you want to create like this big world
link |
through which you can side scroll,
link |
like through which you can walk.
link |
And then you ask yourself a question,
link |
how do I actually build that in a way that's
link |
like the latency is low enough, the hardware can actually
link |
deliver that in such a way that it's a compelling experience.
link |
Yeah, and we knew what we wanted to do
link |
because we were playing all of these console games,
link |
playing all these Nintendo games and arcade games.
link |
Clearly, there was a whole world of awesome things
link |
there that we just couldn't do on the PC, at least initially.
link |
Because every programmer can tell,
link |
it's like if you want to scroll, you can just
link |
redraw the whole screen.
link |
But then it turns out, well, you're
link |
going five frames per second.
link |
That's not an interactive fun experience.
link |
You want to be going 30 or 60 frames per second or something.
link |
And it just didn't feel like that was possible.
link |
It felt like the PCs had to get five times faster for you
link |
to make a playable game there.
link |
And interestingly, I wound up with two completely different
link |
solutions for the scrolling problem.
link |
And this is a theme that runs through everything, where
link |
all of these big technical advancements, it turns out
link |
there's always a couple different ways of doing them.
link |
And it's not like you found the one true way of doing it.
link |
And we'll see this as we go into 3D games and things later.
link |
But so the scrolling, the first set of scrolling tricks
link |
that I got was the hardware had this ability to,
link |
you could shift inside the window of memory.
link |
So the EGA cards at the time had 256 kilobytes of memory.
link |
And it was awkwardly set up in this planar format,
link |
where instead of having 256 or 24 million colors,
link |
you had 16 colors, which is four bits.
link |
So you had four bit planes, 64k a piece.
link |
Of course, 64k is a nice round number
link |
for 16 bit addressing.
link |
So your graphics card had a 16 bit window
link |
that you could look at.
link |
And you could tell it to start the video scan out
link |
anywhere inside there.
link |
So there were a couple games that
link |
had taken this approach of you could make a 2 by 2 screen
link |
or a 1 by 4 screen.
link |
And you could do scrolling really easily like that.
link |
You could just lay it all out and just pan around there.
link |
But you just couldn't make it any bigger
link |
because that's all the memory that was there.
link |
The first insight to the scrolling that I had
link |
was, well, if we make a screen that's just one tile larger,
link |
and we usually had tiles that were 16 pixels by 16 pixels,
link |
the little classic Mario block that you run into,
link |
lots of art gets drawn that way.
link |
And your screen is a certain number of tiles.
link |
But if you had one little buffer region outside of that,
link |
you could easily pan around inside that 16 pixel region.
link |
That could be perfectly smooth.
link |
But then what happens if you get to the edge
link |
and you want to keep going?
link |
The first way we did scrolling was
link |
what I called adaptive tile refresh, which was really
link |
just a matter of you get to the edge
link |
and then you go back to the original point
link |
and then only change the tiles that have actually
link |
that are different between where it was.
link |
In most of the games at the time,
link |
if you think about your classic Super Mario Brothers game,
link |
you've got big fields of blue sky, long rows
link |
of the same brick texture.
link |
And there's a lot of commonality.
link |
It's like a data compression thing.
link |
If you take the screen and you set it down
link |
on top of each other, in general, only about 10%
link |
of the tiles were actually different there.
link |
So this was a way to go ahead and say,
link |
well, I'm going to move it back, and then I'm only
link |
going to change those 10%, 20%, whatever percent tiles there.
link |
And that meant that it was essentially five times faster
link |
than if you were redrawing all of the tiles.
link |
And that worked well enough for us
link |
to do a bunch of these games for Gamer's Edge.
link |
We had a lot of these scrolling games like Slurred Axe
link |
and Shadow Knights and things like that
link |
that we were cranking out at this high rate that
link |
had this scrolling effect on it.
link |
And it worked well enough.
link |
There were design challenges there
link |
where if you made the worst case,
link |
if you made a checkerboard over the entire screen,
link |
you scroll over one and every single tile changes
link |
and your frame rate's now five frames per second
link |
because it had to redraw everything.
link |
So the designers had a little bit
link |
that they had to worry about.
link |
They had to make these relatively plain looking
link |
levels, but it was still pretty magical.
link |
It was something that we hadn't seen before.
link |
And the first thing that we wound up doing with that
link |
was I had just gotten this working,
link |
and Tom Hall was sitting there with me,
link |
and we were looking over at our Super Nintendo on the side
link |
there with Super Mario 3 running.
link |
And we had the technology, we had the tools set up there,
link |
and we stayed up all night.
link |
And we basically cloned the first level
link |
of Super Mario Brothers.
link |
Performance wise as well?
link |
And we had our little character running and jumping in there.
link |
It was close to pixel accurate as far
link |
as all the backgrounds and everything,
link |
but the gaming was just stuff that we cobbled together
link |
from previous games that I had written.
link |
I just really kitbashed the whole thing together
link |
to make this demo.
link |
And that was one of the rare cases
link |
when I said I don't usually do these all night programming
link |
There's probably only two memorable ones
link |
that I can think about.
link |
One was the all nighter to go ahead and get
link |
our Dangerous Dave and copyright infringement
link |
is how we titled it.
link |
Because we had a game called Dangerous Dave running around
link |
with a shotgun shooting things, and we were just
link |
taking our most beloved game at the time there, the Super Mario
link |
3, and sticking Dave inside that with this new scrolling
link |
technology that was going perfectly smooth for them
link |
And Tom and I just blearily the next morning left,
link |
and we left a disk on the desk for John Romero and Jay
link |
Wilbur to see and just said, run this.
link |
And we eventually made it back in later in the day,
link |
and it was like they grabbed us and pulled us into the room.
link |
And that was the point where they were like,
link |
we got to do something with this.
link |
We're going to make a company.
link |
We're going to go make our own games, where
link |
this was something that we were able to just kind of hit them
link |
with a hammer of an experience like, wow, this
link |
is just like so much cooler than what we thought
link |
was possible there.
link |
And initially, we tried to get Nintendo
link |
to let us make Super Mario 3 on the PC.
link |
That's really what we wanted to do.
link |
We were like, hey, we can finish this.
link |
It's line of sight.
link |
This will be great.
link |
And we sent something to Nintendo,
link |
and we heard that it did get looked at in Japan,
link |
and they just weren't interested in that.
link |
But that's another one of those life could
link |
have gone a very different way, where
link |
we could have been like Nintendo's house PC team
link |
And define the direction of Wolfenstein and Doom and Quake
link |
could have been a Nintendo creation.
link |
So at the same time that we were just
link |
doing our first scrolling demos, we
link |
reached out to Scott Miller at Apogee and said,
link |
it's like, hey, we do want to make some games.
link |
These things that you think you want, those are nothing.
link |
What do you see what we can actually do now?
link |
This is going to be amazing.
link |
And he just popped right up and sent a check to us,
link |
where at that point, we still thought
link |
he might be a fraud, that he was just lying about all of this.
link |
But he was totally correct on how much money
link |
he was making with his shareware titles.
link |
And this was his real brainstorm about this,
link |
where shareware was this idea that software
link |
doesn't have a fixed price.
link |
If you use it, you send out of the goodness of your heart
link |
some money to the creator.
link |
And there were a couple of utilities
link |
that did make some significant success like that.
link |
But for the most part, it didn't really work.
link |
Now, there wasn't much software in a pure shareware model
link |
that was successful.
link |
The Apogee innovation was to take something,
link |
call it shareware, split it into three pieces.
link |
You always made a trilogy.
link |
And you would put the first piece out,
link |
but then you buy the whole trilogy
link |
for some shareware amount, which in reality,
link |
it meant that the first part was a demo,
link |
where you kind of like the demo went everywhere for free
link |
and you paid money to get the whole set.
link |
But it was still play to shareware.
link |
And we were happy to have the first one go everywhere.
link |
And it wasn't a crippled demo, where the first episode
link |
of all these trilogies, it was a real complete game.
link |
And probably 20 times as many people played that part of it,
link |
thought they had a great game, had fond memories of it,
link |
but never paid us a dime.
link |
But enough people were happy with that,
link |
where it was really quite successful.
link |
And these early games that we didn't think very much of
link |
compared to commercial quality games,
link |
but they were doing really good business,
link |
some fairly crude things.
link |
And people, it was good business.
link |
People enjoyed it.
link |
And it wasn't like you were taking a crapshoot
link |
on what you were getting.
link |
You just played a third of the experience.
link |
And you loved it enough to hand write out a check
link |
and put it in an envelope and address it and send it out
link |
to Apogee to get the rest of them.
link |
So it was a really pretty feel good business prospect there,
link |
because everybody was happy.
link |
They knew what they were getting when they sent it in.
link |
They would send in fan mail.
link |
If you're going to the trouble of addressing a letter
link |
and filling out an envelope, you write something in it.
link |
And there were just the literal bags of fan mail
link |
for the shareware games, so people loved them.
link |
I should mention that for you, the definition of wealth
link |
is being able to have pizza whenever you want.
link |
For me, there was a dream that I would play shareware games
link |
over and over, the part that's free, over and over.
link |
And it was very deeply fulfilling experience.
link |
But I dreamed of a time when I could actually
link |
afford the full experience.
link |
And this is kind of this dreamland beyond the horizon
link |
when you could find out what else is there.
link |
In some sense, even just playing the shareware was,
link |
it's the limitation of that.
link |
You know, life is limited, eventually we all die.
link |
In that way, shareware was somehow really fulfilling
link |
to have this kind of mysterious thing beyond what's free
link |
It's kind of, I don't know, maybe it's because
link |
a part of my childhood is playing shareware games.
link |
That was a really fulfilling experience.
link |
It's so interesting how that model still brought joy
link |
to so many people.
link |
The 20X people that played it.
link |
I felt very good about that.
link |
I would run into people that would say,
link |
oh, I loved that game that you had early on,
link |
Commander Keen, whatever.
link |
And no, they meant just the first episode
link |
that they got to see everywhere.
link |
That's me, I played the crap out of Commander Keen.
link |
And that was all good.
link |
But so we were in this position where Scott Miller
link |
was just fronting us cash, saying, yeah, make a game.
link |
But we did not properly pull the trigger and say,
link |
all right, we're quitting our jobs.
link |
We were like, we're gonna do both.
link |
We're gonna keep working at Softdisk, working on this.
link |
And then we're going to go ahead and make a new game
link |
for Apogee at the same time.
link |
And this eventually did lead to some legal problems.
link |
And we had trouble.
link |
It all got worked out in the end,
link |
but it was not a good call at the time there.
link |
And your legal mind at the time was not stellar.
link |
You were not thinking in terms of the illegal terms.
link |
No, I definitely wasn't.
link |
And in hindsight, yeah, it's like,
link |
how did we think we were gonna get away
link |
with even using our work computers
link |
to write software for our breakaway new company?
link |
It was not a good plan.
link |
How did Commander Keen come to be?
link |
So the design process, we would start from,
link |
we had some idea of what we wanted to do.
link |
We wanted to do a Mario like game.
link |
It was gonna be a side scroller.
link |
It was gonna use the technology.
link |
We had some sense of what it would have to look like
link |
because of the limitations
link |
of this adaptive tile refresh technology.
link |
It had to have fields of relatively constant tiles.
link |
You couldn't just paint up a background
link |
and then move that around.
link |
The early design or all the design for Commander Keen
link |
really came from Tom Hall,
link |
where he was kind of the main creative mind
link |
for the early id software stuff,
link |
where we had an interesting division of things
link |
where Tom was all creative and design.
link |
I was all programming.
link |
John Romero was an interesting bridge
link |
where he was both a very good programmer
link |
and also a very good designer and artist
link |
and kind of straddled between the areas.
link |
But Commander Keen was very much Tom Hall's baby.
link |
And he came up with all the design and backstory
link |
for the different things of kind of a mad scientist,
link |
little kid with building a rocket ship
link |
and a zap gun and visiting alien worlds
link |
and doing all of this that the background
link |
that we lay the game inside of.
link |
And there's not a whole lot to any of these things.
link |
Design for us was always just what we needed to do
link |
to make the game that was gonna be so much fun to play.
link |
And we laid out our first trilogy of games,
link |
the shareware formula, it's gonna be three pieces.
link |
We make Commander Keen one, two, and three.
link |
And we just really started busting on all that work.
link |
And it went together really quickly.
link |
It was like three months or something
link |
that while we were still making games every month
link |
for Gamer's Edge, we were sharing technology between that.
link |
I'd write a bunch of code for this
link |
and we'd just kind of use it for both.
link |
Again, not a particularly good idea there
link |
that had consequences for us.
link |
But in three months, we got our first game out
link |
and all of a sudden it was three times as successful
link |
as the most successful thing Apogee had had before.
link |
And we were making like $30,000 a month
link |
immediately from the Commander Keen stuff.
link |
And that was again, a surprise to us.
link |
It was more than we thought that that was gonna make.
link |
And we said, well, we're gonna certainly roll
link |
into another set of titles from this.
link |
And in that three months, I had come up
link |
with a much better way of doing the scrolling technology
link |
that was not the adaptive tile refresh,
link |
which in some ways was even simpler.
link |
And these things, so many of the great ideas of technology
link |
are things that are back of the envelope designs.
link |
I make this comment about modern machine learning
link |
where all the things that are really important
link |
practically in the last decade are,
link |
each of them fits on the back of an envelope.
link |
There are these simple little things.
link |
They're not super dense, hard to understand technologies.
link |
And so the second scrolling trick was just a matter of like,
link |
okay, we know we've got this 64K window.
link |
And the question was always like,
link |
well, you could make a two by two,
link |
but you can't go off the edge.
link |
But I finally asked, well, what actually happens
link |
if you just go off the edge?
link |
If you take your start and you say, it's like,
link |
okay, I can move over, I'm scrolling,
link |
I can move over, I can move down, I'm scrolling.
link |
I get to what should be the bottom of the memory window.
link |
It's like, well, what if I just keep going?
link |
And I say, I'm gonna start at,
link |
what happens if I start at FFFE
link |
at the very end of the 64K block?
link |
And it turns out it just wraps back around to the top
link |
And I'm like, oh, well, this makes everything easy.
link |
You can just scroll the screen everywhere
link |
and all you have to draw is just one new line of tiles,
link |
which everything you expose,
link |
it might be unaligned off various parts
link |
of the screen memory, but it just works.
link |
That no longer had the problem of you had to have fields
link |
of the similar colors because it doesn't matter
link |
what you're doing, you could be having
link |
a completely unique world and you're just drawing
link |
the new strip as it comes on.
link |
But it might be, like you said, unaligned,
link |
so it can be all over the place.
link |
Yes, and it turns out it doesn't matter.
link |
I would have two page flipped screens,
link |
as long as they didn't overlap, they moved in series
link |
through this two dimensional window of graphics.
link |
And that was one of those like, well, this is so simple.
link |
This just works, it's faster,
link |
there it seemed like there was no downside.
link |
Funny thing was, it turned out after we shipped titles
link |
with this, there were what they called super VGA cards,
link |
the cards that would allow higher resolutions
link |
and different features that the standard ones didn't.
link |
And on some of those cards,
link |
this was a weird compatibility quirk again,
link |
because nobody thought this was not
link |
what it was designed to do.
link |
And some of those cards had more memory.
link |
They had more than just 256K and four planes.
link |
They had 512K or a megabyte.
link |
And on some of those cards, I scroll my window down
link |
and then it goes into uninitialized memory
link |
that actually exists there
link |
rather than wrapping back around at the top.
link |
And then I was in the tough position
link |
of do I have to track every single one of these?
link |
And it was a mad house back then
link |
with there were 20 different video card vendors
link |
with all slightly different implementations
link |
of their nonstandard functionality.
link |
So either I needed to natively program
link |
all of the VGA cards there to map in that memory
link |
and keep scrolling down through all of that,
link |
or I kind of punted and took the easy solution
link |
of when you finally did run to the edge of the screen,
link |
I accepted a hitch and just copied the whole screen up there.
link |
So on some of those cards, it was a compatibility mode.
link |
In the normal ones, when it all worked fine,
link |
everything was just beautifully smooth.
link |
But if you had one of those cards
link |
where it did not wrap the way I wanted it to,
link |
you'd be scrolling around, scrolling around,
link |
and then eventually you'd have a little hitch
link |
where 200 milliseconds or something
link |
that was not super smooth as it froze a little bit.
link |
And this is the binary thing.
link |
Is it one of the standard screens
link |
or is it one of the weird ones, the Super VGA ones?
link |
And so we would default to,
link |
and I think that was one of those
link |
that changed over the kind of course of deployment
link |
where early on we would have a normal mode
link |
and then you would enable the compatibility flag
link |
if your screen did this crazy flickery thing
link |
when you got to a certain point in the game.
link |
And then later, I think it probably got enabled by default
link |
as just more and more of the cards
link |
kind of did not do exactly the right thing.
link |
And that's the two edged sword
link |
of doing unconventional things with technology
link |
where you can find something that nobody thought about
link |
doing that kind of scrolling trick
link |
when they set up those cards.
link |
But the fact that nobody thought that was the primary reason
link |
when I was relying on that,
link |
then I wound up being broken on some of the later cards.
link |
Let me take a bit of a tangent,
link |
but ask you about the hacker ethic.
link |
Cause you mentioned shareware.
link |
It's an interesting world.
link |
The world of people that make money,
link |
the business and the people that build systems,
link |
And what is the hacker ethic?
link |
You've been a man of the people
link |
and you've embodied at least the part of that ethic.
link |
What does it mean?
link |
What did it mean to you at the time?
link |
What does it mean to you today?
link |
So Steven Levy's book, Hackers,
link |
was a really formative book for me as a teenager.
link |
I mean, I read it several times
link |
and there was all of the great lore
link |
of the early MIT era of hackers
link |
and ending up at the end with,
link |
it kind of went through the early MIT hackers
link |
and then the Silicon Valley hardware hackers
link |
and then the game hackers in part three.
link |
And at that time as a teenager,
link |
I really was kind of bitter in some ways.
link |
I thought I was born too late.
link |
I thought I missed the window there.
link |
And I really thought I belonged in that third section
link |
of that book with the game hackers.
link |
And they were talking about the Williams at Sierra
link |
and origin systems with Richard Garriott.
link |
And it's like, I really wanted to be there.
link |
And I knew that was now a few years in the past.
link |
but the early days, especially the early MIT hacker days,
link |
talking a lot about this sense of the hacker ethic,
link |
that there was this sense that
link |
it was about sharing information, being good,
link |
not keeping it to yourself,
link |
and that it's not a zero sum game,
link |
that you can share something with another programmer
link |
and it doesn't take it away from you.
link |
You then have somebody else doing something.
link |
And I also think that there's an aspect of it
link |
where it's this ability to take joy
link |
in other people's accomplishments,
link |
where it's not the cutthroat bit of like,
link |
I have to be first,
link |
I have to be recognized as the one that did this
link |
but being able to see somebody else do something
link |
and say, holy shit, that's amazing.
link |
And just taking joy in the ability of something amazing
link |
that somebody else does.
link |
And the big thing that I was able to do through ID Software
link |
was this ability to eventually release the source code
link |
like all of our really seminal game titles.
link |
And that was a, it was a stepping stone process
link |
where we were kind of surprised early on
link |
where people were able to hack the existing games.
link |
And of course I had experience with that.
link |
I remember hacking my copies of Ultima.
link |
So I'd give myself 9999 gold and raise my levels
link |
and break out the sector editor.
link |
And so I was familiar with all of that.
link |
So it was just, it was with a smile
link |
when I started to see people doing that to our games,
link |
I am making level editors for Commander Keen
link |
or hacking up Wolfenstein 3D,
link |
but I made the pitch internally
link |
that we should actually release our own tools
link |
for like what we did, what we use to create the games.
link |
And that was a little bit debatable about,
link |
well, will this let other, we'll give people leg up.
link |
what's that gonna mean for the competition?
link |
But the really hard pitch was to actually release
link |
the full source code for the games.
link |
And it was a balancing act with the other people
link |
inside the company where it's interesting
link |
how the programmers generally did get,
link |
certainly the people that I worked closely with,
link |
they did kind of get that hacker ethic bit
link |
where you wanted to share your code.
link |
You were proud of it.
link |
You wanted other people to take it, do cool things with it.
link |
But interestingly, the broader game industry
link |
is a little more hesitant to embrace that
link |
than like the group of people
link |
that we happen to have at id Software,
link |
where it was always a little interesting to me
link |
seeing how a lot of people in the game modding community
link |
were very possessive of their code.
link |
They did not wanna share their code.
link |
They wanted it to be theirs.
link |
It was their claim to fame.
link |
And that was much more like what we tended to see
link |
with artists where the artists understand something
link |
about credit and wanting it to be known as their work.
link |
And a lot of the game programmers
link |
felt a little bit more like artists
link |
than like hacker programmers,
link |
in that it was about building something
link |
that maybe felt more like art to them
link |
than the more tool based and exploration based
link |
kind of hacking culture side of things.
link |
Yeah, it's so interesting that this kind of fear
link |
that credit will not be sufficiently attributed to you.
link |
And that's one of the things that I do bump into a lot
link |
because I try not to go...
link |
I mean, it's easy for me to say
link |
because so much credit is heaped on me
link |
for the id Software side of things.
link |
But when people come up and they wanna pick a fight
link |
and say, no, it's like that wasn't where first person gaming
link |
came from and you can point to some of like things
link |
on obscure titles that I was never aware of
link |
or like the old Playdoh systems
link |
or each personal computer had something that was 3Dish
link |
and moving around.
link |
And I'm happy to say it's like, no, I mean,
link |
I saw Battlezone and Star Wars in the arcades.
link |
I had seen 3D graphics, I had seen all these things
link |
that are standing on the shoulders of lots of other people.
link |
But sometimes these examples they pull out,
link |
it's like, no, I didn't know that existed.
link |
I mean, I had never heard of that before then
link |
that didn't contribute to what I made,
link |
but there's plenty of stuff that did.
link |
And I think there's good cases to be made
link |
that obviously Doom and Quake and Wolfenstein
link |
were formative examples for what everything
link |
that came after that.
link |
But I don't feel the need to go fight
link |
and say claim primacy or initial invention
link |
of anything like that.
link |
But a lot of people do want to.
link |
I think when you fight for the credit in that way
link |
and it does go against the hacker ethic,
link |
you destroy something fundamental about the culture,
link |
about the community that builds cool stuff.
link |
I think credit ultimately,
link |
so I had this sort of,
link |
there's a famous wrestler and freestyle wrestling
link |
called Buvai Sarasateeva.
link |
And he always preached that you should just focus
link |
on the art of the wrestling and let people
link |
write your story however they want.
link |
The highest form of the art is just focusing on the art.
link |
And that is something about the hacker ethic
link |
is just focus on building cool stuff,
link |
sharing it with other cool people,
link |
and credit will get assigned correctly
link |
in the long arc of history.
link |
Yeah, and I generally think that's true.
link |
there's some things, there's a graphics technique
link |
that got labeled CarMax Reverse.
link |
I am literally named,
link |
and it turned out that I wasn't the first person
link |
to figure that out.
link |
Most scientific things or mathematical things you wind up,
link |
it's like, oh, this other person
link |
had actually done that somewhat before.
link |
And then there's things that get attributed to me
link |
like the inverse square root hack that I actually didn't do.
link |
I flat out, that wasn't me.
link |
And it's like, it's weird how the memetic power
link |
of the internet, I cannot convince people of that.
