back to indexKyle Vogt: Cruise Automation | Lex Fridman Podcast #14
link |
The following is a conversation with Kyle Vogt.
link |
He's the president and the CTO of Cruise Automation,
link |
leading an effort to solve one of the biggest
link |
robotics challenges of our time, vehicle automation.
link |
He's a cofounder of two successful companies,
link |
Twitch and Cruise, that have each sold for a billion dollars.
link |
And he's a great example of the innovative spirit
link |
that flourishes in Silicon Valley.
link |
And now is facing an interesting and exciting challenge
link |
of matching that spirit with the mass production
link |
and the safety centered culture of a major automaker,
link |
like General Motors.
link |
This conversation is part of the MIT
link |
Artificial General Intelligence series
link |
and the Artificial Intelligence podcast.
link |
If you enjoy it, please subscribe on YouTube, iTunes,
link |
or simply connect with me on Twitter at Lex Friedman,
link |
And now here's my conversation with Kyle Vogt.
link |
You grew up in Kansas, right?
link |
Yeah, and I just saw that picture you had to hit know
link |
there, so I'm a little bit worried about that now.
link |
So in high school in Kansas City,
link |
you joined Shawnee Mission North High School Robotics Team.
link |
Now that wasn't your high school.
link |
That was the only high school in the area that had a teacher
link |
who was willing to sponsor our first robotics team.
link |
I was gonna troll you a little bit.
link |
Jog your mouth a little bit with that kid.
link |
I was trying to look super cool and intense.
link |
Because this was BattleBots, this is serious business.
link |
So we're standing there with a welded steel frame
link |
and looking tough.
link |
What does that drew you to robotics?
link |
Well, I think, I've been trying to figure this out
link |
for a while, but I've always liked building things
link |
And when I was really, really young,
link |
I wanted the Legos that had motors and other things.
link |
And then, you know, Lego Mindstorms came out
link |
and for the first time you could program Lego contraptions.
link |
And I think things just sort of snowballed from that.
link |
But I remember seeing, you know, the BattleBots TV show
link |
on Comedy Central and thinking that is the coolest thing
link |
in the world, I wanna be a part of that.
link |