link |
You're like the Mark Twain of programming.
link |
Everything just gets attributed to you now,
link |
even though you've never sought the credit of things.
link |
But part of the fact of the humility behind that
link |
is what attracts the attributions.
link |
Let's talk about a game.
link |
I mean, one of the greatest games ever made.
link |
I know you could talk about doing Quake and so on,
link |
but to me, Wolfenstein 3D was like, whoa.
link |
It blew my mind that a world like this could exist.
link |
So how did Wolfenstein 3D come to be
link |
in terms of the programming, in terms of the design,
link |
in terms of some of the memorable technical challenges?
link |
And also, actually just something you haven't mentioned
link |
is how did these ideas come to be inside your mind,
link |
the adaptive side scrolling,
link |
the solutions to these technical challenges?
link |
So I usually can introspectively pull back
link |
pretty detailed accounts of how technology solutions
link |
and design choices on my part came to be,
link |
where technically we had done two games,
link |
3D games like that before,
link |
where Hover Tank was the first one,
link |
which had flat shaded walls,
link |
but did have the scaled enemies inside it.
link |
And then Catacombs 3D, which had textured walls,
link |
scaled enemies, and some more functionality,
link |
like the disappearing walls and some other stuff.
link |
But what's really interesting
link |
from a game development standpoint is those games,
link |
Catacombs 3D, Hover Tank, and Wolfenstein,
link |
they literally used the same code
link |
for a lot of the character behavior
link |
that a 2D game that I had made earlier called Catacombs did,
link |
where it was an overhead view game, kind of like Gauntlet,
link |
you're running around and you can open up doors,
link |
pick up items, basic game stuff.
link |
And the thought was that this exact same game experience
link |
just presented in a different perspective.
link |
It could be literally the same game
link |
just with a different view into it,
link |
would have a dramatically different impact on the players.
link |
So it wasn't a true 3D,
link |
you're saying that you could kind of fake it,
link |
you can like scale enemies,
link |
meaning things that are farther away,
link |
you can make them smaller.
link |
So the game was a 2D map,
link |
like all of our games use the same tool for creation.
link |
We use the same map editor for creating Keen as Wolfenstein
link |
and Hover Tank and Catacombs and all this stuff.
link |
So the game was a 2D grid made out of blocks.
link |
And you could say, well, these are walls,
link |
these are where the enemies start,
link |
then they start moving around.
link |
And these early games like Catacombs,
link |
you played it strictly in a 2D view.
link |
It was a scrolling 2D view,
link |
and that was kind of using an adaptive tile refresh
link |
at the time to be able to do something like that.
link |
And then the thought that these early games,
link |
all it did was take the same basic enemy logic,
link |
but instead of seeing it from the God's eye view on top,
link |
you were inside it and turning from side to side,
link |
yawing your view and moving forwards and backwards
link |
And it's a striking thing where you always talk about
link |
wanting to isolate and factor changes in values.
link |
And this was one of those most pure cases there
link |
where the rest of the game changed very little.
link |
It was our normal kind of change the colors on something
link |
and draw a different picture for it,
link |
but it's kind of the same thing.
link |
But the perspective changed in a really fundamental way.
link |
And it was dramatically different.
link |
I can remember the reactions where the artist, Adrian,
link |
that had been drawing the pictures for it.
link |
We had a cool big troll thing in Catacombs 3D,
link |
and we had these walls that you could get a key
link |
and you could make the blocks disappear,
link |
get really simple stuff.
link |
Blocks could either be there or not there.
link |
So our idea of a door was being able to make a set of blocks
link |
And I remember the reaction
link |
where he had drawn these characters
link |
and he was slowly moving around
link |
and like people had no experience with 3D navigation.
link |
It was all still keyboard.
link |
We didn't even have mice set up at that time,
link |
but slowly moving, going up, picked up a key, go to a wall.
link |
The wall disappears in a little animation
link |
and there's a monster like right there.
link |
And he practically fell out of his chair.
link |
It was just like, ah!
link |
And games just didn't do that.
link |
The games were the God's eye view.
link |
You were a little invested in your little guy.
link |
You can be like happy or sad when things happen,
link |
but you just did not get that kind of startle reaction.
link |
You weren't inside the game.
link |
Something in the back of your brain,
link |
some reptile brain thing is just going,
link |
oh shit, something just happened.
link |
And that was one of those early points where it's like,
link |
yeah, this is gonna make a difference.
link |
This is going to be powerful and it's gonna matter.
link |
Were you able to imagine that in the idea stage or no?
link |
So not that exact thing.
link |
So again, we had cases like the arcade games,
link |
Battle Zone and Star Wars
link |
that you could kind of see a 3D world
link |
and things coming at you and you get some sense of it,
link |
but nothing had done the kind of worlds that we were doing
link |
and the sort of action based things.
link |
3D at the time was really largely
link |
about the simulation thoughts.
link |
And this is something that really
link |
might have trended differently
link |
if not for the id Software approach in the games
link |
where there were flight simulators,
link |
there were driving simulators,
link |
you had like hard drive in and Microsoft flight simulator.
link |
And these were doing 3D and general purpose 3D
link |
in ways that were more flexible
link |
than what we were doing with our games,
link |
but they were looked at as simulations.
link |
They weren't trying to necessarily be fast or responsive
link |
or letting you do kind of exciting maneuvers
link |
because they were trying to simulate reality
link |
and they were taking their cues from the big systems,
link |
the Evans and Sutherlands and the Silicon Graphics
link |
that were doing things.
link |
But we were taking our cues
link |
from the console and arcade games.
link |
We wanted things that were sort of quarter eaters
link |
that were doing fast paced things
link |
that you could smack you around
link |
rather than just smoothly gliding you from place to place.
link |
And you know, a funny thing is,
link |
so much that that built into us
link |
that Wolfenstein still had lives
link |
and you had like one of the biggest power ups
link |
in all these games like was an extra life
link |
because you started off with three lives
link |
and you lose your lives and then it's game over
link |
and there weren't save games in most of this stuff.
link |
It was, it sounds almost crazy to say this,
link |
but it was an innovation in Doom
link |
to not have lives.
link |
You know, you could just play Doom as long as you wanted.
link |
You just restart at the start of the level and why not?
link |
This is like, we aren't trying to take people's quarters.
link |
They've already paid for the entire game.
link |
We want them to have a good time
link |
and you would have some, you know,
link |
some old timer purists that might think
link |
that there's something to the epic journey
link |
of making it to the end,
link |
having to restart all the way from the beginning
link |
after a certain number of tries.
link |
But now more fun is had when you just let people
link |
kind of keep trying when they're stuck
link |
rather than having to go all the way back
link |
and learn different things.
link |
So you've recommended the book,
link |
Game Engine Black Book Wolfenstein 3D
link |
for technical exploration of the game.
link |
So looking back 30 years,
link |
what are some memorable technical innovations
link |
that made this perspective shift into this world
link |
that's so immersive that scares you when a monster appears
link |
or some things you have to solve?
link |
So one of the interesting things
link |
that come back to the theme of deadlines
link |
and resource constraints,
link |
the game Catacombs 3D,
link |
we shipped, we were supposed to be shipping this
link |
for Gamers Edge on a monthly cadence and I had slipped.
link |
I was actually late.
link |
It slipped like six weeks
link |
because this was texture mapped walls
link |
doing stuff that I hadn't done before.
link |
And at the six week point,
link |
it was still kind of glitchy and buggy.
link |
There were things that I knew that if you had a wall
link |
that was like almost edge on, you could slide over to it
link |
and you could see some things freak out
link |
or vanish or not work.
link |
but I was up against the wall.
link |
We had to ship the game.
link |
It was still a lot of fun to play.
link |
It was novel, nobody had seen it.
link |
It gave you that startle reflex reaction.
link |
So it was worth shipping,
link |
but it had these things that I knew were kind of flaky
link |
and janky and not what I was really proud of.
link |
So one of the things that I did very differently
link |
in Wolfenstein was I went,
link |
Catacombs used almost a conventional thing
link |
where you had segments that were one dimensional polygons
link |
basically that were clipped and back faced
link |
and done kind of like a very crude 3D engine
link |
from the professionals,
link |
but I wasn't getting it done right.
link |
I was not doing a good enough job.
link |
I didn't really have line of sight to fix it right.
link |
There's stuff that of course I look back,
link |
it's like, oh, it's obvious how to do this
link |
and do the math right, do your clipping right,
link |
check all of this, how you handle the precision.
link |
But I did not know how to do that at that time.
link |
Was that the first 3D engine you wrote Catacombs 3D?
link |
Yeah, Hover Tank had been a little bit before that,
link |
but that had the flat shaded walls.
link |
So the texture mapping on the walls
link |
was what was bringing in some of these challenges
link |
that was hard for me
link |
and I couldn't solve it right at the time.
link |
Can you describe what flat shading is and texture mapping?
link |
So the walls were solid color,
link |
one of 16 colors in Hover Tank.
link |
So that's easy, it's fast,
link |
you just draw the solid color for everything.
link |
Texture mapping is what we all see today
link |
where you have an image that is stretched and distorted
link |
onto the walls or the surfaces that you're working with.
link |
And it was a long time for me to just figure out
link |
how to do that without it distorting in the wrong ways.
link |
And I did not get it all exactly right in Catacombs
link |
and I had these flaws.
link |
So that was important enough to me
link |
that rather than continuing to bang my head on that
link |
when I wasn't positive I was gonna get it,
link |
I went with a completely different approach
link |
for drawing for figuring out where the walls were
link |
which was a ray casting approach,
link |
which I had done in Catacombs 3D I had a bunch of C code
link |
trying to make this work right and it wasn't working right.
link |
In Wolfenstein I wound up going to a very small amount
link |
So in some ways there should be a slower way of doing it
link |
but by making it a smaller amount of work
link |
that I could more tightly optimize it worked out
link |
and Wolfenstein 3D was just absolutely rock solid.
link |
It was nothing glitched in there.
link |
The game just was pretty much flawless through all of that
link |
and I was super proud of that.
link |
But eventually like in the later games
link |
I went back to the more span based things
link |
where I could get more total efficiency
link |
once I really did figure out how to do it.
link |
So there were two sort of key technical things
link |
to Wolfenstein, one was this ray casting approach
link |
which you still to this day you see people go and say
link |
let's write a ray casting engine
link |
because it's an understandable way of doing things
link |
that lets you make games very much like that.
link |
So you see ray casters in JavaScript,
link |
ray casters in Python, people that are basically going
link |
and re implementing that approach to taking a tiled world
link |
and casting out into it.
link |
It works pretty well but it's not the fastest way
link |
Can you describe what ray casting is?
link |
So you start off and you've got your screen
link |
which is 320 pixels across at the time
link |
if you haven't sized down the window for greater speed
link |
and at every pixel there's gonna be an angle
link |
from you've got your position in the world
link |
and you're going to just run along that angle
link |
and keep going until you hit a block.
link |
So up to 320 times across there
link |
it's gonna throw a cast array out into the world
link |
from wherever your origin is until it runs into a wall
link |
and then it can figure out exactly where on the wall it hits.
link |
The performance challenge of that is as it's going out
link |
every block it's crossing it checks is this a solid wall.
link |
So that means that in like the early Wolfenstein levels
link |
you're in a small jail cell going out into a small hallway
link |
it's super efficient for that
link |
because you're only stepping across three or four blocks
link |
but then if somebody makes a room that covers
link |
our maps were limited to 64 by 64 blocks.
link |
If you made one room that was nothing but walls
link |
at the far space it would go pretty slow
link |
because it would be stepping across 80 tile tests
link |
or something along the way.
link |
By the way the physics of our universe
link |
seems to be competing in this very thing.
link |
So this maps nicely to the actual physics of our world.
link |
Yeah you get like I follow a little bit of something
link |
like Steven Wolfram's work on interconnected
link |
network information states of that
link |
and that's it's beyond what I can have an informed opinion
link |
on but it's interesting that people are considering
link |
things like that and have things that can back it up.
link |
Yeah there's whole different sets
link |
of interesting stuff there.
link |
So Wolfenstein 3D had ray casting.
link |
Ray casting and then the other kind of key aspect
link |
was what I called compiled scalers
link |
where the idea of you'd saw this in the earlier
link |
classic arcade games like Space Harrier and stuff
link |
where you would take a picture which is normally drawn
link |
directly on the screen and then if you have the ability
link |
to make it bigger or smaller big chunky pixels
link |
or fizzly small drop sampled pixels.
link |
That's the fundamental aspect of what our characters
link |
were doing in these 3D games.
link |
You would have it's just like you might have drawn
link |
a tiny little character but now we can make them
link |
really big and make them really small and move it around.
link |
That was the limited kind of 3D that we had for characters.
link |
To make them turn there were literally
link |
eight different views of them.
link |
You didn't actually have a 3D model that would rotate.
link |
You just had these cardboard cutouts.
link |
But that was good enough for that startle fight reaction
link |
and it was kind of what we had to deal with there.
link |
So a straightforward approach to do that
link |
you could just write out your doubly nested loop
link |
of you've got your stretch factor
link |
and it's like you've got a point,
link |
you stretch by a little bit,
link |
it might be on the same pixel,
link |
it might be on the next pixel,
link |
it might have skipped a pixel.
link |
You can write that out but it's not gonna be fast enough
link |
where especially you get a character for that
link |
right in your face monster covering
link |
almost the entire screen.
link |
Doing that with a general purpose scaling routine
link |
would have just been much too slow.
link |
It would have worked when they're small characters
link |
but then it would get slower and slower
link |
as they got closer to you until right at the time
link |
when you most care about having a fast reaction time,
link |
the game would be chunking down.
link |
So the fastest possible way to draw pixels at that time
link |
was to, instead of saying I've got a general purpose
link |
version that can handle any scale,
link |
I used a program to make essentially
link |
a hundred or more separate little programs
link |
that was optimized for I will take an image
link |
and I will draw it 12 pixels tall.
link |
I'll take an image, I'll draw it 14 pixels tall
link |
up by every two pixels even for that.
link |
So you would have the most optimized code
link |
so that in the normal case
link |
where most of the world is fairly large,
link |
like the pixels are big,
link |
we did not have a lot of memory.
link |
So in most cases that meant that you would load
link |
a pixel color and then you would store it multiple times.
link |
So that was faster than even copying an image
link |
in a normal conventional case
link |
because most of the time the image is expanded.
link |
So instead of doing one read, one write for a simple copy,
link |
you might be doing one read and three or four writes
link |
as it got really big.
link |
And that had the beneficial aspect
link |
of just when you needed the performance most
link |
when things are covering the screen,
link |
it was giving you the most acceleration for that.
link |
By the way, were you able to understand this
link |
through thinking about it or were you testing
link |
like the right speed and like that?
link |
This again comes back to, I can find the antecedents
link |
for things like this.
link |
So back in the Apple II days,
link |
the graphics were essentially single bits at a time.
link |
And if you wanted to make your little spaceship,
link |
if you wanted to make it smoothly go across the world,
link |
if you just took the image and you drew it out
link |
at the next location, you would move by seven pixels
link |
at a time, so it'd go chunk, chunk, chunk.
link |
If you wanted to make it move smoothly,
link |
you actually had to make seven versions of the ship
link |
that were pre shifted.
link |
You could write a program that would shift it dynamically,
link |
but on a one megahertz processor,
link |
that's not going anywhere fast.
link |
So if you wanted to do a smooth moving fast action game,
link |
you made separate versions of each of these sprites.
link |
Now there were a few more tricks you could pull
link |
that if it still wasn't fast enough,
link |
you could make a compiled shape
link |
where instead of this program that normally copies an image
link |
and says like, get this byte from here, stored here,
link |
get this byte, store this byte.
link |
If you've got the memory space, you could say,
link |
I'm going to write the program
link |
that does nothing but draw this shape.
link |
It's going to be like,
link |
I'm going to load the immediate value 25,
link |
which is some bit pattern,
link |
and then I'm going to store that at this location.
link |
Rather than loading something from memory
link |
that involved indexing registers and this other slow stuff,
link |
you could go ahead and say,
link |
no, I'm gonna hard code the exact values
link |
of all of the image right into the program.
link |
And this was always a horrible trade off there,
link |
which you didn't have much memory
link |
and you didn't have much speed.
link |
But if you had something that you wanted to go really fast,
link |
you could turn it into a program.
link |
And that was, knowing about that technique
link |
is what made me think about some of these unwinding it
link |
for the PC where people that didn't come
link |
from that background were less likely to think about that.
link |
I mean, there's some deep parallels
link |
probably to human cognition as well.
link |
There's something about optimizing and compressing
link |
the processing of a new information
link |
that requires you to predict the possible ways
link |
in which the game or the world might unroll.
link |
And you have something like compiled scaler is always there.
link |
So you have like, like you have a prediction
link |
of how the world will unroll
link |
and you have some kind of optimized data structure
link |
for that prediction.
link |
And then you can modify if the world turns out
link |
to be different, you can modify a slight way.
link |
And as far as building out techniques,
link |
so much of the brain is about the associative context.
link |
You know, they're just, when you learn something,
link |
it's in the context of something else
link |
and you can have faint, tiny little hints of things.
link |
And I do think there are some deep things
link |
around like sparse distributed memories and boosting.
link |
That's like, if you can just be slightly
link |
above the noise floor of having some hint of something,
link |
you can have things refined into pulling the memory back up.
link |
So having a, being a programmer and having a toolbox
link |
of like all of these things that,
link |
things that I did in all of these previous lives
link |
of programming tasks that still matters to me
link |
about how I'm able to pull up some of these things.
link |
Like in that case, it was something I did on the Apple 2
link |
then being relevant for the PC.
link |
And I have still cases when I would,
link |
when I would work on mobile development then be like,
link |
okay, I did something like this back in the doom days,
link |
but now it's a different environment,
link |
but I have still had that tie.
link |
I can bring it in and I can transform it
link |
into what the world needs right now.
link |
And I do think that's actually one of the very core things
link |
with human cognition and brain like,
link |
you know, brain like functioning
link |
is finding these ways about you've got,
link |
your brain is kind of everything everywhere all at once.
link |
You know, it's, it is just a set of all of this stuff
link |
that is just fetched back by these queries that go into it.
link |
And they can just be slightly above the noise floor
link |
with a random noise in your neurons and synapses
link |
that are affecting exactly what gets pulled up.
link |
So you're saying some of these very specific solutions
link |
for different games,
link |
you find that there's a kernel of a deep idea
link |
that's generalizable to other, to other things.
link |
Yeah, you can't predict what it's going to be,
link |
but that idea of like,
link |
I called out that compiled shaders in the forward
link |
that I wrote for that, the game engine black book,
link |
as you know, this is,
link |
it's kind of an end point of unrolling code,
link |
but that's one of those things that thinking about that
link |
and having that in your mind.
link |
And I'm sure there are some programmers
link |
that, you know, hear about that.
link |
Think about it a little bit.
link |
It's kind of the mind blown moment.
link |
It's like, oh, you can just turn all of that data into code.
link |
And nowadays, you know, you have instruction cache issues
link |
and that's not necessarily the best idea,
link |
but there are different,
link |
it's an idea that has power
link |
and has probably relevance in some other areas.
link |
Maybe it's in a hardware point of view
link |
that there's a way you approach building hardware
link |
that has that same,
link |
you don't even have to think about iterating.
link |
You just bake everything all the way into it in one place.
link |
What is the story of how you came to program Doom?
link |
What are some memorable technical challenges
link |
or innovations within that game?
link |
So the path that we went after Wolfenstein got out
link |
and we were on this crazy arc
link |
where Keen one through three, more success than we thought,
link |
Keen four through six, even more success,
link |
Wolfenstein even more success.
link |
So we were on this crazy trajectory for things.
link |
So actually our first box commercial project
link |
was the Commander Keen game,
link |
but then Wolfenstein was going to have a game called
link |
Spear of Destiny, which was a commercial version,
link |
So the rest of the team took the game engine
link |
pretty much as it was and started working on that.
link |
We got new monsters,
link |
but it's basically reskins of the things there.
link |
And there's a really interesting aspect about that
link |
that I didn't appreciate until much, much later
link |
about how Wolfenstein clearly did tap out its limit
link |
about what you want to play,
link |
all the levels and a couple of our license things.
link |
There was a hard creative wall
link |
that you did not really benefit much
link |
by continuing to beat on it.
link |
But a game like Doom and other more modern games
link |
like Minecraft or something,
link |
there's kind of a Turing completeness level
link |
of design freedom that you get in games
link |
that Wolfenstein clearly sat on one side of.
link |
All the creative people in the world
link |
could not go and do a masterpiece
link |
just with the technology that Wolfenstein had.
link |
Wolfenstein could do Wolfenstein,
link |
but you really couldn't do something crazy and different.
link |
But it didn't take that much more capability
link |
to get to Wolfenstein with the freeform lines
link |
and a little bit more artistic freedom
link |
to get to the point where people still announce
link |
new Doom levels today, all these years after
link |
without having completely tapped out the creativity.
link |
How did you put it?
link |
Yeah, Turing complete design space.
link |
Where it's like, you know,
link |
we have the kind of computational universality
link |
on a lot of things and how different
link |
subsurface work. For creativity.
link |
But yeah, there's things where a box can be too small,
link |
but above a certain point,
link |
you kind of are at the point where you really have
link |
almost unbounded creative ability there.
link |
And Doom was the first time you crossed that line.
link |
Yeah, where there were thousands of Doom levels created,
link |
and some of them still have something new and interesting
link |
to say to the world about it.
link |
Is that line, can you introspect what that line was?
link |
Is it in the design space?
link |
Is it something about the programming capabilities
link |
that you were able to add to the game?
link |
So the graphics fidelity was a necessary part
link |
because the block limitations in Wolfenstein,
link |
what we had right there was not enough.
link |
The full scale blocks, although Minecraft really did show
link |
that perhaps blocks stacked in 3D
link |
and at one quarter of the scale of that,
link |
or one eighth in volume,
link |
is then sufficient to have all of that.
link |
But the wall sized blocks that we had in Wolfenstein
link |
was too much of a creative limitation.
link |
We licensed the technology to a few other teams.
link |
None of them made too much of a dent with that.
link |
It just wasn't enough creative ability.
link |
But a little bit more,
link |
whether it was the variable floors and ceilings
link |
and arbitrary angles in Doom,
link |
or the smaller voxel blocks in Minecraft,
link |
is then enough to open it up
link |
to just worlds and worlds of new capabilities.
link |
What is binary space partitioning?
link |
So the. Which is one of the technologies.
link |
Yeah, so jump around a little bit on the story path there.
link |
So while the team was working on Spirit Destiny
link |
for Wolfenstein, we had met another development team,
link |
Raven Software, while we were in Wisconsin.
link |
And they were doing, they had RPG background
link |
and I still kind of loved that.
link |
And I offered to do a game engine for them
link |
to let them do a 3D rendered RPG
link |
instead of the, like most RPG games were kind of hand drawn.
link |
They made it look kind of 3D,
link |
but it was done just all with artist work
link |
rather than a real engine.
link |
And after Wolfenstein, this was still a tile based world,
link |
but I added floors and ceilings and some lighting
link |
and the ability to have some sloped floors
link |
in different areas.
link |
And that was my intermediate step
link |
for a game called Shadowcaster.
link |
And it had slowed down enough.
link |
It was not fast enough to do our type of action things.
link |
So they had the screen crop down a little bit.
link |
So you couldn't go the full screen width
link |
like we would try to do in Wolfenstein, but I learned a lot.
link |
I got the floors and ceilings and lightings
link |
and it looked great.
link |
They were great artists up there.
link |
And it was an inspiration for us
link |
to look at some of that stuff.
link |
But I had learned enough from that,
link |
that I had the plan for,
link |
I knew faster ways to do the lighting and shadowing.
link |
And I wanted to do this free form geometry.
link |
I wanted to break out of this tile based
link |
90 degree world limitations.
link |
So we had, that was when we got our next stations
link |
and we were working with these higher powered systems.
link |
And we built an editor that let us draw
link |
kind of arbitrary line segments.
link |
And I was working hard to try to make something
link |
that could render this fast enough.
link |
I was pushing myself pretty hard.
link |
And we were at a point where we could see some things
link |
that looked amazingly cool,
link |
but it wasn't really fast enough for the way I was doing it.
link |
For this flexibility, it was no longer,
link |
I couldn't just ray cast into it.
link |
And I had these very complex sets of lines
link |
and simple little worlds were okay.
link |
But the cool things that we wanted to do
link |
just weren't quite fast enough.
link |
And I wound up taking a break at that point.
link |
And I did the port.
link |
I did two ports of our games,
link |
Wolfenstein to the Super Nintendo.
link |
It was a crazy difficult thing to do,
link |
which was an even slower processors,
link |
like a couple of megahertz processor.
link |
And it had been this whole thing
link |
where we had farmed out the work and it wasn't going well.
link |
And I took it back over
link |
and trying to make it go fast on there
link |
where it really did not have much processing power.
link |
The pixels were stretched up hugely
link |
and it was pretty ugly when you looked at it.
link |
But in the end, it did come out fast enough to play
link |
and still be kind of fun from that.
link |
But that was where I started using BSP trees
link |
or binary space partitioning trees.
link |
It was one of those things I had to make it faster there.
link |
It was a stepping stone where it was reasonably easy
link |
to understand in the grid world of Wolfenstein
link |
where it was all still 90 degree angles.
link |
BSP trees were, I eased myself into it with that.
link |
And it was a big success.
link |
Then when I came back to working on Doom,
link |
I had this new tool in my toolbox.
link |
It was gonna be a lot harder
link |
with the arbitrary angles of Doom.
link |
This was where I really started grappling
link |
with epsilon problems.
link |
And just up until that point,
link |
I hadn't really had to deal with the fact
link |
that I am so many numeric things.
link |
This almost felt like a betrayal to me
link |
where people had told me that I had mathematicians
link |
up on a bit of a pedestal where I was,
link |
people think I'm a math wizard and I'm not.
link |
I really, everything that I did was really done
link |
with a solid high school math understanding.
link |
Algebra two trigonometry and that was what got me
link |
all the way through Doom and Quake and all of that,
link |
of just understanding basics of matrices
link |
and knowing it well enough to do something with it.
link |
What's the epsilon problems you ran into?
link |
So when you wind up taking like a sloped line
link |
and you say, I'm going to intersect it
link |
with another sloped line,
link |
then you wind up with something
link |
that's not going to be on these nice grid boundaries.
link |
With the Wolfenstein tile maps,
link |
all you've got is horizontal and vertical lines
link |
looking at it from above.
link |
And if you cut one of them, it's just obvious
link |
the other one gets cut exactly at that point.
link |
But when you have angled lines,
link |
you're doing a kind of a slope intercept problem
link |
and you wind up with rational numbers there
link |
where things that are not going to evenly land on an integer
link |
or even on any fixed point value that you've got.
link |
So everything winds up having to snap
link |
to some fixed point value.
link |
So the lines slightly change their angle.
link |
You wind up, if you cut something here,
link |
this one's going to bend a little this way
link |
and it's not going to be completely straight.
link |
And then you come down to all these questions of,
link |
well, this one is a point on an angled line.
link |
You can't answer that in finite precision
link |
unless you're doing something with actual rational numbers.
link |
And later on, I did waste far too much time
link |
chasing things like that.
link |
How do you do precise arithmetic with rational numbers?
link |
And it always blows up eventually, exponentially
link |
So these kinds of things are impossible with computers.
link |
So they're possible.
link |
Again, there are paths to doing it,
link |
but you can't fit them conveniently
link |
in any of the numbers.
link |
You need to start using big nums
link |
and different factor trackings of different things.
link |
So you have to, if you have any elements of OCD
link |
and you want to do something perfectly,
link |
you're screwed if you're working with floating point.
link |
So you had to deal with this for the first time.
link |
And there were lots of challenges there about like,
link |
okay, they build this cool thing.
link |
And the way the BSP trees work is it basically
link |
takes the walls and it carves other walls by those walls
link |
in this clever way that you can then
link |
take all of these fragments.
link |
And then you can for sure, from any given point,
link |
get an ordering of everything in the world.
link |
And you can say, this goes in front of this,
link |
goes in front of this, all the way back to the last thing.
link |
And that's super valuable for graphics
link |
where kind of a classic graphics algorithm
link |
would be painter's algorithm.
link |
You paint the furthest thing first,
link |
and then the next thing, and then the next thing,
link |
and then it comes up and it's all perfect for you.
link |
That's slow because you don't want to have to have drawn
link |
everything like that,
link |
but you can also flip it around
link |
and draw the closest thing to you.
link |
And then if you're clever about it,
link |
you can figure out what you need to draw
link |
that's visible beyond that.
link |
And that's what BSP trees allow you to do.
link |
Yeah, so it's combined with a bunch of other things,
link |
but it gives you that ordering.
link |
It's a clever way of doing things.
link |
And I remember I had learned this
link |
from one of my graphics Bible at the time,
link |
a book called Foley and Van Damme.
link |
And again, it was a different world back there.
link |
There was a small integer number of books.
link |
And this book that was big fat college textbook
link |
that I had read through many times.
link |
I didn't understand everything in it.
link |
Some of it wasn't useful to me,
link |
but they had the little thing about finite orderings
link |
of you draw a little T shaped thing
link |
and you can say you can make a fixed ahead of time order
link |
from this and you can generalize this with the BSP trees.
link |
And I got a little bit more information about that.
link |
And it was kind of fun later while I was working on Quake,
link |
I got to meet Bruce Naylor,
link |
who is one of the original researchers
link |
that developed those technologies for academic literature.
link |
And that was kind of fun,
link |
but I was very much just finding a tool
link |
that can help me solve what I was doing.
link |
And I was using it in this very crude way
link |
in a two dimensional fashion, rather than the general 3D.
link |
The Epsilon problems got much worse
link |
in Quake and three dimensionals
link |
when things angle in every way.
link |
But eventually I did sort out
link |
how to do it reliably on Doom.
link |
There were still a few edge cases in Doom
link |
that were not absolutely perfect
link |
where they even got terminologies in the communities.
link |
Like when you got to something where it was messed up,
link |
it was a hall of mirrors effect
link |
because you'd sweep by and it wouldn't draw something there.
link |
And you would just wind up with the leftover remnants
link |
as you flipped between the two pages.
link |
But BSP trees were important for it.
link |
But it's again worth noting that after we did Doom,
link |
our major competition came from Ken Silverman
link |
and his build engine, which was used for Duke Nukem 3D
link |
and some of the other games for 3D Realms.
link |
And he used a completely different technology,
link |
nothing to do with BSP trees.
link |
So there's not just a one true way of doing things.
link |
There were critical things about
link |
to make any of those games fast.
link |
You had to separate your drawing into,
link |
you drew vertical lines and you drew horizontal lines,
link |
just kind of changing exactly
link |
what you would draw with them.
link |
That was critical for the technologies at that time.
link |
And like all the games that were kind of like that
link |
wound up doing something similar,
link |
but there were still a bunch of other decisions
link |
that could be made.
link |
And we made good enough decisions on everything on Doom.
link |
We brought in multiplayer significantly
link |
and it was our first game that was designed
link |
to be modified by the user community
link |
where we had this whole setup of our WAD files and PWADs
link |
and things that people could build with tools
link |
that we released to them.
link |
And they eventually rewrote to be better
link |
than what we released, but they could build things
link |
and you could add it to your game
link |
without destructively modifying it,
link |
which is what you had to do in all the early games.
link |
You literally hacked the data files or the executable
link |
before while Doom was set up in this flexible way
link |
so that you could just say,
link |
run the normal game with this added on on top
link |
and it would overlay just the things
link |
that you wanted to there.
link |
Would you say that Doom was kind of the first
link |
true 3D game that you created?
link |
So no, it's still, Doom would usually be called
link |
a two and a half D game
link |
where it had three dimensional points on it.
link |
And this is another one of these kind of pedantic things
link |
that people love to argue about,
link |
about what was the first 3D game.
link |
I still, like every month probably I hear from somebody
link |
about, well, was Doom really a 3D game or something?
link |
And I give the point where characters had three coordinates.
link |
So you had like an X, Y and Z,
link |
the cacodamon could be coming in very high
link |
and come down towards you.
link |
The walls had three coordinates on them.
link |
So on some sense it's a 3D game engine,
link |
but it was not a fully general 3D game engine.
link |
You could not build a pyramid in Doom
link |
because you couldn't make a sloped wall,
link |
which was slightly different
link |
where in that previous shadow caster game,
link |
I could have vertexes and have a sloped floor there,
link |
but the changes that I made for Doom to get higher speed
link |
and a different set of flexibility traded away that ability,
link |
but you literally couldn't make that.
link |
You could not, you could make different heights of passages,
link |
but you could not make a bridge over another area.
link |
You could not go over and above it.
link |
So it still had some 2D limitations to it.
link |
That's more about the building
link |
versus the actual experience.
link |
Cause the experience is.
link |
It felt like things would come at you,
link |
but again, you couldn't look up either.
link |
I am, you know, you could only pitch.
link |
It was four degrees of freedom
link |
rather than six degrees of freedom.
link |
You did not have the ability to tilt your head this way
link |
or pitch up and down.
link |
So that takes us to Quake.
link |
What was the leap there?
link |
What was some fascinating technical challenges
link |
and there were a lot or not challenges,
link |
but innovations that you've come up with.
link |
So Quake was kind of the first thing
link |
where I did have to kind of come face to face
link |
with my limitations,
link |
where it was the first thing
link |
where I really did kind of give it my all
link |
and still come up, you know, come up a little bit short
link |
in terms of what and when I wanted to get it done.
link |
And the company ran,
link |
they had some serious stresses through the whole project
link |
and we bit off a lot.
link |
So the things that we set out to do
link |
was it was going to be really a true 3D engine
link |
where it could do six degree of freedom.
link |
You could have all the viewpoints.
link |
You could model anything.
link |
It had a really remarkable new lighting model
link |
with the surface caching and things.
link |
That was one of those where it was starting
link |
to do some things that they weren't doing
link |
even on the very high end systems.
link |
And it was going to be completely programmable
link |
in the modding standpoint,
link |
where the thing that you couldn't do in Doom,
link |
you could replace almost all of the media,
link |
but you couldn't really change the game.
link |
There were still some people that were doing
link |
the hex setting of the executable,
link |
the dehacked things where you could change
link |
a few things about rules
link |
and people made some early capture the flag type things
link |
by hacking the executable,
link |
but it wasn't really set out to do that.
link |
Quake was going to have its own programming language
link |
that the game was going to be implemented in it.
link |
And that would be able to be overwritten
link |
just like any of the media.
link |
Code was going to be data for that.
link |
And you would be able to have expansion packs
link |
that changed fundamental things and mods and so on.
link |
And the multiplayer was going to be playable
link |
over the internet.
link |
It was going to support a client server
link |
rather than peer to peer.
link |
So we had the possibility of supporting
link |
larger numbers of players in disparate locations
link |
with this full flexibility of the programming overrides
link |
with full six degree of freedom modeling and viewing.
link |
And with this fancy new light mapped
link |
kind of surface caching side.
link |
And this was one of those things that
link |
if I could go back and tell younger me
link |
to do something differently,
link |
it would have been to split those innovations up
link |
into two phases in two separate games.
link |
It will be phase one and phase two.
link |
So it probably would have been taking
link |
the Doom rendering engine and bringing in
link |
the TCP IP client server.
link |
Focusing on the multiplayer.
link |
And the Quake C or would have been Doom C
link |
programming language there.
link |
So I would have split that into programming language
link |
and networking with the same Doom engine
link |
rather than forcing everybody to go towards
link |
the Quake engine, which really meant getting a Pentium.
link |
While it ran on a 486, it was not a great experience there.
link |
We could have made more people happier
link |
and gotten two games done in 50% more time.
link |
So speaking of people happier, our mutual friend,
link |
Joe Rogan, it seems like the most important moment
link |
of his life is centered around Quake.
link |
So it was a definitive part of his life.
link |
So would he agree with your thinking that they should split?
link |
So he is a person who loves Quake
link |
and played Quake a lot.
link |
Would he agree that you should have done the Doom engine
link |
and focus on the multiplayer for phase one?
link |
Or in your looking back, is the 3D world
link |
that Quake created was also fundamental
link |
to the enriching experience?
link |
You know, I would say that what would have happened
link |
is you would have had a Doom looking
link |
but Quake feeling game eight months earlier
link |
and then maybe six months after Quake actually shipped,
link |
then there would have been the full running on a Pentium,
link |
six degree of freedom graphics engine type things there.
link |
So it's not that it wouldn't have been there.
link |
It would have been something amazingly cool earlier
link |
and then something even cooler somewhat later
link |
where I would much rather have gone
link |
and done two one year development efforts.
link |
I've cycled them through.
link |
I've been a little more pragmatic about that
link |
rather than killing us ourselves on the whole Quake
link |
But I would say it's obviously things worked out well
link |
in the end, but looking back and saying,
link |
how would I optimize and do things differently?
link |
That did seem to be a clear case where I going ahead
link |
and we had enormous momentum on Doom.
link |
We did Doom two as the kind of commercial boxed version
link |
after our shareware success with the original,
link |
but we could have just made another Doom game
link |
adding those new features in.
link |
It would have been huge.
link |
We would have learned all the same lessons, but faster.
link |
And it would have given six degree of freedom
link |
and Pentium class systems a little bit more time
link |
to get mainstream because we did cut out a lot of people
link |
with the hardware requirements for Quake.
link |
Was there any dark moments for you personally,
link |
psychologically in having such harsh deadlines
link |
and having to solve so many difficult technical challenges?
link |
So I've never really had really dark black places.
link |
I mean, I can't necessarily put myself
link |
in anyone else's shoes,
link |
but I understand a lot of people have significant challenges
link |
with kind of their mental health and wellbeing.
link |
And I've been super stressed.
link |
I've been unhappy as a teenager in various ways,
link |
but I've never really gone to a very dark place.
link |
I just seem to be largely immune to what really wrecks people.
link |
I mean, I've had plenty of time when I'm very unhappy
link |
and miserable about something,
link |
but it's never hit me like,
link |
I believe it winds up hitting some other people.
link |
I've borne up well under whatever stresses
link |
have kind of fallen on me.
link |
And I've always coped best on that
link |
when all I need to do is usually
link |
just kind of bear down on my work.
link |
I pull myself out of whatever hole I might be slipping into
link |
by actually making progress.
link |
I mean, maybe if I was in a position
link |
where I was never able to make that progress,
link |
I could have slid down further,
link |
but I've always been in a place where,
link |
okay, a little bit more work,
link |
maybe I'm in a tough spot here,
link |
but I always know if I just keep pushing,
link |
eventually I break through and I make progress,
link |
I feel good about what I'm doing.
link |
And that's been enough for me so far in my life.
link |
Have you seen it in the distance,
link |
like ideas of depression or contemplating suicide?
link |
Have you seen those things far?
link |
So what was interesting, when I was a teenager,
link |
I was probably on some level a troubled youth.
link |
I was unhappy most of my teenage years.
link |
I really, I wanted to be on my own
link |
doing programming all the time.
link |
As soon as I was 18, 19, even though I was poor,
link |
I was doing exactly what I wanted and I was very happy,
link |
but high school was not a great time for me.
link |
And I had a conversation with like the school counselor
link |
and they're kind of running their script.
link |
It's like, okay, it's kind of a weird kid here.
link |
Let's carefully probe around.
link |
It's like, do you ever think about ending it all?
link |
I'm like, no, of course not.
link |
Never, not at all.
link |
This is temporary, things are going to be better.
link |
And that's always been kind of the case for me.
link |
And obviously that's not that way for everyone
link |
and other people do react differently.
link |
What was your escape from the troubled youth,
link |
like music, video games, books?
link |
How did you escape from a world
link |
that's full of cruelty and suffering and that's absurd?
link |
Yeah, I mean, I was not a victim of cruelty and suffering.
link |
It's like, I was an unhappy, somewhat petulant youth
link |
in my point where I'm not putting myself up
link |
with anybody else's suffering,
link |
but I was unhappy objectively.
link |
And the things that I did
link |
that very much characterized my childhood
link |
were I had books, comic books,
link |
Dungeons and Dragons, arcade games, video games.
link |
Like some of my fondest childhood memories
link |
are the convenience stores, the 711s and Quick Trips,
link |
because they had a spinner rack of comic books
link |
and they had a little side room
link |
with two or three video games, arcade games in it.
link |
And that was very much my happy place.
link |
If I could, I get my comic books
link |
and if I could go to a library
link |
and go through those little zero, zero, zero section
link |
where computer books were supposed to be.
link |
And there were a few sad little books there,
link |
but still just being able to sit down and go through that.
link |
And I read a ridiculous number of books,
link |
both fiction and nonfiction as a teenager.
link |
And my rebelling in high school
link |
was just sitting there with my nose in a book,
link |
ignoring the class through lots of it.
link |
And teachers had a range of reactions to that,
link |
some more accepting of it than others.
link |
I'm with you on that.
link |
So let us return to Quake for a bit
link |
with the technical challenges.
link |
What everything together from the networking
link |
to the graphics, what are some things you remember
link |
that were innovations you had to come up with
link |
in order to make it all happen?
link |
Yeah, so there were a bunch of things on Quake
link |
where on the one hand,
link |
the idea that I built my own programming language
link |
to implement the game in.
link |
Looking back, and I try to tell people,
link |
it's like every high level programmer
link |
sometime in their career goes through
link |
and they invent their own language.
link |
It just seems to be a thing that's pretty broadly done.
link |
People will be like,
link |
I'm gonna go write a computer programming language.
link |
And I don't regret having done it,
link |
but after that, I switched from Quake C,
link |
my quirky little pseudo object oriented
link |
or entity oriented language there.
link |
Quake two went back to using DLLs with C
link |
and then Quake three,
link |
I implemented my own C interpreter or compiler,
link |
which was a much smarter thing to do
link |
that I should have done originally for Quake.
link |
But building my own language was an experience.
link |
I learned a lot from that.
link |
And then there was a generation of game programmers
link |
that learned programming with Quake C,
link |
which I feel kind of bad about,
link |
because we give JavaScript a lot of crap,
link |
but Quake C was nothing to write home about there.
link |
But it allowed people to do magical things.
link |
You get into programming,
link |
not because you love the BNF syntax of a language,
link |
it's because the language lets you do something
link |
that you cared about.
link |
And here's very much,
link |
you could do something
link |
in a whole beautiful three dimensional world.
link |
And the idea and the fact that the code
link |
for the game was out there,
link |
you could say, I liked the shotgun,
link |
but I want it to be more bad ass.
link |
You go in there and say,
link |
okay, now it does 200 points damage.
link |
And then you go around with a big grin on your face,
link |
blowing up monsters all over the game.
link |
So yeah, it is not what I would do today
link |
going back with that language,
link |
but that was a big part of it.
link |
Learning about the networking stuff,
link |
because it's interesting where I learned these things
link |
So I would get a book on networking and find something,
link |
I read all about it and learn, okay, packets,
link |
they can be out of order or lost or duplicated.
link |
These are all the things
link |
that can theoretically happen to packets.
link |
So I wind up spending all this time thinking about
link |
how do we deal about all of that?
link |
And it turns out, of course, in the real world,
link |
those are things that yes,
link |
theoretically can happen with multiple routes,
link |
but they really aren't things that your 99.999%
link |
of your packets have to deal with.
link |
So there was learning experiences about lots of that.
link |
Like why, when TCP is appropriate versus UDP
link |
and how if you do things in UDP,
link |
you wind up reinventing TCP badly in almost all cases.
link |
So there's good arguments for using both
link |
for different game technology,
link |
different parts of the game process,
link |
transitioning from level to level and all.
link |
But the graphics were the showcase
link |
of what Quake was all about.
link |
It was this graphics technology that nobody had seen there.
link |
And it was a while before
link |
there were competitive things out there.
link |
And it went a long time internally really not working
link |
where we were even building levels
link |
where the game just was not at all shippable
link |
with large fractions of the world,
link |
like disappearing, not being there,
link |
or being really slow in various parts of it.
link |
And it was this act of faith.
link |
It's like, I think I'm gonna be able to fix this.
link |
I think I'm gonna be able to make this work.
link |
And lots of stuff changed
link |
where the level designers would build something
link |
and then have to throw it away as something fundamental
link |
and the kind of graphics or level technology changed.
link |
And so there were two big things
link |
that contributed to making it possible at that timeframe.
link |
There was certainly
link |
hardcore optimized low level assembly language.
link |
And this was where I had hired Michael Abrash
link |
away from Microsoft.
link |
And he had been one of my early inspirations
link |
where back in the soft disk days,
link |
the library of magazines that they had,
link |
some of my most treasured ones
link |
were Michael Abrash's articles in Dr. Dobbs journal.
link |
And it was amazing after all of our success in Doom,
link |
we were able to kind of hit him up and say,
link |
hey, we'd like you to come work at id Software.
link |
And he was in the senior technical role at Microsoft.
link |
And he was on track for,
link |
and this was right when Microsoft was starting to take off.
link |
And I did eventually convince him
link |
that what we were doing
link |
was gonna be really amazing with Quake.
link |
It was gonna be something nobody had seen before.
link |
It had these aspects of what we were talking about.
link |
We had metaverse talk back then.
link |
We had read Snow Crash and we knew about this.
link |
And Michael was big into the science fiction
link |
and we would talk about all that
link |
and kind of spin this tale.
link |
And it was some of the same conversations
link |
that we have today about the metaverse,
link |
about how you could have different areas
link |
linked together by portals
link |
and you could have user generated content
link |
and changing out all of these things.
link |
So you really were creating the metaverse with Quake.
link |
And we talked about things like,
link |
Duke used to be advertised as a virtual reality experience.
link |
That was the first wave of virtual reality
link |
was in the late 80s and early 90s,
link |
you had like the Lawn Mower Man movie
link |
and you had time in Newsweek
link |
talking about the early VPL headsets.
link |
And of course that cratered so hard
link |
that people didn't wanna look at virtual reality
link |
for decades afterwards,
link |
where it was just, it was smoke and mirrors.
link |
It was not real in the sense
link |
that you could actually do something
link |
real and valuable with it.
link |
But still we had that kind of common set of talking points.
link |
And we were talking about what these games could become
link |
and how you'd like to see people
link |
building all of these creative things.
link |
Because we were seeing an explosion of work
link |
with Doom at that time,
link |
where people were doing amazingly cool things.
link |
Like we saw cooler levels that we had built
link |
coming out of the user community.
link |
And then people finding ways
link |
to change the characters in different ways.
link |
And we knew what we were doing in Quake
link |
was removing those last things.
link |
There was some quirky things
link |
with a couple of the data types
link |
that didn't work right for overriding.
link |
And then the core thing about the programming model.
link |
And I was definitely going to hit all of those in Quake.
link |
But the graphics side of it was still,
link |
I knew what I wanted to do.
link |
And it was one of these hubris things
link |
where it's like, well, so far I've been able
link |
to kind of kick everything that I set out to go do.
link |
But Quake was definitely a little bit more
link |
than could be comfortably chewed at that point.
link |
And, but Michael was one of the strongest programmers
link |
and graphics programmers that I knew.
link |
And he was one of the people that I trusted
link |
to write assembly code better than I could.
link |
And there's a few people that I can point to
link |
about things like this where I'm a world class optimizer.
link |
I mean, I make things go fast,
link |
but I recognize there's a number of people
link |
that can write tighter assembly code,
link |
tighter SIMD code or tighter CUDA code
link |
My best strengths are a little bit more at the system level.
link |
I mean, I'm good at all of that,
link |
but the most leverage comes from making the decisions
link |
that are a little bit higher up
link |
where you figure out how to change your large scale problems
link |
so that these lower level problems are easier to do
link |
or it makes it possible to do them in a uniquely fast way.
link |
So most of my big wins in a lot of ways
link |
from all the way from the early games through VR
link |
and the aerospace work that I'm doing or did,
link |
and hopefully the AI work that I'm working on now
link |
is finding an angle on something
link |
that means you trade off something
link |
that you maybe think you need,
link |
but it turns out you don't need.
link |
And by making a sacrifice in one place,
link |
you can get big advantages in another place.
link |
Is it clear at which level of the system
link |
those big advantages can be gained?
link |
It's not always clear.
link |
And that's why the thing that I try
link |
to make one of my core values,
link |
and I proselytize to a lot of people
link |
is trying to know the entire stack,
link |
trying to see through everything that happens.
link |
And it's almost impossible
link |
on like the web browser level of things
link |
where there's so many levels to it,
link |
but you should at least understand what they all are,
link |
even if you can't understand
link |
all the performance characteristics at each level,
link |
but it goes all the way down to literally the hardware.
link |
So what is this chip capable of?
link |
And what is this software that you're writing capable of?
link |
And then with this architecture you put on top of that,
link |
then the ecosystem around it,
link |
all the people that are working on it.
link |
So there are all these decisions
link |
and they're never made in a globally optimal way,
link |
but sometimes you can drive a thread
link |
of global optimality through it.
link |
You can't look at everything, it's too complicated,
link |
but sometimes you can step back up
link |
and make a different decision.
link |
And we kind of went through this on the graphic side
link |
on Quake, where in some ways it was kind of bad
link |
where Michael would spend his time writing,
link |
like I'd rough out the basic routines,
link |
like, okay, here's our span rasterizer.
link |
And he would spend a month writing
link |
this beautiful cycle optimized piece of assembly language
link |
that does what I asked it to do.
link |
And he did it faster than like my original code would do,
link |
or probably what I would be able to do
link |
even if I had spent that month on it.
link |
But then we'd have some cases when I'd be like,
link |
okay, well, I figured out at this higher level
link |
instead of drawing these in a painter's order here,
link |
I do a span buffer and it cuts out 30% or 40%
link |
of all of these pixels,
link |
but it means you need to rewrite kind of this interface
link |
And I could tell that wore on him a little bit,
link |
but in the end it was the right thing to do
link |
where we wound up changing that rasterization approach
link |
and we wound up with a super optimized
link |
assembly language core loop
link |
and then a good system around it
link |
which minimized how much that had to be called.
link |
And so in order to be able to do
link |
this kind of system level thinking,
link |
whether we're talking about game development,
link |
aerospace, nuclear energy, AI, VR,
link |
you have to be able to understand the hardware,
link |
the low level software, the high level software,
link |
the design decisions, the whole thing,
link |
the full stack of it.
link |
Yeah, and that's where a lot of these things
link |
When you're really, when you're bringing the future forward,
link |
I mean, there's a pace that everything
link |
just kind of glides towards
link |
where we have a lot of progress
link |
that's happening at such a different,
link |
so many different ways you kind of slide towards progress
link |
just left to your own, programs just get faster.
link |
For a while it wasn't clear
link |
if they were gonna get fatter more than they get,
link |
quicker than they get faster and it cancels out,
link |
but it is clear now in retrospect,
link |
no, programs just get faster
link |
and have gotten faster for a long time.
link |
But if you wanna do something
link |
like back at that original talking about scrolling games,
link |
say what, this needs to be five times faster.
link |
Well, we can wait six years
link |
and just it'll naturally get that much faster at that time
link |
or you come up with some really clever way of doing it.
link |
So there are those opportunities like that
link |
in a whole bunch of different areas.
link |
Now, most programmers don't need to be thinking about that.
link |
There's not that many,
link |
there's a lot of opportunities for this,
link |
but it's not everyone's work a day type stuff.
link |
So everyone doesn't have to know how all these things work.
link |
They don't have to know how their compiler works,
link |
how the processor chip manages cache eviction
link |
and all these low level things.
link |
But sometimes there are powerful opportunities
link |
that you can look at and say,
link |
we can bring the future five years faster.
link |
We can do something that,
link |
wouldn't it be great if we could do this?
link |
Well, we can do it today
link |
if we make a certain set of decisions.
link |
And it is in some ways smoke and mirrors
link |
where you say it's like,
link |
Doom was a lot of smoke and mirrors
link |
where people thought it was more capable
link |
than it actually was,
link |
but we picked the right smoke and mirrors
link |
to deploy in the game where by doing this,
link |
people will think that it's more general.
link |
We are gonna amaze them with what they've got here
link |
and they won't notice
link |
that it doesn't do these other things.
link |
So smart decision making at that point,
link |
that's where that kind of global holistic top down view
link |
And I'm really a strong believer
link |
that technology should be sitting at that table
link |
having those discussions
link |
because you do have cases where you say,
link |
well, you wanna be the Jonathan Ivy or whatever,
link |
where it's a pure design solution.
link |
And that's in some cases now
link |
where you truly have almost infinite resources.
link |
Like if you're trying to do a scrolling game on the PC now,
link |
you don't even have to talk to a technology person.
link |
You can just have,
link |
any intern can make that go run as fast as it needs to there
link |
and it can be completely design based.
link |
But if you're trying to do something that's hard,
link |
either that can't be done for resources
link |
like VR on a mobile chip set,
link |
or that we don't even know how to do yet,
link |
like artificial general intelligence,
link |
it's probably going to be a matter
link |
of coming at it from an angle.
link |
Like, I mean, for AGI,
link |
we have some of like what are some of the hudder principles
link |
about how you can AXI or some of the,
link |
there are theoretical ways that you can say,
link |
this is the optimal learning algorithm
link |
that can solve everything, but it's completely impractical.
link |
You just can't do that.
link |
So clearly you have to make some concessions
link |
for general intelligence
link |
and nobody knows what the right ones are yet.
link |
So people are taking different angles of attack.
link |
I hope I've got something clever
link |
to come up with in that space.
link |
It's been surprising to me.
link |
And I think perhaps it is a principle of progress
link |
that smoke and mirrors somehow
link |
is the way you build the future.
link |
You kind of fake it till you make it
link |
and you almost always make it.
link |
And I think that's going to be the way we achieve AGI.
link |
That's going to be the way we build consciousness
link |
into our machines.
link |
There's philosophers debate about the Turing test
link |
is essentially about faking it till you make it.
link |
You start by faking it.
link |
And I think that always leads to making it
link |
because if we look at history.
link |
Arguments when, as soon as people start talking about
link |
qualia and consciousness and Chinese rooms and things,
link |
it's like, I just check out.
link |
I just don't think there's any value in those conversations.
link |
It's just like, go ahead, tell me it's not going to work.
link |
I'm going to do my best to try to make it work anyways.
link |
I don't know if you work with legged robots.
link |
There's a bunch of these.
link |
They sure as heck make me feel like they're cautious
link |
in a certain way that's not here today,
link |
but is you could see the kernel.
link |
It's like the flame, the beginnings of a flame.
link |
We don't have line of sight,
link |
but there's glimmerings of light in the distance
link |
for all of these things.
link |
Yeah, I'm hearing murmuring in a distant room.
link |
Well, let me ask you a human question here.
link |
You've in the game design space,
link |
you've done a lot of incredible work throughout,
link |
but in terms of game design, you have changed the world.
link |
And there's a few people around you that did the same.
link |
So famously there's some animosity, there's much love,
link |
but there's some animosity between you and John Romero.
link |
What is at the core of that animosity and human tension?
link |
So there really hasn't been for a long time.
link |
And even at the beginning, it's like,
link |
yes, I did push Romero out of the company.
link |
And this is one of the things that I look back,
link |
if I could go back telling my younger self
link |
some advice about things,
link |
the original founding kind of corporate structure
link |
of id Software really led to a bunch of problems.
link |
We started off with us as equal partners
link |
and we had a buy sell agreement
link |
because we didn't want outsiders
link |
to be telling us what to do inside the company.
link |
And that did lead to a bunch of the problems
link |
where I was sitting here going,
link |
it's like, all right, I'm working harder than anyone.
link |
I'm doing these technologies, nobody's done before,
link |
but we're all equal partners.
link |
And then I see somebody that's not working as hard.
link |
And I mean, I can't say I was the most mature about that.
link |
I was 20 something years old
link |
and it did bother me when I'm like, everybody,
link |
okay, we need to all pull together
link |
and we've done it before everybody.
link |
We know we can do this if we get together
link |
and we grind it all out,
link |
but not everybody wanted to do that for all time.
link |
And I was the youngest one of the crowd there.
link |
I had different sets of kind of backgrounds and motivations
link |
and left at that point where it was, all right,
link |
either everybody has to be contributing like up to this level
link |
or they need to get pushed out was not,
link |
that was not a great situation.
link |
And I look back on it and know that we pushed people
link |
out of the company that could have contributed
link |
if there was a different framework for them.
link |
And the modern kind of Silicon Valley,
link |
like let your stock vest over a time period
link |
and maybe it's non voting stock
link |
and all those different things.
link |
We knew nothing about any of that.
link |
I mean, we didn't know what we were doing
link |
in terms of corporate structure or anything.
link |
So if you think the framework was different,
link |
some of the human tension could have been a little bit.
link |
It almost certainly would have.
link |
I mean, I look back at that
link |
and it's like even trying to summon up in my mind,
link |
it's like, I know I was really, really angry about,
link |
like Romero not working as hard as I wanted him to work
link |
or not carrying his load on the design for Quake
link |
and coming up with things there.
link |
But he was definitely doing things.
link |
He made some of the best levels there.
link |
He was working with some of our external teams
link |
like Raven on the licensing side of things,
link |
but there were differences of opinion about it,
link |
but he landed right on his feet.
link |
He went and he got $20 million from Eidos
link |
to go do Ion Storm and he got to do things his way
link |
and spun up three teams simultaneously
link |
because that was always one of the challenging things
link |
in it where we were doing these single string,
link |
one project after another.
link |
And I think some of them wanted to grow the company more
link |
and I didn't because I knew people that were saying that,
link |
oh, companies turn to shit when you got 50 employees.
link |
It's just a different world there.
link |
And I loved our little dozen people working on the projects,
link |
but you can look at it and say,
link |
well, business realities matter.
link |
It's like, you're super successful here
link |
and we could take a swing and a miss on something,
link |
but you do it a couple of times and you're out of luck.
link |
There's a reason companies try to have multiple teams
link |
running at one time.
link |
And so that was, again,
link |
something I didn't really appreciate back then.
link |
So if you look past all that,
link |
you did create some amazing things together.
link |
What did you love about John Romero?
link |
What did you respect and appreciate about him?
link |
What did you admire about him?
link |
What did you learn from him?
link |
When I met him, he was the coolest programmer
link |
He had done all of this stuff.
link |
He had made all of these games.
link |
He had worked at one of the companies
link |
that I thought was the coolest at Origin Systems.
link |
And he knew all this stuff.
link |
He made things happen fast.
link |
And he was also kind of a polymath about this
link |
where he could do, he drew his own art.
link |
He made his own levels,
link |
as well as we worked on sound design systems
link |
on top of actually being a really good programmer.
link |
And we went through a little,
link |
it was kind of fun where one of the early things
link |
that we did where there was kind of the young buck bit
link |
going in where I was the new guy.
link |
And he was the top man programmer at the soft disk area.
link |
And eventually we had sort of a challenge over the weekend
link |
that we were gonna like race to implement this game,
link |
to port one of our PC games back down to the Apple II.
link |
And that was where we finally kind of became clear.
link |
It's like, okay, Carmack stands a little bit apart
link |
on the programming side of things.
link |
But Romero then very gracefully moved into,
link |
well, he'll work on the tools, he'll work on the systems,
link |
do some of the game design stuff,
link |
as well as contributing on,
link |
starting to lead the design aspects of a lot of things.
link |
So he was enormously valuable in the early stuff.
link |
And so much of Doom and even Quake have his stamp on it
link |
But he wasn't at the same level of focus
link |
that I brought to the work that we were doing there.
link |
And he really did, we hit such a degree of success
link |
that it was all in the press about that.
link |
The rockstar game programmers.
link |
I mean, it's the Beatles problem.
link |
Yeah, I mean, he ate it up and he did personify.
link |
There was the whole game developers with Ferraris
link |
that we had there.
link |
And I thought that led to some challenges there.
link |
But so much of the stuff that was great in the games
link |
did come from him.
link |
And I would certainly not take that away from him.
link |
And even after we parted ways and he took his swing
link |
with Eidos, in some ways, he was ahead of the curve
link |
with mobile gaming as well, where
link |
one of his companies after Eidos was working on feature phone
link |
And I wound up doing some of that
link |
just before the iPhone crossing over into the iPhone phase
link |
And that was something that clearly
link |
did turn out to be a huge thing, although he
link |
was too early for what he was working on at that time.
link |
We've had pretty cordial relationships
link |
where I was happy to talk with him any time I'd
link |
run into him at a conference.
link |
I have actually had some other people just say,
link |
it's like, oh, you shouldn't go over there
link |
and give him the time of day.
link |
Or felt that Masters of Doom played things up in a way
link |
that I shouldn't be too happy with.
link |
But I'm OK with all of that.
link |
So you've still got love in your heart.
link |
Yeah, I mean, I just talked with him like last year,
link |
or I guess it was even this year,
link |
about mentioning that I'm going off doing this AI stuff.
link |
I'm going big into artificial intelligence.
link |
And he had a bunch of ideas for how
link |
AI is going to play into gaming.
link |
And asked if I was interested in collaborating.
link |
And it's not in line with what I'm doing.
link |
But I do wish almost everyone the best.
link |
I mean, I know I may not have partnered
link |
on the best of terms with some people.
link |
But I was thrilled to see Tom Hall writing VR games now.
link |
He wrote, I'm working on a game called Demio, which
link |
is really an awesome VR game.
link |
It's like Dungeons and Dragons.
link |
We all used to play Dungeons and Dragons together.
link |
That was one of the things.
link |
That was what we did on Sundays in the early days.
link |
I would Dungeon Master, and they'd all play.
link |
And so it really made me smile seeing
link |
Tom involved with an RPG game in virtual reality.
link |
You were the CTO of Oculus VR since 2013,
link |
and maybe less than a year involved in a bit in 2019.
link |
Oculus was acquired by Facebook Now Meta in 2014.
link |
You've spoken brilliantly about both the low level details,
link |
the experimental design, and the big picture
link |
vision of virtual reality.
link |
Let me ask you about the metaverse, the big question
link |
here, both philosophically and technically.
link |
How hard is it to build the metaverse?
link |
What is the metaverse in your view?
link |
You started with discussing and thinking about Quake
link |
as a kind of a metaverse.
link |
As you think about it today, what is the metaverse,
link |
the thing that could create this compelling user value,
link |
this experience that will change the world,
link |
and how hard is it to build it?
link |
So the term comes from Neal Stephenson's book Snow Crash,
link |
which many of us had read back in the 90s.
link |
It was one of those kind of formative books.
link |
And there was this sense that the possibilities
link |
and kind of the freedom and unlimited capabilities
link |
to build a virtual world that does whatever you want,
link |
whatever you ask of it, has been a powerful draw
link |
for generations of developers, game developers specifically,
link |
and people that are thinking about more general purpose
link |
So we were talking about that back in the Doom and Quake days
link |
about how do you wind up with an interconnected set of worlds
link |
that you kind of visit from one to another.
link |
And as web pages were becoming a thing,
link |
you start thinking about what is the interactive kind
link |
of 3D based equivalent of this.
link |
And there were a lot of really bad takes.
link |
You had like Vermol and virtual reality markup languages.
link |
And there's aspects like that that came from people saying,
link |
well, what kind of capabilities should we
link |
develop to enable this?
link |
And that kind of capability first work
link |
has usually not panned out very well.
link |
On the other hand, we have successful games
link |
that started with things like Doom and Quake and communities
link |
that formed around those, whether it
link |
was server lists in the early days
link |
or literal portaling between different games,
link |
and then modern things that are on completely different order
link |
of magnitude like Minecraft and Fortnite that
link |
have 100 million plus users.
link |
I still think that that's the right way
link |
to go to build the metaverse is you build something that's
link |
amazing that people love and people wind up
link |
spending all their time in because it's awesome.
link |
And you expand the capabilities of that.
link |
So even if it's a very basic experience, if it's awesome.
link |
Minecraft is an amazing case study in so many things.
link |
That's basic as it gets.
link |
What's been able to be done with that is really enlightening.
link |
And there are other cases where, like right now,
link |
Roblox is basically a game construction kit aimed at kids.
link |
And that was a capability first play.
link |
And it's achieving scale that's on the same order
link |
So it's not impossible, but my preferred bet
link |
would be you make something amazing that people love
link |
and you make it better and better.
link |
And that's where I could say we could have gone back
link |
and followed a path like that in the early days
link |
if you just take the same game, whether it's
link |
when Activision demonstrated that you could make
link |
Call of Duty every year.
link |
And not only is it not bad, people love it.
link |
And it's very profitable.
link |
The idea that you could have taken something like that,
link |
take a great game, release a new version every year
link |
that lets the capabilities grow and expand
link |
to start saying it's like, OK, it's
link |
a game about running around and shooting things,
link |
but now you can bring your media into it.
link |
You can add persistence of social signs of life
link |
or whatever you want to add to it.
link |
I still think that's quite a good position to take.
link |
And I think that while Meta is doing a bottoms up capability
link |
approach with Horizon Worlds where it's
link |
a fairly general purpose, creators
link |
can build whatever they want in their sort of thing,
link |
it's hard to compare and compete with something
link |
like Fortnite, which also has enormous amounts of creativity
link |
even though it was not designed originally
link |
as a general purpose sort of thing.
link |
So we have examples on both sides.
link |
Me personally, I would have bet on trying
link |
to do entertainment, valuable destination first,
link |
and expanding from there.
link |
So can you imagine the thing that will be kind of,
link |
if we look back a couple of centuries from now
link |
and you think about the experiences that
link |
marked the singularity, the transition where most
link |
of our world moved into virtual reality,
link |
what do you think those experiences will look like?
link |
So I do think it's going to be kind of like the way the web
link |
slowly took over, where you're the frog
link |
in the pot of water that's slowly heating up,
link |
where having lived through all of that,
link |
I remember when it was shocking to start
link |
seeing the first website address on a billboard when you're
link |
like, hey, my computer world is infecting the real world.
link |
This is spreading out in some way.
link |
But when you look back and say, well,
link |
what made the web take off?
link |
And it wasn't a big bang sort of moment there.
link |
It was a bunch of little things that turned out
link |
not to even be the things that are relevant now
link |
that brought them into it.
link |
So I wonder if, I mean, like you said, you're not a historian.
link |
So maybe there is a historian out there
link |
that could really identify that moment, data driven way.
link |
It could be like MySpace or something like that.
link |
Maybe the first major social network that really reached
link |
into non geek world or something like that.
link |
I think that's kind of the fallacy of historians, though,
link |
looking for some of those kind of primary dominant causes
link |
where so many of these things are like we
link |
see an exponential curve.
link |
But it's not because like one thing is going exponential.
link |
It's because we have hundreds of little sigmoid curves
link |
overlapped on top of each other.
link |
And they just happen to keep adding up
link |
so that you've got something kind of going exponential
link |
at any given point.
link |
But no single one of them was the critical thing.
link |
There were dozens and dozens of things.
link |
I mean, seeing the transitions of stuff
link |
like as obviously MySpace giving way to other things,
link |
but even like blogging giving way to social media
link |
and getting resurrected in other guises.
link |
And the memes with the dancing baby GIF
link |
or whatever the all your base now belong to us.
link |
Whatever those early memes that led to the modern memes
link |
and the different evolution of humor on the internet
link |
that I'm sure the historians will also write books about
link |
from the different website that support
link |
to create the infrastructure for that humor like Reddit
link |
and all that kind of stuff.
link |
So people will go back, and they will name
link |
firsts and critical moments.
link |
But it's probably going to be a poor approximation of what
link |
And we've already seen in the VR space
link |
where it didn't play out the way we thought it would in terms
link |
of what was going to be like when the modern era of VR
link |
basically started with my E3 demo of Doom 3
link |
on the Rift prototype.
link |
So we're like first person shooters in VR,
link |
match made in heaven, right?
link |
And that didn't work out that way at all.
link |
They have the most comfort problems with it.
link |
And then the most popular virtual reality app
link |
is Beat Saber, which nobody predicted back then.
link |
What's that make you like from first principles
link |
if you were to reverse engineer that?
link |
Why are these silly fun games the most?
link |
It actually makes very clear sense
link |
when you analyze it from hindsight
link |
and look at the engineering reasons
link |
where it's not just that it was a magical quirky idea.
link |
It was something that played almost perfectly
link |
to what turned out to be the real strengths of VR
link |
where the one thing that I really underestimated
link |
importance in VR was the importance of the controllers.
link |
I was still thinking we could do a lot more
link |
with the game pad and just the amazingness
link |
of taking any existing game,
link |
being able to move your head around and look around,
link |
that that was really amazing.
link |
But the controllers were super important.
link |
But the problem is so many things
link |
that you do with the controllers just suck.
link |
It feels like it breaks the illusion
link |
like trying to pick up glasses with the controllers
link |
where you're like, oh, use the grip button
link |
when you're kind of close and it'll snap into your hand.
link |
All of those things are unnatural actions
link |
that you do them and it's still part of the VR experience.
link |
But Beat Saber winds up playing only to the strengths.
link |
It completely hides all the weaknesses of it
link |
because you are holding something in your hand.
link |
You keep a solid grip on it the whole time.
link |
It slices through things without ever bumping into things.
link |
You never get into the point where,
link |
I'm knocking on this table,
link |
but in VR, my hand just goes right through it.
link |
So you've got something that slices through.
link |
So it's never your brain telling you,
link |
oh, I should have hit something.
link |
You've got a lightsaber here.
link |
It's just, you expect it to slice through everything.
link |
Audio and music turned out to be a really powerful aspect
link |
of virtual reality where you're blocking the world off
link |
and constructing the world around you.
link |
And being something that can run efficiently
link |
on even this relatively low powered hardware
link |
and can have a valuable loop in a small amount of time
link |
where a lot of modern games,
link |
you're supposed to sit down and play it for an hour
link |
just to get anywhere.
link |
Sometimes a new game takes an hour
link |
to get through the tutorial level.
link |
And that's not good for VR for a couple reasons.
link |
You do still have the comfort issues
link |
if you're moving around at all,
link |
but you've also got just discomfort from the headset,
link |
battery lifespan on the mobile versions.
link |
So having things that do break down
link |
into three and four minute windows of play,
link |
that turns out to be very valuable
link |
from a gameplay standpoint.
link |
So it winds up being kind of a perfect storm
link |
of all of these things that are really good.
link |
It doesn't have any of the comfort problems.
link |
You're not navigating around.
link |
You're standing still.
link |
All the stuff flies at you.
link |
It has placed audio strengths.
link |
It adds the whole fitness in VR.
link |
Nobody was thinking about that back at the beginning.
link |
And it turns out that that is an excellent
link |
daily fitness thing to be doing.
link |
If you go play an hour of Beat Saber
link |
or Supernatural or something,
link |
that is legit solid exercise.
link |
And it's more fun than doing it
link |
just about any other way there.
link |
So that's kind of the arcade stage of things.
link |
If I were to say with my experience with VR,
link |
the thing that I think is powerful is the,
link |
maybe it's not here yet,
link |
but the degree to which it is immersive
link |
in the way that Quake is immersive.
link |
It takes you to another world.
link |
For me, because I'm a fan of role playing games,
link |
the Elder Scrolls series,
link |
like Skyrim or even Daggerfall,
link |
it just takes you to another world.
link |
And when you're not in that world,
link |
you miss not being there.
link |
And then you just, you kind of want to stay there forever
link |
because life is shitty.
link |
And you just want to go to this place.
link |
Is that there was a time when we were kind of asked
link |
to come up with like, what's your view about VR?
link |
And my pitch was that it should be better
link |
inside the headset than outside.
link |
It's the world as you want it.
link |
And everybody thought that was dystopian.
link |
And like, that's like, oh,
link |
you're just going to forget about the world outside.
link |
And I don't get that mindset where the idea
link |
that if you can make the world better inside the headset
link |
than outside, you've just improved the person's life
link |
that has a headset that can wear it.
link |
And there are plenty of things that we just can't do
link |
for everyone in the real world.
link |
Everybody can't have Richard Branson's Private Island,
link |
but everyone can have a private VR Island
link |
and it can have the things that they want on it.
link |
And there's a lot of these kind of rivalrous goods
link |
in the real world that VR can just be better at.
link |
We can do a lot of things like that
link |
that can be very, very rich.
link |
So yeah, I want the,
link |
I think it's going to be a positive thing,
link |
this world where people want to go back into their headset,
link |
where it can be better than somebody that's living
link |
in a tiny apartment can have a palatial estate
link |
in virtual reality.
link |
They can have all their friends from all over the world
link |
come over and visit them without everybody getting
link |
on a plane and meeting in some place
link |
and dealing with all the other logistics hassles.
link |
There is real value in the presence that you can get
link |
for remote meetings.
link |
It's all the little things that we need to sort out,
link |
but those are things that we have line of sight on.
link |
People that have been in like a good VR meeting
link |
using work rooms where you can say,
link |
oh, that was better than a Zoom meeting.
link |
But of course it's more of a hassle to get into it.
link |
Not everyone has the headset.
link |
Interoperability is worse.
link |
You can't have, you cap out at a certain number.
link |
There's all these things that need to be fixed,
link |
but that's one of those things you can look at and say,
link |
we know there's value there.
link |
We just need to really grind hard,
link |
file off all the rough edges and make that possible.
link |
So you do think we have line of sight
link |
because there's a reason like,
link |
I do this podcast in person, for example.
link |
So doing it remotely, it's not the same.
link |
And if somebody were to ask me why it's not the same,
link |
I wouldn't be able to write down exactly why.
link |
But you're saying that it's possible,
link |
whatever the magic is for in person interaction,
link |
that immersiveness of the experience,
link |
we are almost there.
link |
Yes, so the idea of like,
link |
I'm doing a VR interview with someone.
link |
I'm not saying it's here right now,
link |
but you can see glimmers of what it should be.
link |
And we largely know what would need to be fixed
link |
and improved to, like you say,
link |
there's a difference between at remote interview
link |
doing a podcast over Zoom or something and face to face.
link |
There's that sense of presence, that immediacy,
link |
the super low latency responsiveness,
link |
being able to see all the subtle things there,
link |
just occupying the same field of view.
link |
And all of those are things
link |
that we absolutely can do in VR.
link |
And that simple case of a small meeting
link |
with a couple of people, that's the much easier case
link |
than everybody thinks the Ready Player One multiverse
link |
with a thousand people going across a huge bridge
link |
to amazing places,
link |
that's harder in a lot of other technical ways.
link |
Not to say we can't also do that,
link |
but that's further away and has more challenges.
link |
But this small thing about being able to have a meeting
link |
with one or a few people and have it feel real,
link |
feel like you're there,
link |
like you have the same interactions and talking with them,
link |
you get subtle cues as we start getting eye
link |
and face tracking and some of the other things
link |
on high end headsets.
link |
A lot of that is going to come over
link |
and it doesn't have to be as good.
link |
This is an important thing that people miss
link |
where there was a lot of people that,
link |
especially rich people that would look at VR and say,
link |
it's like, oh, this just isn't that good.
link |
And I'd say, it's like, well,
link |
you've already been courtside backstage and on pit row
link |
and you've done all of these experiences
link |
because you get to do them in real life.
link |
But most people don't get to.
link |
And even if the experience is only half as good,
link |
if it's something that they never would have gotten
link |
to do before, it's still a very good thing.
link |
And as we can just, we can push that number up over time.
link |
It has a minimum viable value level
link |
when it does something that is valuable enough to people,
link |
as long as it's better inside the headset on any metric
link |
than it is outside and people choose to go there,
link |
we're on the right path.
link |
And we have a value gradient
link |
that I'm just always hammering on.
link |
We can just follow this value gradient,
link |
just keep making things better
link |
rather than going for that one, close your eyes,
link |
swing for the fences, kind of silver bullet approach.
link |
Well, I wonder if there's a value gradient
link |
for in person meetings, because if you get that right,
link |
I mean, that would change the world.
link |
Yeah. That it doesn't need to,
link |
I mean, you don't need a ready player one.
link |
But I wonder if there's that value gradient
link |
you can follow along, because if there is and you follow it,
link |
then there'll be a certain like phase shift
link |
at a certain point where people will shift
link |
from Zoom to this.
link |
I wonder, what are the bottlenecks?
link |
Is it software, is it hardware?
link |
Is it like, is it all about latency?
link |
So I have big arguments internally
link |
about strategic things like that,
link |
where like the next headset that's coming out
link |
that we've made various announcements about
link |
is gonna be a higher end headset,
link |
more expensive, more features.
link |
Lots of people wanna make those trade offs.
link |
We'll see what the market has to say
link |
about the exact trade offs we've made here.
link |
But if you wanna replace Zoom,
link |
you need to have something that everybody has.
link |
So you like cheaper.
link |
I like cheaper because also lighter and cheaper
link |
wind up being a virtuous cycle there
link |
where expensive and more features
link |
tends to also lead towards heavier.
link |
And it just kind of goes, it's like, let's add more features.
link |
The features are not, they have physical presence
link |
and weight and draw from batteries and all of those things.
link |
So I've always favored a lower end, cheaper, faster approach.
link |
That's why I was always behind the mobile side of VR
link |
rather than the higher end PC headsets.
link |
And I think that's proven out well.
link |
But there's, ideally we have a whole range of things,
link |
but if you've only got one or two things,
link |
it's important that those two things cover
link |
the scope that you think is most important.
link |
When we're in a world when it's like cell phones
link |
and there's 50 of them on the market
link |
covering every conceivable ecological niche you want,
link |
that's gonna be great,
link |
but we're not gonna be there for a while.
link |
Where are the bottlenecks?
link |
Is it the hardware or the software?
link |
Yeah, so right now you can play,
link |
you can get work rooms on Quest
link |
and you can set up these things
link |
and it's a pretty good experience.
link |
It's surprisingly good.
link |
I haven't tried it.
link |
It's surprisingly good.
link |
Yeah, the voice latency is better on that
link |
than a lot better than a Zoom meeting.
link |
So you've got a more, a better sense of immediacy there.
link |
The expressions that you get from the current hardware
link |
with just kind of your controllers and your head
link |
is pretty realistic feeling.
link |
You've got a pretty good sense of being there
link |
with someone with it.
link |
Are these like avatars of people?
link |
Do you get to see their body?
link |
And they're sitting around a table?
link |
And it feels better than Zoom?
link |
It feels better than you, yeah,
link |
better than you'd expect for that.
link |
It is definitely, yeah, I'd say it's quite a bit better
link |
than Zoom when everything's working right,
link |
but there's still all the rough edges of,
link |
the reason Zoom became so successful
link |
is because they just nailed the usability of everything.
link |
It's high quality with a absolutely first rate experience.
link |
And we are not there yet with any of the VR stuff.
link |
I'm trying to push hard to get,
link |
I keep talking about it's like,
link |
it needs to just be one click to make everything happen.
link |
And we're getting there in our home environment,
link |
not the whole work rooms application,
link |
but the main home where you can now kind of go over
link |
and click and invite,
link |
and it still winds up taking five times longer
link |
but we're getting close to that where you click there,
link |
they click on their button and then they're sitting there
link |
in this good presence with you,
link |
but latencies need to get a lot better.
link |
User interface needs to get a lot better.
link |
Ubiquity of the headsets needs to get better.
link |
We need to have a hundred million of them out there
link |
just so that everybody knows somebody
link |
that uses this all the time.
link |
Well, I think it's a virtuous cycle
link |
because I do think the interface
link |
is the thing that makes or breaks this kind of revolution.
link |
It's so interesting how like you said one click,
link |
but it's also like how you achieve that one click.
link |
What is, can I ask a dark question?
link |
Maybe let's keep it outside of meta,
link |
but this is about meta, but also Google and big company.
link |
Are they able to do this kind of thing?
link |
It seems like, let me put on my cranky old man hat,
link |
is they seem to not do a good job
link |
of creating these user friendly interfaces
link |
as they get bigger and bigger as a company.
link |
Like Google has created some of the greatest interfaces ever
link |
early on in its, I mean, creating Gmail,
link |
just so many brilliant interfaces
link |
and it just seems to be getting crappier
link |
and crappier at that.
link |
Same with meta, same with Microsoft.
link |
It's just, it seems to get worse and worse at that.
link |
I don't know what it is,
link |
because you've become more conservative,
link |
careful, risk averse.
link |
Can you speak to that?
link |
It's been really eye opening to me
link |
working inside a tech titan,
link |
where I had my small companies
link |
and then we're acquired by a midsize game publisher
link |
and then Oculus getting acquired by meta
link |
and meta has grown by a factor of many
link |
just in the eight years since the acquisition.
link |
So I did not have experience with this
link |
and it was interesting because I remember like previously
link |
my benchmark for kind of use of resources
link |
was some of the government programs I interacted with
link |
on the aerospace side.
link |
And I remember thinking there was,
link |
okay, there's an air force program
link |
and they spent $50 million and they didn't launch anything.
link |
They didn't even build anything.
link |
It was just kind of like they made a bunch of papers
link |
and had some parts in a warehouse and nothing came of it.
link |
It's like $50 million and I've had to radically recalibrate
link |
my sense of like how much money can be spent
link |
with mediocre resources.
link |
Where on the plus side, VR has turned out,
link |
we've built pretty much exactly what,
link |
we just passed the 10 year mark then
link |
from like my first demo of the Rift.
link |
And if I could have said what I wanted to have,
link |
it would have been a standalone inside out tracked
link |
4K resolution headset that I,
link |
that could still plug into a PC for high end rendering.
link |
And that's exactly what we've got on Quest 2 right now.
link |
Yes, first of all, let's pause on that
link |
with me being cranky and everything.
link |
It's what Meta achieved with Oculus and so on is incredible.
link |
I mean, this is, when I thought about the future of VR,
link |
this is what I imagined in terms of hardware, I would say.
link |
And maybe in terms of the experience as well,
link |
but it's still not there somehow.
link |
On the one hand, we did kind of achieve it and win.
link |
And we've got, we've sold, we're a success right now,
link |
but the amount of resources that have gone into it,
link |
it winds up getting cluttered up in accounting
link |
where Mark did announce that they spent $10 billion a year,
link |
like on Reality Labs.
link |
Now Reality Labs covers a lot.
link |
It was, VR was not the large part of it.
link |
It also had Portal and Spark
link |
and the big AR research efforts.
link |
And it's been expanding out to include AI
link |
and other things there where there's a lot going on there.
link |
But $10 billion was just a number
link |
that I had trouble processing.
link |
It's just, I feel sick to my stomach
link |
thinking about that much money being spent.
link |
But that's how they demonstrate commitment to this,
link |
where it's not more so than like, yeah,
link |
Google goes and cancels all of these projects,
link |
different things like that,
link |
while Meta is really sticking with the funding of VR
link |
and AR is still further out with it.
link |
So there's something to be said for that.
link |
It's not just gonna vanish, the work's going in.
link |
I just wish it could be,
link |
all those resources could be applied more effectively
link |
because I see all these cases.
link |
I point out these examples of how a third party
link |
that we're kind of competing with in various ways.
link |
There's a number of these examples
link |
and they do work with a 10th of the people
link |
that we do internally and a lot of it comes from,
link |
yes, the small company can just go do it
link |
while in a big company, you do have to worry about,
link |
is there some SDK internally that you should be using
link |
because another team's making it,
link |
you have to have your cross functional group meetups
link |
for different things.
link |
You do have more concerns about privacy
link |
or diversity and equity and safety of different things,
link |
parental issues and things that a small startup company
link |
can just kind of cowboy off and do something interesting.
link |
And there's a lot more that is a problem
link |
that you have to pay attention to in the big companies,
link |
but I'm not willing to believe that we are within
link |
even a factor of two or four
link |
of what the efficiency could be.
link |
I am constantly kind of crying out for it's like,
link |
we can do better than this.
link |
And you wonder what the mechanisms
link |
to unlock that efficiency are.
link |
I don't, there is some sense in a large company
link |
that like an individual engineer might not believe
link |
that they can change the world.
link |
Maybe you delegate a little bit of the responsibility
link |
to be the one who changes the world in a big company,
link |
I think, but the reality is like the world will get changed
link |
by a single engineer anyway.
link |
So if whether inside Google or inside a startup,
link |
it doesn't matter, it's just like Google
link |
and Meta needs to help those engineers believe
link |
they're the ones that are gonna decrease that latency.
link |
It'll take one John Carmack, the 20 year old Carmack
link |
that's inside Meta right now to change everything.
link |
And I try to point that out and push people.
link |
It's like, try to go ahead.
link |
And when you see some, because there is,
link |
you get the silo mentality where you're like,
link |
okay, I know something's not right over there,
link |
but I'm staying in my lane here.
link |
And there's a couple of people that I can think about
link |
that are willing to just like hop all over the place
link |
and man, I treasure them.
link |
The people that are just willing to, they're fearless.
link |
They will go over and they will go rebuild the kernel
link |
and change this distribution and go in
link |
and hack a firmware over here to get something done right.
link |
And that is relatively rare.
link |
There's thousands of developers
link |
and you've got a small handful that are willing
link |
to operate at that level.
link |
And it's potentially risky for them.
link |
The politics are real in a lot of that.
link |
And I'm in the very much the privileged position
link |
of I'm more or less untouchable there
link |
where I've been dinged like twice
link |
for it's like you said something insensitive in that post
link |
and you should probably not say that.
link |
But for the most part, yes, I get away with,
link |
every week I'm posting something pretty loud
link |
and opinionated internally.
link |
And I think that's useful for the company,
link |
but yeah, it's rare to have a position like that.
link |
And I can't necessarily offer advice
link |
for how someone can do that.
link |
Well, you could offer advice to a company in general
link |
to give a little bit of freedom for the young wild,
link |
like the wildest ideas come from the young minds.
link |
And so you need to give the young minds freedom
link |
to think big and wild and crazy.
link |
And for that, they have to be opinionated.
link |
They have to think crazy ideas and thoughts
link |
and pursue them with a full passion
link |
without being slowed down by bureaucracy or managers
link |
and all that kind of stuff.
link |
Obviously startups really empower that,
link |
but big companies could too.
link |
And that's a design challenge for big companies
link |
to see how can you enable that?
link |
How can you empower that?
link |
Yeah, because the big company,
link |
there are so many resources there.
link |
And they do, amazing things do get accomplished,
link |
but there's so much more that could come out of that.
link |
And I'm always hopeful.
link |
I'm an optimist in almost everything.
link |
I think things can get better.
link |
I think that they can improve things
link |
that you go through a path
link |
and you're learning kind of what does and doesn't work.
link |
And I'm not ready to be fatalistic
link |
about the kind of the outcome of any of that.
link |
Me neither, I know too many good people
link |
inside of those large companies that are incredible.
link |
You have a friendship with Elon Musk.
link |
Often when I talk to him,
link |
he'll bring up how incredible of an engineer
link |
and just a big picture thinker you are.
link |
His huge amount of respect for you.
link |
I just, I've never been a fly in the wall
link |
between a discussion between the two of you.
link |
I just wonder, is there something you guys debate,
link |
argue about, discuss?
link |
Is there some interesting problems
link |
that the two of you think about?
link |
You come from different worlds.
link |
Maybe there's some intersection in aerospace.
link |
Maybe there's some intersection
link |
in your new efforts in artificial intelligence
link |
in terms of thinking.
link |
Is there something interesting you could say
link |
about sort of the debates the two of you have?
link |
So I think in some ways,
link |
we do have a kind of similar background
link |
where we're almost exactly the same age.
link |
And we had kind of similar programming backgrounds
link |
on the personal computers
link |
and even some of the books that we would read
link |
and things that would kind of turn us into the people
link |
that we are today.
link |
And I think there is a degree of sensibility similarities
link |
where we kind of call bullshit on the same things
link |
and kind of see the same opportunities
link |
in different technology.
link |
And there's that sense of,
link |
I always talk about the speed of light solutions for things.
link |
And he's thinking about kind of minimum manufacturing
link |
and engineering and operational standpoints for things.
link |
And so, I mean, I first met Elon
link |
right at the start of the aerospace era
link |
where I wasn't familiar with,
link |
I was still in my game dev bubble.
link |
I really wasn't familiar with all the startups
link |
that were going and being successful
link |
and what went on with PayPal
link |
and all of his different companies.
link |
But I met him as I was starting to do armadillo aerospace.
link |
And he came down with kind of his right hand propulsion guy.
link |
And we talked about rockets.
link |
What can we do with this?
link |
And it was kind of specific things
link |
about like how are our flight computers set up?
link |
What are different propellant options?
link |
You know, what can happen with different ways
link |
of putting things together?
link |
And then in some ways,
link |
he was certainly the biggest player
link |
in the sort of alt space community
link |
that was going on in the early 2000s.
link |
He was the most well funded,
link |
although his funding in the larger scheme of things
link |
compared to like a NASA or something like that
link |
It was a lot more than I had at the time.
link |
But it was interesting.
link |
I had a point years later when I realized,
link |
okay, my financial resources at this point
link |
are basically what Elon's was
link |
when he went all in on SpaceX and Tesla.
link |
And there's, I think in many corners,
link |
he does not get the respect that he should
link |
about being a wealthy person that could just retire.
link |
And he went all in where he was really going to,
link |
he could have gone bust.
link |
And there's plenty of people,
link |
you'd look at the sad athletes or entertainers
link |
that had all the money in the world and blew it.
link |
And he could have been the business case example of that.
link |
But the things that he was doing,
link |
space exploration, electrification of transportation,
link |
solar city type things,
link |
these are big world level things.
link |
And I have a great deal of admiration
link |
that he was willing to throw himself so completely
link |
into that because in contrast with myself,
link |
I was doing armadillo aerospace with this tightly bounded,
link |
it was John's crazy money at the time
link |
that had a finite limit on it.
link |
It was never going to impact me or my family
link |
if it completely failed.
link |
And I was still hedging my bets,
link |
working at id Software at the time
link |
when he had been really all in there.
link |
And I have a huge amount of respect for that.
link |
And people do not,
link |
the other thing I get irritated with is people that say,
link |
it's like, oh, Elon's just a business guy.
link |
He just got like, he was gifted the money
link |
and he's just kind of investing in all of this
link |
when he was really deeply involved
link |
in a lot of the decisions.
link |
Not all of them were perfect,
link |
but he cared very much about engine material selection,
link |
propellant selection.
link |
And for years he'd be kind of telling me,
link |
it's like, get off that hydrogen peroxide stuff.
link |
It's like, liquid oxygen is the only proper oxidizer
link |
And the times that I've gone through the factories
link |
with him, we're talking very detailed things
link |
about like how this weld is made,
link |
how this sub assembly goes together,
link |
what are like startup shut down behaviors
link |
of the different things.
link |
So he is really in there at a very detailed level.
link |
And I think that he is the best modern example now
link |
of someone that tries to,
link |
that can effectively micromanage some decisions on things
link |
on both Tesla and SpaceX to some degree
link |
where he cares enough about it.
link |
I worry a lot that he stretched too thin
link |
that you get Boring Company and Neuralink and Twitter
link |
and all the other possible things there
link |
where I know I've got limits on how much I can pay attention
link |
to that I have to kind of box off different amounts of time.
link |
And I look back at like at my aerospace side of things,
link |
it's like, I did not go all in on that.
link |
I did not commit myself at a level
link |
that it would have taken to be successful there.
link |
And I, yeah, and it's kind of a weird thing
link |
just like having a discussion with him.
link |
He's the richest man in the world right now,
link |
but he operates on a level that is still very much
link |
in my wheelhouse on a technical side of things.
link |
That doing that systems level type of thinking
link |
where you can go to the low level details
link |
and go up high to the big picture.
link |
Do you think in aerospace arena
link |
in the next five, 10 years,
link |
do you think we're gonna put a human on Mars?
link |
Like, what do you think is the interesting?
link |
No, I do, in fact, I made a bet with someone
link |
with a group of people kind of this
link |
about whether boots on Mars by 2030.
link |
And this was kind of a fun story
link |
because I was at an Intel sponsored event
link |
and we had a bunch of just world class, brilliant people.
link |
And we were talking about computing stuff,
link |
but the after dinner conversation was like,
link |
what are some other things?
link |
How are they gonna go in the future?
link |
And one of the ones tossed up on the whiteboard
link |
was like boots on Mars by 2030.
link |
And most of the people in the room thought, yes,
link |
I thought that like, SpaceX is kicking ass.
link |
We've got all this possible stuff.
link |
Seems likely that it's gonna go that way.
link |
And I said, no, I think less than 50% chance
link |
that it's going to make it there.
link |
And people were kind of like,
link |
oh, why the pessimism or whatever?
link |
And of course I'm an optimist at almost everything,
link |
but for me to be the one kind of outlier saying,
link |
no, I don't think so.
link |
Then I started saying some of the things I said,
link |
well, let's be concrete about it.
link |
Let's bet $10,000 that it's not gonna happen.
link |
And this was really a startling thing to see that I,
link |
again, room full of brilliant people,
link |
but as soon as like money came on the line
link |
and they were like, do I wanna put 10,000?
link |
I was not the richest person in the room.
link |
There were people much better off than I was.
link |
There's a spectrum, but as soon as they started thinking,
link |
it's like, oh, I could lose money
link |
by keeping my position right now.
link |
And all these engineers, they engaged their brain.
link |
They started thinking, it's like, okay,
link |
launch windows, launch delays,
link |
like how many times would it take to get this right?
link |
What historical precedents do we have?
link |
And then it mostly came down to, it's like,
link |
well, what about in transit by 2030?
link |
And then what about different things
link |
or would you go for 2032?
link |
But one of the people did go ahead
link |
and was optimistic enough to make a bet with me.
link |
So I have a $10,000 bet that by 2030,
link |
I think it's gonna happen shortly thereafter.
link |
I think there will probably be infrastructure on Mars
link |
by 2030, but I don't think that we'll have humans
link |
I think it's possible, but I think it's less
link |
than a 50% chance, so I felt safe making that bet.
link |
Well, I think you had an interesting point.
link |
Correct me if I'm wrong.
link |
That's a dark one.
link |
That should perhaps help people appreciate Elon Musk,
link |
which is, in this particular effort,
link |
Elon is critical to the success.
link |
SpaceX seems to be critical to, you know,
link |
humans on Mars by 2030 or thereabouts.
link |
So if something happens to Elon,
link |
then all of this collapses.
link |
And this is in contrast to the other $10,000 bet
link |
I made kind of recently, and that was self driving cars
link |
at like a level five running around cities.
link |
And people have kind of nitpicked that
link |
that we probably don't mean exactly level five,
link |
but the guy I'm having the bet with is,
link |
we're gonna be, we know what we mean about this.
link |
Yeah, coding horror and stack overflow and all.
link |
But yeah, I mean, he doesn't think that people
link |
are gonna be riding around in robo taxis in 2030
link |
in major cities, just like you take an Uber now.
link |
And I think it will.
link |
You think it will.
link |
And the difference is, everybody looks at this,
link |
it's like, oh, but Tesla's been wrong for years.
link |
They've been promising it for years and it's not here yet.
link |
And the reason this is different than the bet with Mars
link |
is Mars really is more than is comfortable
link |
a bet on Elon Musk.
link |
That is his thing and he is really going to move
link |
heaven and earth to try to make that happen.
link |
And perhaps not even SpaceX, perhaps just Elon Musk.
link |
Yeah, because if Elon went away and SpaceX went public
link |
and got a board of directors,
link |
there are more profitable things they could be doing
link |
than focusing on human presence on Mars.
link |
So this really is a sort of personal thing there.
link |
And in contrast with that, self driving cars
link |
have a dozen credible companies working really hard.
link |
And while, yes, it's going slower
link |
than most people thought it would,
link |
betting against that is a bet against almost the entire world
link |
in terms of all of these companies
link |
that have all of these incentives.
link |
It's not just one guy's passion project.
link |
And I do think that it is solvable.
link |
Although I recognize it's not 100% chance
link |
because it's possible the long tail of self driving problems
link |
winds up being an AGI complete problem.
link |
I think there's plenty of value to mine out of it with narrow AI.
link |
And I think that it's going to happen probably more so
link |
than people expect.
link |
But it's that whole sigmoid curve
link |
where you overestimate the near term progress
link |
and you underestimate the long term progress.
link |
And I think self driving is going to be like that.
link |
And I think 2030 is still a pretty good bet.
link |
Yeah, unfortunately, self driving is a problem
link |
that is safety critical, meaning that if you don't do it well,
link |
But the other side of that is people are terrible drivers.
link |
So it is not going to be,
link |
that's probably going to be the argument that gets it through
link |
is like we can save 10,000 lives a year
link |
by taking imperfect self driving cars
link |
and letting them take over a lot of driving responsibilities.
link |
It's like 30,000 people a year die in auto accidents right now in America.
link |
And a lot of those are preventable.
link |
And the problem is you'll have people that
link |
every time a Tesla crashes into something,
link |
you've got a bunch of people that literally have vested interests
link |
shorting Tesla to come out and make it the worst thing in the world.
link |
And people will be fighting against that.
link |
But optimist in me again,
link |
I think that we will have systems that are statistically safer than human drivers
link |
and we will be saving thousands and thousands of lives every year
link |
when we can hand over more of those responsibilities to it.
link |
I do still think as a person who studied this problem very deeply
link |
from a human side as well,
link |
it's still an open problem how good slash bad humans are driving.
link |
It's a kind of funny thing we say about each other.
link |
Oh, humans suck at driving.
link |
Everybody except you, of course.
link |
Like we think we're good at driving.
link |
But I, after really studying it,
link |
I think you start to notice, you know,
link |
because I watched hundreds of hours of humans driving
link |
with the projects of this kind of thing.
link |
You've noticed that even with the distraction,
link |
even with everything else,
link |
humans are able to do some incredible things
link |
with the attention,
link |
even when you're just looking at the smartphone,
link |
just to get cues from the environment
link |
to make last second decisions,
link |
to use instinctual type of decisions that actually save your ass
link |
time and time and time again.
link |
And are able to do that with so much uncertainty around you
link |
in such tricky dynamic environments.
link |
I don't know exactly how hard is it to beat that kind of skill
link |
of common sense reasoning.
link |
This is one of those interesting things that there have been
link |
a lot of studies about how experts in their field
link |
usually underestimate the progress that's going to happen
link |
because an expert thinks about all the problems they deal with
link |
and they're like, damn, I'm going to have a hard time
link |
solving all of this.
link |
And they filter out the fact that they are one expert
link |
in a field of thousands.
link |
And you think about, yeah, I can't do all of that.
link |
And you sometimes forget about the scope of the ecosystem
link |
that you're embedded in.
link |
And if you think back eight years, very specifically,
link |
the state of AI and machine learning, where was that?
link |
We had just gotten Resnets probably at that point.
link |
And you look at all the amazing magical things
link |
that have happened in eight years,
link |
and they do kind of seem to be happening a little faster
link |
in recent years also.
link |
And you project that eight more years into the future,
link |
where, again, I think there's a 50% chance
link |
we're going to have signs of life of AGI,
link |
which we can put through driver's ed if we need to
link |
to actually build self driving cars.
link |
And I think that the narrow systems are going to have
link |
real value demonstrated well before then.
link |
So signs of life in AGI, you've mentioned that,
link |
okay, first of all, you're one of the most brilliant people
link |
You could be solving a number of different problems,
link |
as you've mentioned.
link |
Your mind was attracted to nuclear energy.
link |
Obviously, virtual reality with the metaverse
link |
is something you could have a tremendous impact on.
link |
So I do want to say a quick thing about nuclear energy,
link |
where this is something that so precisely feels like
link |
aerospace before SpaceX, where from everything that I know
link |
about all of these, the physics of this stuff hasn't changed.
link |
And the reasons why things are expensive now
link |
are not fundamental.
link |
Somebody should be going into a really hard Elon Musk style
link |
at fission, economical fission, not fusion,
link |
where the fusion is the kind of the darling of people
link |
that want to go and do nuclear because it doesn't have
link |
the taint that fission has in a lot of people's minds.
link |
But it's an almost absurdly complex thing where nuclear fusion,
link |
as you look at the tokamaks or any of the things that people
link |
are building, and it's doing all of this infrastructure
link |
just at the end of the day to make something hot
link |
that you can then turn into energy
link |
through a conventional power plant.
link |
And all of that work, which we think
link |
we've got line of sight on, but even if it comes out,
link |
then you have to do all of that immensely complex expensive
link |
stuff just to make something hot, where nuclear fission
link |
is basically you put these two rocks together
link |
and they get hot all by themselves.
link |
That is just that much simpler.
link |
It's just orders of magnitude simpler.
link |
And the actual rocks, the refined uranium,
link |
is not very expensive.
link |
It's a couple percent of the cost of electricity.
link |
That's why I made that point where you could have something
link |
which was five times less efficient than current systems.
link |
And if the rest of the plant was a whole bunch cheaper,
link |
you could still be super, super valuable.
link |
So how much of the pie do you think could be solved
link |
by nuclear energy by fission?
link |
So how much could it become the primary source
link |
of energy on Earth?
link |
It could be most of it.
link |
Like the reserves of uranium, as it stands now,
link |
could not power the whole Earth.
link |
But you get into breeder reactors and thorium
link |
and things like that that you do for conventional fission.
link |
There is enough for everything.
link |
Now, I mean, solar, photovoltaic has been amazing.
link |
One of my current projects is working on an off grid system.
link |
And it's been fun just kind of, again, putting my hands
link |
on all this, stripping the wires and wiring things together
link |
and doing all of that.
link |
And just having followed that a little bit
link |
from the outside over the last couple decades,
link |
there's been semiconductor like magical progress
link |
in what's going on there.
link |
So I'm all for all of that.
link |
But it doesn't solve everything.
link |
And nuclear really still does seem like the smart money
link |
bet for what you should be getting for baseband
link |
on a lot of things.
link |
And solar may be cheaper for peaking over air conditioning
link |
loads during the summer and things
link |
that you can push around in different ways.
link |
But it's one of those things that's
link |
it's just strange how we've had the technology sitting there.
link |
But these non technical reasons on the social optics of it
link |
has been this major forcing function for something
link |
that really should be at the cornerstone of all
link |
of the world's concerns with energy.
link |
It's interesting how the non technical factors have really
link |
dominated something that is so fundamental to the existence
link |
of the human race as we know it today.
link |
And much of the troubles of the world,
link |
including wars in different parts of the world like Ukraine
link |
And yeah, it's just sitting right there to be solved.
link |
That said, I mean, to me personally,
link |
I think it's clear that if AGI were to be achieved,
link |
that would change the course of human history.
link |
So AGI wise, I was making this decision about what
link |
do I want to focus on after VR.
link |
And I'm still working on VR regularly.
link |
I spend a day a week kind of consulting with Meta.
link |
And Boz styles me the consulting CTO.
link |
It's kind of like the Sherlock Holmes that
link |
comes in and consults on some of the specific tough issues.
link |
And I'm still pretty passionate about all of that.
link |
But I have been figuring out how to compartmentalize and force
link |
that into a smaller box to work on some other things.
link |
And I did come down to this decision
link |
between working on economical nuclear fission
link |
or artificial general intelligence.
link |
And the fission side of things, I've
link |
got a bunch of interesting things going that way.
link |
But that would be a fairly big project thing to do.
link |
I don't think it needs to be as big as people expect.
link |
I do think something original SpaceX sized, you build it,
link |
power your building off of it, and then the government,
link |
I think, will come around to what you need to.
link |
Everybody loves an existence proof.
link |
I think it's possible somebody should be doing this.
link |
But it's going to involve some politics.
link |
It's going to involve decent sized teams
link |
and a bunch of this cross functional stuff
link |
that I don't love.
link |
While the artificial general intelligence side of things,
link |
it seems to me like this is the highest leverage
link |
moment for potentially a single individual,
link |
potentially in the history of the world,
link |
where the things that we know about the brain, about what we
link |
can do with artificial intelligence,
link |
nobody can say absolutely on any of these things.
link |
But I am not a madman for saying that it is likely
link |
that the code for artificial general intelligence
link |
is going to be tens of thousands of lines of code,
link |
not millions of lines of code.
link |
This is code that conceivably one individual could write,
link |
unlike writing a new web browser or operating system.
link |
And based on the progress that AI, machine learning,
link |
has made in the recent decade, it's
link |
likely that the important things that we don't know
link |
are relatively simple.
link |
There's probably a handful of things.
link |
My bet is that I think there's less than six key insights that
link |
Each one of them can probably be written
link |
on the back of an envelope.
link |
We don't know what they are.
link |
But when they're put together in concert with GPUs at scale
link |
and the data that we all have access to,
link |
that we can make something that behaves like a human being
link |
or like a living creature and that can then
link |
be educated in whatever ways that we need to get to the point
link |
where we can have universal remote workers where anything
link |
that somebody does mediated by a computer
link |
and doesn't require physical interaction that an AGI will
link |
We can already simulate the equivalent of the Zoom meetings
link |
with avatars and synthetic deep fakes and whatnot.
link |
We can definitely do that.
link |
We have superhuman capabilities on any narrow thing
link |
that we can formalize and make a loss function for.
link |
But there's things we don't know how to do now.
link |
But I don't think they are unapproachably hard.
link |
Now, that's incredibly hubristic to say that it's like,
link |
but I think that what I said a couple years ago
link |
is a 50% chance that somewhere there will be signs of life
link |
And I've probably increased that slightly.
link |
I may be at 55%, 60% now because I
link |
do think there's a little sense of acceleration there.
link |
So I wonder what the, and by the way,
link |
you also written that I bet with hindsight
link |
we will find that clear antecedents of all
link |
the critical remaining steps for AGI
link |
are already buried somewhere in the vast literature of today.
link |
So the ideas are already there.
link |
I think that's likely the case.
link |
One of the things that appeals to so many people,
link |
including me, about the promise of AGI
link |
is we know that we're only drinking
link |
from a straw from the fire hose of all the information out there.
link |
I mean, you look at just in a very narrowly bounded field
link |
like machine learning, like you can't read all the papers that
link |
come out all the time.
link |
You can't go back and read all the clever things
link |
that people did in the 90s or earlier
link |
that people have forgotten about because they didn't pan out
link |
at the time when they were trying to do them with 12 neurons.
link |
So this idea that, yeah, I think there
link |
are gems buried in some of the older literature that was not
link |
the path taken by everything.
link |
And you can see a herd mentality on the things that
link |
It's almost funny to see.
link |
It's like, oh, Google does something,
link |
and OpenAI does something, Meta does something.
link |
They're the same people that all talk to each other,
link |
and they're all one upping each other,
link |
and they're all capable of implementing
link |
each other's work given a month or two
link |
after somebody has an announcement of that.
link |
But there's a whole world of possible approaches
link |
to machine learning.
link |
And I think that we probably will, in hindsight,
link |
go back and see it's like, yeah, that was kind of clearly
link |
predicted by this early paper here.
link |
And this turns out that if you do this and this
link |
and take this result from animal training
link |
and this thing from neuroscience over here
link |
and put it together and set up this curriculum for them
link |
to learn in, that that's kind of what it took.
link |
You don't have too many people now
link |
that are still saying it's not possible
link |
or it's going to take hundreds of years.
link |
And 10 years ago, you would get a collection of experts,
link |
and you would have a decent chunk on the margin that either
link |
say not possible or a couple hundred years,
link |
might be centuries.
link |
And the median estimate would be like 50, 70 years.
link |
And it's been coming down.
link |
And I know with me saying eight years for something,
link |
that still puts me on the optimistic side,
link |
but it's not crazy out in the fringes.
link |
And just being able to look at that at a meta level
link |
about the trend of the predictions going down there,
link |
the idea that something could be happening relatively soon.
link |
Now, I do not believe in fast takeoffs.
link |
That's one of the safety issues that people say,
link |
it's like, oh, it's going to go, boom,
link |
and the AI is going to take over the world.
link |
There's a lot of reasons I don't think
link |
that's a credible position.
link |
And I think that we will go from a point
link |
where we start seeing things that credibly look
link |
like animals behaviors.
link |
And I have a human voice box wired into them.
link |
It's like I tried to get Elon to say,
link |
it's like your pig in Neuralink, give it a human voice box
link |
and let it start learning human words.
link |
I think animal intelligence is closer to human intelligence
link |
than a lot of people like to think.
link |
And I think that culture and modalities of IO
link |
make the Gulf seem a lot bigger than it actually is.
link |
There's just that smooth spectrum
link |
of how the brain developed and cortexes and scaling
link |
of different things going on there.
link |
Culture modalities of IO, yes.
link |
Language is sort of lost in translation,
link |
conceals a lot of intelligence.
link |
So when you think about signs of life for AGI,
link |
you're thinking about human interpretable signs.
link |
So the example I give, if we get to the point
link |
where you've got a learning disabled toddler,
link |
some kind of real special needs child
link |
that can still interact with their favorite TV show
link |
and video game and can be trained and learn
link |
in some appreciably human like way.
link |
At that point, you can deploy an army of engineers,
link |
cognitive scientists, developmental education people.
link |
And you've got so many advantages there
link |
unlike real education where you can do rollbacks
link |
and AB testing and you can find a golden path
link |
through a curriculum of different things.
link |
If you get to that point, learning disabled toddler,
link |
I think that it's gonna be a done deal.
link |
But do you think we'll know when we see it?
link |
So there's been a lot of really interesting
link |
general learning progress from DeepMind,
link |
OpenAI a little bit too.
link |
I tend to believe that Tesla Autopilot
link |
deserves a lot more credit than it's getting
link |
for making progress on the general,
link |
on doing the multitask learning thing
link |
and increasing the number of tasks
link |
and automating that process of sort of learning
link |
from the edge, discovering the edge cases
link |
and learning from the edge cases.
link |
That is, it's really approaching from a different angle,
link |
the general learning problem of AGI.
link |
But the more clear approach comes from DeepMind
link |
where you have these kind of game situations
link |
and you build systems there.
link |
But I don't know, people seem to be quite...
link |
Yeah, there will always be people that just won't believe it
link |
and I fundamentally don't care.
link |
I mean, I don't care if they don't believe it.
link |
You know, when it starts doing people's jobs
link |
and I don't care about the philosophical zombie argument at all.
link |
But will you, do you think you will notice
link |
that something special has happened here?
link |
Because to me, I've been noticing a lot of special things.
link |
I think a lot of credit should go to DeepMind for AlphaZero.
link |
That was truly special.
link |
The self play mechanisms achieve sort of solve problems
link |
that used to be thought unsolvable like the game of Go.
link |
Also, I mean, protein folding, starting to get into that space
link |
where learning is doing...
link |
At first there's not, it wasn't end to end learning
link |
and now it's end to end learning of a very difficult
link |
previously thought unsolvable problem of protein folding.
link |
And so, yeah, where do you think
link |
would be a really magical moment for you?
link |
There have been incredible things happening in recent years.
link |
Like you say, all of the things from DeepMind and OpenAI
link |
that have been huge showpiece things.
link |
But when you really get down to it and you read the papers
link |
and you look at the way the models are going,
link |
it's still like a feed forward.
link |
You push something in, something comes out on the end.
link |
I mean, maybe there's diffusion models
link |
or Monte Carlo tree rollouts and different things going on,
link |
but it's not a being.
link |
It's not close to a being.
link |
That's going through a lifelong learning process.
link |
Do you want something that kind of gives signs of a being?
link |
What's the difference between a neural network,
link |
a feed forward neural network and a being?
link |
Fundamentally, the brain is a recurrent neural network
link |
generating an action policy.
link |
I mean, it's implemented on a biological substrate.
link |
And it's interesting thinking about things like that
link |
where we know fundamentally the brain is not
link |
a convolutional neural network or a transformer.
link |
Those are specialized things that are very valuable
link |
for what we're doing, but it's not the way the brain's doing.
link |
Now, I do think consciousness and AI in general
link |
is a substrate independent mechanism
link |
where it doesn't have to be implemented the way the brain is,
link |
but if you've only got one existence proof,
link |
there's certainly some value in caring about what it says and does.
link |
And so the idea that anything that can be done with a narrow AI
link |
that you can quantify up a loss function for or reward mechanism,
link |
you're almost certainly going to be able to produce something
link |
that's more resource effective to train and deploy
link |
and use in an inference mode.
link |
You know, train a whole lot using an inference.
link |
But a living being is going to be something that's a continuous,
link |
lifelong learned task agnostic thing.
link |
So the lifelong learning is really important too.
link |
And the long term memory.
link |
So memory is a big weird part of that puzzle.
link |
And we've got, you know, again, I have all the respect in the world
link |
for the amazing things that are being done now,
link |
but sometimes they can be taken a little bit out of context
link |
with things like there's some smoke and mirrors going on,
link |
like the Gato, the recent work, the multitask learning stuff.
link |
You know, it's amazing that it's one model that plays all the Atari games
link |
as well as doing all of these other things.
link |
But of course, it didn't learn to do all of those.
link |
It was instructed in doing that by other reinforcement learners
link |
going through and doing that.
link |
And even in the case of all the games,
link |
it's still going with a specific hand coded reward function
link |
in each of those Atari games where it's not that, you know,
link |
it just wants to spend its summer afternoon playing Atari
link |
because that's the most interesting thing for it.
link |
So it's, again, not a general, it's not learning the way humans learn.
link |
And there's, I believe, a lot of things that are challenging
link |
to make a loss function for that you can train
link |
through these existing conventional things.
link |
We're going to chip away at all the things that people do
link |
that we can turn into narrow AI problems
link |
and billions of, like trillions of dollars of value
link |
are going to be created by that.
link |
But there's still going to be a set of things
link |
and we've got questionable cases like the self driving car
link |
where it's possible, it's not my bet, but it's plausible
link |
that the long tail could be problematic enough
link |
that that really does require a full on artificial general intelligence.
link |
The counter argument is that data solves almost everything.
link |
There's an interpolation problem if you have enough data
link |
and Tesla may be able to get enough data
link |
from all of their deployed stuff to be able to work like that,
link |
And there are all the other problems about, like,
link |
say you want to have a strategy meeting
link |
and you want to go ahead and bring in all of your remote workers
link |
and your consultants, and you want a world where
link |
some of those could be AIs that are talking and interacting with you
link |
in an area that is too murky to have a crisp loss function,
link |
but they still have things that on some level,
link |
they're rewarded on some internal level
link |
for building a valuable to humans kind of life
link |
and ability to interact with things.
link |
See, I still think that self driving cars,
link |
solving that problem will take us very far towards AGI.
link |
You might not need AGI, but I am really inspired
link |
by what Autopilot is doing.
link |
Waymo, so some of the other companies,
link |
I think Waymo leads the way there is also really interesting,
link |
but they don't have quite as ambitious of an effort
link |
in terms of learning based sort of data hungry approach to driving,
link |
which I think is very close to the kind of thing
link |
that would take us far towards AGI.
link |
Yeah, and it's a funny thing because as far as I can tell,
link |
Elon is completely serious about all of his concerns about AGI
link |
being an existential threat.
link |
And I tried to draw him out to talk about AGI
link |
and he just didn't want to.
link |
And I think that I get that little fatalistic sense from him.
link |
It's weird because his company could very well be the leading company
link |
leading towards a lot of that where Tesla being a super pragmatic company
link |
that's doing things because they really want to solve this actual problem.
link |
It's a different vibe than the research oriented companies
link |
where it's a great time to be an AI researcher.
link |
You've got your pick of trillion dollar companies
link |
that will pay you to kind of work on the problems you're interested in,
link |
but that's not necessarily driving hard towards the core problem
link |
of AGI as something that's going to produce a lot of value
link |
by doing things that people currently do or would like to do.
link |
I mean, I have a million questions to you about your ideas about AGI,
link |
but do you think it needs to be embodied?
link |
Do you think it needs to have a body to start to notice the signs of life
link |
and to develop the kind of system that's able to reason,
link |
perceive the world in the way that an AGI should and act in the world?
link |
So should we be thinking about robots
link |
or can this be achieved in a purely digital system?
link |
So I have a clear opinion on that and that's that no,
link |
it does not need to be embodied in the physical world
link |
where you could say most of my career is about making simulated virtual worlds
link |
in games or virtual reality.
link |
And so on a fundamental level,
link |
I believe that you can make a simulated environment
link |
that provides much of the value of what the real environment does
link |
and restricting yourself to operating at real time
link |
in the physical world with physical objects,
link |
I think is an enormous handicap.
link |
I mean, that's one of the real lessons driven home by all my aerospace work
link |
is that, you know, reality is a bitch in so many ways there
link |
where dealing with all the mechanical components,
link |
like everything fails, Murphy's law,
link |
even if you've done it right before on your fifth one,
link |
it might come out differently.
link |
So yeah, I think that anybody that is all in on the embodied aspect of it,
link |
they are tying a huge weight to their ankles
link |
and I think that I would almost count them out,
link |
anybody that's making that a cornerstone of their belief about it,
link |
I would almost write them off as being worried about them getting to AGI first.
link |
I was very surprised that Elon's big on the humanoid robots.
link |
I mean, like the NASA Robonaut stuff was almost a gag line,
link |
like what are you doing people?
link |
Well, that's very interesting because he has a very pragmatic view of that.
link |
That's just a way to solve a particular problem in a factory.
link |
Now, I do think that once you have an AGI,
link |
robotic bodies, humanoid bodies are going to be enormously valuable.
link |
I just don't think they're helpful getting to AGI.
link |
Well, he has a very sort of practical view, which I disagree with and argue with him,
link |
but it's a practical view that there's, you know,
link |
you could transfer the problem of driving to the problem of robotic manipulation
link |
because so much of it is perception.
link |
It's perception and action and it's just a different context
link |
and so you can apply all the same kind of data engine learning processes
link |
to a different environment.
link |
And so why not apply it to the humanoid robot environment?
link |
But I think, I do think that there's a certain magic to the embodied robot.
link |
That may be the thing that finally convinces people.
link |
But again, I don't really care that much about convincing people.
link |
You know, the world that I'm looking towards is, you know,
link |
you go to the website and say,
link |
I want five Frank 1As to work on my team today and they all spin up
link |
and they start showing up in your Zoom meetings.
link |
To push back, but also to agree with you.
link |
But first to push back, I do think you need to convince people
link |
for them to welcome that thing into their life.
link |
I think there's enough businesses that operate on an objective kind of
link |
profit loss sort of basis that, I mean, if you look at how many things,
link |
again, talking about the world as an evolutionary space there,
link |
when you do have free markets and you have entrepreneurs,
link |
you are going to have people that are going to be willing to go out
link |
and try whatever crazy things.
link |
And when it proves to be beneficial, you know,
link |
there's fast followers in all sorts of places.
link |
Yeah, and you're saying that, I mean, you know, Quake and VR
link |
is a kind of embodiment, but just in a digital world.
link |
And if you're able to demonstrate, if you're able to do something
link |
productive in that kind of digital reality,
link |
then AGI doesn't need to have a body.
link |
Yeah, it's like one of the really practical technical questions
link |
that I kind of keep arguing with myself over.
link |
If you're doing a training and learning and you've got like,
link |
you can watch Sesame Street, you can play Master System games or something,
link |
is it enough to have just a video feed that is that video coming in?
link |
Or should it literally be on a virtual TV set in a virtual room,
link |
even if it's, you know, a simple room, just to have that sense of
link |
you're looking at a 2D projection on a screen versus having the screen
link |
beam directly into your retinas.
link |
And, you know, I think it's possible to maybe get past some of these
link |
signs of life of things with the just kind of projected directly
link |
into the receptor fields, but eventually for more kind of human
link |
emotional connection for things, probably having some VR room
link |
with a lot of screens in it for the AI to be learning in is likely helpful.
link |
It may be a world of different AIs interacting with each other.
link |
That self play I do think is one of the critical things where
link |
socialization wise, one of the other limitations I set for myself
link |
thinking about these is I need something that is at least potentially
link |
real time because I want, it's nice you can always slow down time,
link |
you can run on a subscale system and test an algorithm at some lower level.
link |
And if you've got extra horsepower, running it faster than real time
link |
But I want to be able to have the AIs either socially interact with
link |
each other or critically with actual people, your sort of child
link |
development psychiatrist that comes in and interacts and does the
link |
good boy, bad boy sort of thing as they're going through and
link |
exploring different things.
link |
And it's nice to, I come back to the value of constraints in a lot of ways.
link |
And if I say, well, one of my constraints is real time operation.
link |
I mean, it might still be a huge data center full of computers,
link |
but it should be able to interact on a Zoom meeting with people.
link |
And that's how you also do start convincing people, even if it's not
link |
a robot body moving around, which eventually gets to irrefutable levels.
link |
But if you can go ahead and not just type back and forth to a GPT bot
link |
on something, but you're literally talking to them in an embodied
link |
over Zoom form and working through problems with them or exploring
link |
situations, having conversations that are fully stateful and learned.
link |
I think that that's a valuable thing.
link |
So I do keep all of my eyes on things that can be implemented within
link |
sort of that 30 frames per second kind of work.
link |
And I think that's feasible.
link |
Do you think the most compelling experiences that first will be
link |
for pleasure or for business as they ask in airports?
link |
So meaning is if it's interacting with AI agents, will it be sort of
link |
like friends, entertainment, almost like a therapist or whatever,
link |
that kind of interaction?
link |
Or is it in the business setting, something like you said, brainstorming
link |
different ideas, sort of this is all a different formulation of kind of
link |
a Turing test or the spirit of the original Turing test.
link |
Where do you think the biggest benefit will first come?
link |
So it's going to start off hugely expensive.
link |
I mean, you're going to if we're still all guessing about what compute
link |
is going to be necessary.
link |
I fall on the side of I don't think you run the numbers and you're like
link |
86 billion neurons, 100 trillion synapses.
link |
I don't think those all need to be weights.
link |
I don't think we need models that are quite that big evaluated quite
link |
I based that on we've got reasonable estimates of what some parts of
link |
We don't have the neocortex formula, but we kind of get some of the
link |
other sensory processing and it doesn't feel like we need to.
link |
We can simulate that in computers for less weights.
link |
But still, it's probably going to be thousands of GPUs to be running
link |
a human level AGI, depending on how it's implemented.
link |
That might give you sort of a clan of 128 kind of run in batch people,
link |
depending on whether there's sparsity in the way that the weights are
link |
and things are set up.
link |
If it is a reasonably dense thing, then just the memory bandwidth
link |
tradeoffs means you get 128 of them at the same time.
link |
And either it's all feeding together, learning in parallel or kind of
link |
all running together, kind of talking to a bunch of people.
link |
But still, if you've got thousands of GPUs necessary to run these
link |
things, it's going to be kind of expensive where it might start off
link |
$1,000 an hour for even post development or something for that, which
link |
would be something that you would only use for a business, something
link |
where you think they're going to help you make a strategic decision
link |
or point out something super important.
link |
But I also am completely confident that we will have another factor
link |
of 1,000 in cost performance increase in AGI type calculations.
link |
Not in general computing necessarily, but there's so much more that
link |
we can do with packaging, making those right tradeoffs, all those
link |
same types of things that in the next couple of decades, 1,000 X easy.
link |
And then you're down to $1 an hour.
link |
And then you're kind of like, well, I should have an entourage of
link |
AIs that are following me around, helping me out on anything that
link |
I want them to do.
link |
That's one interesting trajectory.
link |
But I'll push back because I have, so in that case, if you want to
link |
pay thousands of dollars, it should actually provide some value.
link |
I think it's easier for cheaper to provide value via a dumb AI that
link |
will take us towards AGI to just have a friend.
link |
I think there's an ocean of loneliness in the world.
link |
And I think an effective friend that doesn't have to be perfect,
link |
that doesn't have to be intelligent, that has to be empathic,
link |
having emotional intelligence, having ability to remember things,
link |
having ability to listen.
link |
Most of us don't listen to each other.
link |
One of the things that love and when you care about somebody,
link |
when you love somebody is when you listen.
link |
And that is something we treasure about each other.
link |
And if an AI can do that kind of thing.
link |
I think that provides a huge amount of value and very importantly,
link |
provides value in its ability to listen and understand
link |
versus provide really good advice.
link |
I think providing really good advice is another next level step
link |
that I think is just easier to do companionship.
link |
Yeah, I wouldn't disagree.
link |
I mean, I think that there's very few things that I would argue
link |
can't be reduced to some kind of a narrow AI.
link |
I think we can do a trillion dollars of value easily
link |
and all the things that can be done there.
link |
And a lot of it can be done with smoke and mirrors
link |
without having to go the whole thing.
link |
I mean, there's going to be the equivalent of the doom version
link |
for the AGI that's not really AGI, it's all smoke and mirrors,
link |
but it happens to do enough valuable things that it's enormously
link |
useful and valuable to people.
link |
But at some point, you do want to get to the point
link |
where you have the fully general thing
link |
and you stop making bespoke specialized systems for each thing
link |
and you wind up start using the higher level language
link |
instead of writing everything in assembly language.
link |
What about consciousness?
link |
The C word, do you think that's fundamental to solving AGI
link |
or is it a quirk of human cognition?
link |
So I think most of the arguments about consciousness
link |
don't have a whole lot of merit.
link |
I think that consciousness is kind of the way the brain feels
link |
when it's operating.
link |
And this idea that, you know, I do generally subscribe
link |
to sort of the pandemonium theories of consciousness
link |
where there's all these things bubbling around
link |
and I think of them as kind of slightly randomized sparse
link |
distributed memory bit strings of things that are kind of happening,
link |
recalling different associative memories,
link |
and eventually you get some level of consensus
link |
and it bubbles up to the point of being a conscious thought there.
link |
And the little bits of stochasticity that are sitting on in this
link |
as it cycles between different things and recalls different memory,
link |
that's largely our imagination and creativity.
link |
So I don't think there's anything deeply magical about it,
link |
certainly not symbolic.
link |
I think it is generally the flow of these associations
link |
drawn up with stochastic noise overlaid on top of them.
link |
And I think so much of that is like it depends on
link |
what you happen to have in your field of view
link |
as some other thought was occurring to you
link |
that overlay and blend into the next key that queries your memory for things.
link |
And that kind of determines how your chain of consciousness goes.
link |
So that's kind of the qualia, the subjective experience of it
link |
is not essential for intelligence.
link |
So I don't think there's anything really important there.
link |
What about some other human qualities,
link |
like fear of mortality and stuff like that?
link |
Like the fact that this ride ends, is that important?
link |
Like you talked so much about this conversation
link |
about the value of deadlines and constraints.
link |
Do you think that's important for intelligence?
link |
That's actually a super interesting angle
link |
that I don't usually take on that about
link |
has death being a deadline that forces you to make better decisions.
link |
Because I have heard people talk about how if you have immortality,
link |
people are going to stop trying and working on things
link |
because they've got all the time in the world.
link |
But I would say that I don't expect it to be a super critical thing
link |
that a sense of mortality and impending death is necessary there
link |
because those are things that they do wind up providing reward signals to us
link |
and we will be in control of the reward signals.
link |
And there will have to be something fundamental
link |
that causes, that engenders curiosity and goal setting and all of that.
link |
Something is going to play in there at the reward level,
link |
whether it's positive or negative or both.
link |
I don't have any strong opinions on exactly what it's going to be,
link |
but that's that type of thing where I doubt
link |
that might be one of those half dozen key things
link |
that has to be sorted out on exactly what the master reward,
link |
the meta reward over all of the local task specific rewards have to be.
link |
That could be that big negative reward of death.
link |
Maybe not death, but ability to walk away from an interaction.
link |
So it bothers me when people treat AI systems like servants.
link |
So it doesn't bother me, but I mean it really is drawing the line
link |
between what an AI system could be.
link |
It's limiting the possibility of what an AI system could be.
link |
It's treating them as justice tools.
link |
Now that's, of course, from a narrow AI perspective,
link |
there's so many problems that narrow AI could solve,
link |
just like you said, in its form of a tool.
link |
But it could also be a being, which is much more than a tool.
link |
And to become a being, you have to respect that thing for being a being.
link |
And for that, it has to be able to make its own decisions,
link |
to walk away, to say, I had enough of you.
link |
I would like to break up with you now.
link |
You've not treated me well, and I would like to move on.
link |
So I think that actually that choice to end things.
link |
So a couple of things on that.
link |
So on the one hand, it is kind of disturbing
link |
when you see people being like people that are mean to robots
link |
and mean to Alexa or whatever.
link |
And that seems to speak badly about humanity.
link |
But there's also the exact opposite side of that
link |
where you have so many people that imbue humanity in inanimate objects
link |
or things that are toys or that are relatively limited.
link |
So I think there may even be more danger
link |
about people putting more emotional investment
link |
into a lot of these proto AIs in different ways.
link |
And then the AI would manipulate that.
link |
But as far as like the AI ethnic sides of things,
link |
I really stay away from any of those discussions
link |
or even really thinking about it.
link |
It's similar with the safety things where I think it's just premature.
link |
And there's a certain class of people
link |
that enjoy thinking about impractical things,
link |
things that are not in the world and of pragmatic effect around you.
link |
And I think that, again, because I don't think there's going to be a fast takeoff,
link |
I think we actually will have time to have these debates
link |
when we know the shape of what we're debating.
link |
And some people do take a principled approach
link |
that they think it's going to go too fast,
link |
that you really do need to get ahead of it,
link |
that you need to be thinking about this
link |
because we have slow processes of coming to any kind of consensus
link |
or even coming up with ideas about this.
link |
And maybe that's true.
link |
I wouldn't put any of my money or funding into something like that
link |
because I don't think it's a problem yet.
link |
And I think that we will have these signs of life
link |
when we've got our learning disabled toddler,
link |
we should really start talking about some of the safety and ethics issues,
link |
but probably not before then.
link |
Can you elaborate briefly about why you don't think there'll be a fast takeoff?
link |
Is there some deep intuition you have about it?
link |
Is it because it's grounded in the physical world or why?
link |
Yeah, so it is my belief that we're going to start off
link |
with something that requires thousands of GPUs.
link |
And I don't know if you've tried to go get a thousand GPU instance
link |
on a cloud anytime recently,
link |
but these are not things that you can just go spin up hundreds of.
link |
There are real challenges to, I mean,
link |
these things are going to take data centers
link |
and data centers take years to build.
link |
The last few years, we've seen a few of them kind of coming up,
link |
going in different places.
link |
They're big engineering efforts.
link |
You can hear people bemoan about the fact that I know
link |
that the network was wired all wrong
link |
and it took them a month to go unwire it and rewire it the right way.
link |
These aren't things that you can just magic into existence.
link |
And the ideas of like the old tropes about it's going to escape
link |
onto the internet and take over other systems,
link |
the fast takeoff ones are clearly nonsense
link |
because you just can't open TCP connections above a certain rate,
link |
no matter how smart you are, even if you have perfect hacking ability,
link |
that take over the world in an instant sort of thing
link |
just isn't plausible at all.
link |
And even if you had access to all of the resources,
link |
these are going to be specialized systems
link |
where you're going to wind up with something
link |
that is architected around exactly this chip with this interconnect,
link |
and it's not just going to be able to be plopped somewhere else.
link |
Now, interestingly, it is going to be something
link |
that the entire code for all of it will easily fit on a thumb drive.
link |
That's total spy movie, thriller sorts of things
link |
where you could have, hey, we cracked the secret to AGI
link |
and it fits on this thumb drive and anyone could steal it.
link |
Now, they're still going to have to build the right data center to deploy it
link |
and have the right kind of life experience curriculum
link |
to take it up to the point where it's valuable.
link |
But the real core of it, the magic that's going to happen there
link |
is going to be very small.
link |
It's, again, tens of thousands of lines of code,
link |
not millions of lines of code.
link |
It is possible to imagine a world, as you mentioned in the spy thriller view,
link |
if it's just a few lines of code,
link |
we can imagine a world where the surface of computation is growing,
link |
maybe growing exponentially,
link |
meaning the refrigerators start getting a GPU.
link |
First of all, the smartphones, the billions of smartphones.
link |
But maybe if they become highways
link |
through which code can spread across the entirety of the computation surface,
link |
then you don't any longer have to book AWS GPUs.
link |
There were real fundamental issues there.
link |
When you start getting down to taking an actual problem
link |
and putting it on an abstract machine like that,
link |
that has not worked out well in practice.
link |
And the idea that there was always,
link |
like it's always been easy to come up with ways to compute faster,
link |
say more flops or more giga ops or whatever there.
link |
That's usually the easy part.
link |
But you then have interconnect and then memory for what goes into it.
link |
And when you talk about saying, well, cell phones,
link |
well, you're limited to like a 5G connection or something on that.
link |
And if you take your calculation
link |
and you factor it across a million cell phones
link |
instead of a thousand GPUs in a warehouse,
link |
you might be able to have some kind of a substrate like that,
link |
but it could be operating then at one one thousandth the speed.
link |
And so, yes, you could have an AGI working there,
link |
but it wouldn't be a real time AGI.
link |
It would be something that is operating at really a snail's pace,
link |
much, much slower than kind of human level thought for things.
link |
I'm not worried about that problem.
link |
You're transferring the problem into the interconnect,
link |
the communication, the shared memory,
link |
the collective intelligence aspect of it,
link |
which is extremely difficult as well.
link |
I mean, it's back to the very earliest days of supercomputers.
link |
You still have the balance between bandwidth, storage and computation.
link |
And sometimes they're easier to get one or the other,
link |
but it's been remarkably constant across all those years
link |
that you still need all three.
link |
What do your efforts now, you mentioned to me
link |
that you're really committing to AI at this stage.
link |
What do you see your life in the next few months, years look like?
link |
What do you hope to achieve here?
link |
So I literally just this week signed a term sheet
link |
to take some investment money for my company
link |
where the last two years I had backed off from Meta
link |
and I was still doing my consulting CTO role there,
link |
but I had styled it as I was going to take
link |
the Victorian gentleman scientist route
link |
where I was going to be the wealthy person
link |
that was going to go pursue science
link |
and learn about this and do experiments.
link |
And honestly, I'm surprised there aren't more people like that
link |
that are like me, technical people that made a bunch of money
link |
and are interested in some of these,
link |
possibly the biggest leverage point in human history.
link |
I mean, I know of, I've heard of a couple organizations
link |
that are basically led by one rich techie guy
link |
that gets a few people around him to try to work on this,
link |
but I'm surprised that there's not more,
link |
that there aren't like a dozen of them.
link |
I mean, maybe people still think that it's an unapproachable problem,
link |
that it's kind of beyond their ability to get a wrench on
link |
and have some effect on like whatever startups they've run before.
link |
But that was my kind of, like with all the stuff I've learned,
link |
whether it's gaming, aerospace, whatever,
link |
I go through a larval phase where I'm like,
link |
okay, I'm sucking up all of this information trying to see,
link |
is this something that I can actually do?
link |
Is this something that's practical to devote a large chunk of my life to?
link |
And I've gone through that with the AI machine learning space of things.
link |
And I think I've got my arms around it.
link |
I've got the measure of it where some of the most brilliant people
link |
in the world are working on this problem,
link |
but nobody knows exactly the path that it's going on.
link |
We're throwing a lot of things at the wall and seeing what sticks.
link |
But I have another interesting thing just learning about all of this,
link |
the contingency of your path to knowledge
link |
and talking about the associations and the context that you have with them
link |
where people that learn in the same path will have similar thought processes.
link |
And I think it's useful that I come at this from a different background,
link |
a different history than the people that have had
link |
the largely academic backgrounds for this
link |
where I have huge blind spots that they could easily point out,
link |
but I have a different set of experiences in history
link |
and approaches to problems in systems engineering
link |
that might turn out to be useful.
link |
And I can afford to take that bet where I'm not going to be destitute.
link |
I have enough money to fund myself working on this for the rest of my life.
link |
But what I was finding is that I was still not committing
link |
where I had a foot firmly in the VR and meta side of things
link |
where in theory, I've got a very nice position there.
link |
I only have to work one day a week for my consulting role,
link |
but I was engaging every day.
link |
I'd still be like my computer's there.
link |
I'd be going and checking the workplace and notes
link |
and testing different things and communicating with people.
link |
But I did make the decision recently that I'm going to get serious.
link |
I'm still going to keep my ties with meta,
link |
but I am seriously going for the AGI side of things.
link |
And it's actually a really interesting point
link |
because a lot of the machine learning, the AI community is quite large,
link |
but really, basically almost everybody has taken the same trajectory
link |
through life in that community.
link |
And it's so interesting to have somebody like you
link |
with a fundamentally different trajectory,
link |
and that's where the big solutions can come
link |
because there is a kind of silo,
link |
and it is a bunch of people kind of following the same kind of set of ideas.
link |
And I was really worried that I didn't want to come off
link |
as an arrogant outsider for things
link |
where I have all the respect in the world for the work.
link |
It's been a miracle decade.
link |
We're in the midst of a scientific revolution happening now,
link |
and everybody doing this is, you know, these are the Einsteins and Boers
link |
and whatevers of our modern era.
link |
And I was really happy to see that the people that I've sat down and talked with,
link |
everybody does seem to really be quite great about,
link |
just happy to talk about things, willing to acknowledge that
link |
we don't know what we're doing, we're figuring it out as we go along.
link |
And I mean, I've got a huge debt on this where this all really started for me
link |
because Sam Altman basically tried to recruit me to open AI.
link |
And it was at a point when I didn't know anything about
link |
what was really going on in machine learning.
link |
And in fact, it's funny how the first time you reached out to me,
link |
it's like four years ago for your AI podcast.
link |
Yeah, for people who are listening to this should know that,
link |
first of all, obviously I've been a huge fan of yours for the longest time,
link |
but we've agreed to talk like, yeah, like four years ago,
link |
back when this was called the artificial intelligence podcast,
link |
we wanted to do a thing and you said yes.
link |
And I said, it's like, I don't know anything about modern AI.
link |
I said, I could kind of take an angle on machine perception
link |
because I'm doing a lot of that with the sensors and the virtual reality,
link |
but we could probably find something to talk about.
link |
And that's where, when did Sam talk to you about open AI, around the same time?
link |
It was a little bit, it was a bit after that.
link |
So I had done the most basic work, I had kind of done the neural networks
link |
from scratch where I had gone and written it all in C just to make sure
link |
I understood backpropagation at the lowest level and my nuts and bolts approach.
link |
But after Sam approached me, it was flattering to think that he thought
link |
that I could be useful at open AI largely for kind of like systems optimization
link |
sorts of things without being an expert.
link |
But I asked Ilya Sutskever to give me a reading list and he gave me a binder
link |
full of all the papers that like, okay, these are the important things.
link |
If you really read and understand all of these, you'll know like 80% of what most
link |
of the machine language researchers work on.
link |
And I went through and I read all those papers multiple times
link |
and highlighted them and went through and kind of figured the things out there
link |
and then started branching out into my own sets of research on things
link |
and eventually started writing my own experiments and doing kind of figuring out,
link |
you know, finding out what I don't know, what the limits of my knowledge are
link |
and starting to get some of my angles of attack on things,
link |
the things that I think are a little bit different from what people are doing.
link |
And I've had a couple of years now, like two years since I kind of left
link |
the full time position at Meta and now I've kind of pulled the trigger and said,
link |
I'm going to get serious about it but some of my lessons all the way back
link |
to Armadillo Aerospace about how I know I need to be more committed to this
link |
where there is that, you know, it's both a freedom and a cost in some ways
link |
when you know that you're wealthy enough to say it's like this doesn't really mean anything.
link |
I can spend, you know, I can spend a million dollars a year for the rest of my life
link |
and it doesn't mean anything, it's fine.
link |
But that is an opportunity to just kind of meander
link |
and I could see that in myself when I'm doing some things.
link |
It's like, oh, this is a kind of interesting, curious thing.
link |
Let's look at this for a little while, let's look at that.
link |
It's not really bearing down on the problem.
link |
So there's a few things that I've done that are kind of tactics for myself
link |
to make me more effective.
link |
Like one thing I noticed I was not doing well is I had a Google Cloud account
link |
to get GPUs there and I was finding I was very rarely doing that
link |
for no good psychological reasons where I'm like, oh, I can always think of something to do
link |
other than to spin up instances and run an experiment.
link |
I can keep working on my local Titans or something.
link |
But it was really stupid.
link |
I mean, it was not a lot of money.
link |
I should have been running more experiments there.
link |
So I thought to myself, well, I'm going to go buy a quarter million dollar DGX station.
link |
I'm going to just like sit it right there and it's going to mock me if I'm not using it.
link |
If the fans aren't running on that thing, I'm not properly utilizing it.
link |
And that's been helpful.
link |
You know, I've done a lot more experiments since then.
link |
It's been it's been interesting where I thought I'd be doing all this low level
link |
envy link optimized stuff.
link |
But 90 percent of what I do is just spin up four instances of an experiment
link |
with different hyper parameters on it.
link |
You're doing like really sort of building up intuition by doing
link |
ML experiments of different kinds.
link |
But so the next big thing, though, is I am you know, I decided that I was going
link |
to take some take some investor money because I have an overactive sense
link |
of responsibility about other people's money.
link |
And it's like I I don't want I mean, a lot of my my push and my passionate
link |
entreaties for things that matter.
link |
It's like I don't want to have wasted his money investing in Oculus.
link |
I want it to work out.
link |
I want it to change the world.
link |
I want it to be worth all of this time, money and effort going into it.
link |
And I expect that it's going to be that like that with my with my company
link |
where it's a huge forcing function.
link |
This investment investors that are going to expect something of me now.
link |
We've all had the conversation that this is a low probability long term bet.
link |
It's not something that there's a million things I could do that I would have line
link |
of sight on the value proposition for.
link |
This isn't that I think there are there are unknown unknowns in the way.
link |
But it's one of these things that it's, you know, it's hyperbole.
link |
But it's potentially one of the most important things humans ever do.
link |
And it's something that I think is within our lifetimes, if not within a decade
link |
So, yeah, this is just now happening.
link |
Like term sheet, like the ink is barely virtually barely dry.
link |
I mean, as I mentioned to you offline, like somebody I admire and somebody
link |
you know, Andre Karpathy, I think the two of you different trajectories in life,
link |
but approach problems similarly in that he calls stuff from scratch up all the time.
link |
And I he's created a bunch of little things outside of even outside the course
link |
at Stanford that have been tremendously useful to build up intuition about stuff,
link |
but also to help people.
link |
And they're all in the realm of A.I.
link |
Do you see yourself potentially doing things like this?
link |
You know, not necessarily solving a gigantic problem, but on the journey,
link |
on the path to that building up intuitions and sharing code or ideas or systems
link |
that give inklings of AGI, but also kind of are useful to people in some way.
link |
So, yeah, first of all, Andre is awesome.
link |
I learned a lot when I was going through my larval phase from his blog posts
link |
and his Stanford course and, you know, super valuable.
link |
I got to meet him first a couple of years ago when I was first kind of starting off
link |
on my gentleman scientist bit.
link |
And just just a couple of months ago when he went out on his sabbatical,
link |
he stopped by in Dallas and we talked for a while and I had a great time with him.
link |
And then when I heard he actually left Tesla, I did, of course,
link |
along with a hundred other people say, hey, if you ever want to work with me,
link |
it would be an honor.
link |
So, he thinks that he's going to be doing this educational work,
link |
but I think someone's going to make him an offer he can't refuse
link |
before he gets too far along on it.
link |
Oh, his current interest is educational.
link |
So, yeah, he's a special mind.
link |
Is there something you could speak to what makes him so special?
link |
So, you know, like he did, he was very much a programmer's programmer
link |
that was doing machine learning work rather than,
link |
it's a different feel than an academic where you can see it in paper sometimes
link |
where somebody that's really a mathematician or a statistician at heart
link |
and they're doing something with machine learning.
link |
But, you know, Andre is about getting something done.
link |
And you could see it in like all of his earliest approaches to it's like,
link |
okay, here's how reinforcement learning works.
link |
Here's how recurrent neural networks work.
link |
Here's how transformers work.
link |
Here's how crypto works.
link |
And, yeah, it's just, he's a hacker's, you know,
link |
one of his old posts was like a hacker's guide to machine learning.
link |
And, you know, he deprecated that and said,
link |
don't really pay attention to what's in here,
link |
but it's that thought that carries through in a lot of it
link |
where it is that back again to that hacker mentality
link |
and the hacker ethic with what he's doing and sharing all of it.
link |
Yeah, and a lot of his approach to a new thing, like you said,
link |
larva stage is let me code up the simplest possible thing to build up intuition about it.
link |
Yeah, like I say, I sketch with structs and things.
link |
When I'm just thinking about a problem, I'm thinking in some degree of code.
link |
You are also, among many things, a martial artist, both Judo and Jiu Jitsu.
link |
How has this helped make you the person you are?
link |
So, I mean, I was a competent club player in Judo and grappling.
link |
I mean, I was, you know, by no means any kind of a superstar, but it was,
link |
I went through a few phases with it where I did some when I was quite young,
link |
a little bit more when I was 17.
link |
And then I got into it kind of seriously in my mid 30s.
link |
And, you know, I went pretty far with it.
link |
And I was pretty good at some of the things that I was doing.
link |
And I did appreciate it quite a bit where, I mean, on the one hand,
link |
it's always if you're going to do exercise or something,
link |
it's a more motivating form of exercise.
link |
If someone is crushing you, you are motivated to do something about that,
link |
to up your attributes and be better about getting out of that.
link |
Up your attributes, yes.
link |
But there's also that sense that I was not a sports guy.
link |
I did do wrestling in junior high.
link |
And I often wish that, I think I would have been good for me
link |
if I'd carried that on into high school and had a little bit more of that.
link |
I mean, it's like I, you know, filch a little bit of wrestling vibe
link |
with always going on about embracing the grind and like that push
link |
that I associate with the wrestling team that in hindsight,
link |
I wish I had gone through that and pushed myself that way.
link |
But even getting back into Judo and Jiu Jitsu in my mid 30s,
link |
as usually the old man on the mat with that, there was still the, you know,
link |
the sense that, you know, working out with the group
link |
and having the guys that you're beating each other up with it,
link |
but you just feel good coming out of it.
link |
And I can remember those driving home aching in various ways
link |
and just thinking, it's like, oh, that was really great.
link |
And, you know, it's mixing with a bunch of people that had nothing to do
link |
with any of the things that I worked with.
link |
You know, every once in a while, someone would be like,
link |
oh, you're the Doom guy.
link |
But for the most part, it was just different slice of life.
link |
You know, a good thing.
link |
And I made the call when I was 40.
link |
That's like, maybe I'm getting a little old for this.
link |
I had separated a rib and tweaked a few things.
link |
And I got out of it without any really bad injuries.
link |
And it was like, have I dodged enough bullets?
link |
Should I, you know, should I hang it up?
link |
I've gone a couple of times in the last decade
link |
trying to get my kids into it a little bit.
link |
I didn't really stick with any of them,
link |
but it was fun to get back on the mats.
link |
It really hurts for a while when you haven't gone for a while.
link |
But I still debate this pretty constantly.
link |
My brother's only a year younger than me,
link |
and he's going kind of hard in jiu jitsu right now.
link |
You know, he won a few medals at the last tournament he was at.
link |
So he's competing, too.
link |
Yeah, and I was thinking, yeah, I guess we're in the executive division
link |
if you're over 50 or over 45 or something.
link |
And it's not out of the question that I go back at some point to do some of this.
link |
But again, I'm just reorganizing my life around more focus.
link |
Probably not going to happen.
link |
I'm pushing my exercise around to give me a longer,
link |
uninterrupted intellectual focus time,
link |
pushing it to the beginning or the end of the game.
link |
Like running and stuff like that, walking.
link |
Yeah, I got running and calisthenics and some things like that.
link |
It allows you to still think about a problem.
link |
But if you're going to a judo club or something, you've got it fixed.
link |
It's going to be 7 o clock or whatever, 10 o clock on Saturday.
link |
Although, I talked about this a little bit when I was on Rogan.
link |
And shortly after that, Carlos Machado did reach out.
link |
And I had trained with him for years back in the day.
link |
And he was like, hey, we've got kind of a small private club
link |
with a bunch of executive type people.
link |
And it does tempt me.
link |
Yeah, I don't know if you know him, but John Donahart moved here to Austin
link |
with Gordon Ryan and a few other folks.
link |
He has a very interesting way, very deep systematic way of thinking about jiu jitsu
link |
that reveals the chest of it, the science of it.
link |
And I do think about that more as kind of an older person
link |
considering the martial arts where I can remember the very earliest days
link |
getting back into judo and I'm like, teach me submissions right now.
link |
It's like, learn the arm bar, learn the choke.
link |
But as you get older, you start thinking more about,
link |
it's like, okay, I really do want to like learn the entire canon of judo.
link |
It's like all the different things there and like all the different approaches for it.
link |
Not just the, you know, if you want to compete,
link |
there's just a handful of things you learn really, really well.
link |
But sometimes there's interest in learning a little bit more of the scope there
link |
and figuring some things out from, you know, at one point I had wasn't exactly a spreadsheet,
link |
but I did have a big long text file with like, here's the things that I learned
link |
and here are like ways you chain this together.
link |
And while, when I went back a few years ago,
link |
it was good to see that I whipped myself back into reasonable shape
link |
about doing the basic grappling.
link |
But I know there was a ton of the subtleties that were just, that were gone,
link |
but could probably be brought back reasonably quickly.
link |
And there's also the benefit.
link |
I mean, you're exceptionally successful now.
link |
And the problem, the old problem of the ego.
link |
I still pushed kind of harder than I should.
link |
I mean, that was, I was one of those people that I, you know,
link |
I'm on the smaller side for a lot of the people competing and I would, you know,
link |
I'd go with all the big guys and I'd go hard and I'd push myself a lot.
link |
And that would be one of those where I would, you know,
link |
I'd be dangerous to anyone for the first five minutes,
link |
but then sometimes after that I'm already dead.
link |
And I knew it was terrible for me because it made the, you know,
link |
it meant I got less training time with all of that when you go and you just gas
link |
out relatively quickly there.
link |
And I like to think that I would be better about that where after I gave up judo,
link |
I started doing the half marathons and tough butters and things like that.
link |
And so when I did go back to the local judokai club, I thought it's like, oh,
link |
I should have better cardio for this because I'm a runner now and I do all of
link |
this and didn't work out that way.
link |
It was the same old thing where just push really hard, strain really hard.
link |
And of course, when I worked with good guys like Carlos, it's like just the
link |
whole flow like water thing is real and he's just like.
link |
That's true with judo too.
link |
Some of the best people like I've trained with Olympic gold medalists and for
link |
some reason with them, everything's easier.
link |
Everything is you actually start to feel the science of it, the music of it,
link |
Everything is effortless.
link |
You understand that there's an art to it.
link |
It's not just an exercise.
link |
It was interesting where I did go to the Kodokan in Japan.
link |
I kind of the birthplace of judo and everything.
link |
And I remember I rolled with one old guy.
link |
I didn't didn't start standing, just started on groundwork and it was and it
link |
was striking how different it was from Carlos.
link |
He was still he was better than me and he got my arm and I had to tap there.
link |
But it was a completely different style where I just felt like I could do
link |
He was just enveloping me and just like slowly grounded down with my arm and
link |
bent it while with Carlos, you know, he's just loose and free.
link |
And you always thought like, oh, you're just going to go grab something.
link |
But you never had any chance to do it.
link |
But it was very different feeling.
link |
That's a good summary of the difference between jiu jitsu and judo and jiu
link |
There's it is a dance and you feel like there's a freedom and actually anybody
link |
like the Gordon Ryan, one of the best the best grappler in the world.
link |
Nogi grappler in the world. There's a feeling like you can do anything.
link |
But when you actually try to do something, you can't just magically doesn't
link |
But with the best judo players in the world, it does feel like there's a
link |
blanket that weighs a thousand pounds on top of you.
link |
And there's not a feeling like you can do anything.
link |
You just you're trapped.
link |
And that's a that's a style that's a difference in the style of martial
link |
arts. But it's also once you start to study, you understand it all has to
link |
do with human movement and the physics of it and the leverage and all that
link |
And that's like that's super fascinating.
link |
At the end of the day, for me, the biggest benefits and the humbling
link |
aspect when another human being kind of tells you that, you know, there's
link |
a hierarchy or there's a you're not that special.
link |
And in the most extreme case, when you tap to a choke, you are basically
link |
living because somebody let you live.
link |
And that's that is one of those.
link |
If you think about it, that is a closer brush with mortality than than most
link |
And that kind of humbling act is good to take to your work then where it's
link |
harder to get humbled.
link |
Nobody's nobody that does any martial art is coming out thinking I'm the
link |
best in the world at anything because everybody loses.
link |
Let me ask you for advice.
link |
What advice would you give to young people today about life, about career,
link |
how they can have a job, how they can have an impact, how they can have a
link |
life they could be proud of.
link |
So it was kind of fun.
link |
I got invited to give the commencement speech back at the I went to college
link |
for two semesters and dropped out and went on to do my tech stuff.
link |
But they still wanted me to come back and give a commencement speech.
link |
And I've got that pinned on my Twitter account.
link |
I still feel good about everything that I said there.
link |
And my biggest point was that the path for me might not be the path for
link |
And in fact, the advice, the path that I took and even the advice that I
link |
would give based on my experience and learnings probably isn't the best
link |
advice for everyone.
link |
Because what I did was all about this knowledge in depth.
link |
It was about not just having this surface level ability to make things do
link |
what I want, but to really understand them through and through.
link |
To let me do the systems engineering work and to sometimes find these
link |
inefficiencies that can be bypassed.
link |
And the whole world doesn't need that.
link |
Most programmers or engineers of any kind don't necessarily need to do that.
link |
They need to do a little job that's been parceled out to them.
link |
Let people depend on you.
link |
Do quality work with all of that.
link |
But people that do have an inclination for wanting to know things deeper and
link |
learn things deeper, there are just layers and layers of things out there.
link |
If you're the right person that is excited about that, the world's never
link |
been like this before.
link |
It's better than ever.
link |
I mean, everything that was wonderful for me is still there, and there's
link |
whole new worlds to explore on the different things that you can do.
link |
And that, you know, it's hard work.
link |
Embrace the grind with it.
link |
And understand as much as you can.
link |
And then be prepared for opportunities to present themselves.
link |
Where you can't just say this is my goal in life and just push at that.
link |
I mean, you might be able to do that, but you're going to make more total
link |
progress if you say I'm preparing myself with this broad set of tools.
link |
And then I'm being aware of all the way things are changing as I move through
link |
the world and as the whole world changes around me.
link |
And then looking for opportunities to deploy the tools that you've built.
link |
And there's going to be more and more of those types of things there where an
link |
awareness of what's happening, where the inefficiencies are, what things can
link |
be done, what's possible versus what's current practice.
link |
And then finding those areas where you can go and make an adjustment and make
link |
something that may affect millions or billions of people in the world.
link |
When, maybe from your own example, how were you able to recognize this about
link |
yourself that you saw the layers in a particular thing and you were drawn to
link |
discovering deeper and deeper truths about it?
link |
Is that something that was obvious to you that you couldn't help or is there some
link |
actions you had to take to actually allow yourself to dig deep?
link |
So in the earliest days of personal computers, I remember the reference
link |
manuals and the very early ones even had schematics of computers in the
link |
background, in the back of the books, as well as firmware listings and things.
link |
And I could look at that.
link |
And at that time, when I was a younger teenager, I didn't understand a lot of
link |
that stuff, how the different things worked.
link |
I was pulling out the information that I could get, but I always wanted to know
link |
There was kind of magical information sitting down there.
link |
It's like the elder lore that some gray beard wizard is the keeper of.
link |
And so I always felt that pull for wanting to know more, wanting to explore
link |
the mysterious areas there.
link |
And that followed right in through all the things that got the value,
link |
exploring the video cards leading to the scrolling advantages,
link |
exploring some of the academic papers and things, learning about BSP trees
link |
and the different things that I could do with those systems and just the huge
link |
larval phases going through aerospace, just reading bookshelves full of books.
link |
Again, that point where I have enough money, I can buy all the books I want.
link |
It was so valuable there where I was terrible with my money when I was a kid.
link |
My mom thought I would always be broke because I'd buy my comic books
link |
and just be out of money.
link |
But it was like all the pizza I want, all the Diet Coke I want,
link |
video games and then books.
link |
And it didn't take that much.
link |
As soon as I was making 27K a year, I felt rich and I was just getting
link |
all the things that I wanted.
link |
But that sense of books have always been magical to me.
link |
And that was one of the things that really made me smile is Andre had said
link |
when he came over to my house, he said he found my library inspiring.
link |
And it was great to see.
link |
I used to look at him.
link |
He's kind of a younger guy.
link |
I sometimes wonder if younger people these days have the same relationship
link |
with books that I do where they were such a cornerstone for me in so many ways.
link |
But that sense that, yeah, I always wanted to know it all.
link |
And that was like one of the last things I said.
link |
You can't know everything, but you should convince yourself
link |
that you can know anything.
link |
Any one particular thing, it was created and discovered by humans.
link |
You can find out what you need on there.
link |
And you can learn it deeply.
link |
And you can drive a nail down through whatever layer cake problem space
link |
you've got and learn a cross section there.
link |
And not only can you have an impact doing that,
link |
you can attain happiness doing that.
link |
There's something so fulfilling about becoming a craftsman of a thing.
link |
And I don't want to tell people that, look, this is a good career move.
link |
Just grit your teeth and bear it.
link |
And I do think it is possible sometimes to find the joy in something.
link |
Like it might not immediately appeal to you.
link |
But I had told people early on, like in software times,
link |
that a lot of game developers are in it just because they
link |
are so passionate about games.
link |
But I was always really more flexible in what appealed to me,
link |
where I said, I think I could be quite engaged doing
link |
operating system work or even database work.
link |
I would find the interest in that.
link |
Because I think most things that are significant in the world
link |
have a lot of layers and complexity to them
link |
and a lot of opportunities hidden within them.
link |
So that would probably be the most important thing to encourage to people
link |
is that you can weaponize curiosity.
link |
You can deploy your curiosity to find,
link |
to kind of like make things useful and valuable to you,
link |
even if they don't immediately appear that way.
link |
Deploy your curiosity.
link |
Yeah, that's very true.
link |
We've mentioned this debate point,
link |
whether mortality or fear of mortality is fundamental to creating an AGI.
link |
But let's talk about whether it's fundamental to human beings.
link |
Do you think about your own mortality?
link |
And you probably always have to like take with a grain of salt
link |
anything somebody says about fundamental things like that.
link |
But I don't think about really aging, impending death,
link |
having a legacy with my children, things like that.
link |
And clearly it seems most of the world does a lot more than I do.
link |
So I mean I think I'm an outlier in that where it doesn't wind up
link |
being a real part of my thinking and motivation about things.
link |
So daily existence is about sort of the people you love
link |
and the problems before you.
link |
And I'm focused on what I'm working on right now.
link |
I do take that back.
link |
There's one aspect where the kind of finiteness of the life does impact me,
link |
and that is about thinking about the scope of the problems that I'm working on.
link |
When I decided to work on it, when I was like nuclear fission or AGI,
link |
these are big ticket things that impact large fractions of the world.
link |
And I was thinking to myself at some level that, okay,
link |
I may have a couple more swings at bat with me at full capability,
link |
but yes, my mental abilities will decay with age, mostly inevitably.
link |
I don't think it's a 0% chance that we will address some of that
link |
before it becomes a problem for me.
link |
I think exciting medical stuff in the next couple decades.
link |
But I do have this kind of vague plan that when I'm not at the top of my game
link |
and I don't feel that I'm in a position to put a dent in the world some way,
link |
that I'll probably wind up doing some kind of recreational retro programming
link |
or I'll work on something that I would not devote my life to now,
link |
but I can while away my time as the old man gardening in the code worlds.
link |
And then to step back even bigger, let me ask you about why we're here,
link |
What's the meaning of it all?
link |
What's the meaning of life, John Carmack?
link |
So very similar with that last question.
link |
I know a lot of people fret about this question a lot,
link |
and I just really don't.
link |
I really don't give a damn.
link |
We are biological creatures that happenstance of evolution.
link |
We have innate drives that evolution crafted for survival
link |
and passing on of genetic codes.
link |
I don't find a lot of value in trying to go much deeper than that.
link |
I have my motivations, some of which are probably genetically coded
link |
and many of which are contingent on my upbringing
link |
and the path that I've had through my life.
link |
I don't run into like spades of depression or envy
link |
or anything that winds up being a challenge
link |
and forcing a degree of soul searching with things like that.
link |
I seem to be okay kind of without that.
link |
As a brilliant ant in the ant colony without looking up to the sky
link |
wondering why the hell am I here again.
link |
So the why of it, the incredible mystery of the fact that we started,
link |
first of all, the origin of life on Earth,
link |
and from that, from single cell organisms,
link |
the entirety of the evolutionary process took us somehow
link |
to this incredibly intelligent thing that is able to build Wolfenstein 3D
link |
and Doom and Quake and take a crack at the problem of AGI
link |
and things that eventually supersede human beings.
link |
That doesn't, the why of it is...
link |
It's been my experience that people that focus on,
link |
that don't focus on the here and now right in front of them
link |
tend to be less effective.
link |
I mean, it's not 100%, you know, vision matters to some people,
link |
but it doesn't seem to be a necessary motivator for me
link |
and I think that the process of getting there is usually done.
link |
It's like the magic of gradient descent.
link |
People just don't believe that just looking locally
link |
gets you to all of these spectacular things.
link |
That's been, you know, the decades of looking at
link |
really some of the smartest people in the world
link |
that would just push back forever against this idea
link |
that it's not this grand, sophisticated vision of everything,
link |
but little tiny steps, local information winds up
link |
leading to all the best answers.
link |
So the meaning of life is following locally
link |
wherever the gradient descent takes you.
link |
This was an incredible conversation,
link |
officially the longest conversation I've ever done on the podcast,
link |
which means a lot to me because I get to do it with one of my heroes.
link |
John, I can't tell you how much it means to me
link |
that you would sit down with me.
link |
You're an incredible human being.
link |
I can't wait what you do next, but you've already changed the world.
link |
You're an inspiration to so many people.
link |
And again, we haven't covered like most of what I was planning to talk about,
link |
so I hope we get a chance to talk someday in the future.
link |
I can't wait to see what you do next.
link |
Thank you so much again for talking to me.
link |
Thank you very much.
link |
Thanks for listening to this conversation with John Carmack.
link |
To support this podcast, please check out our sponsors in the description.
link |
And now, let me leave you with some words from John Carmack himself.
link |
Focused hard work is the real key to success.
link |
Keep your eyes on the goal,
link |
and just keep taking the next step towards completing it.
link |
If you aren't sure which way to do something,
link |
do it both ways and see which works better.
link |
Thank you for listening and hope to see you next time.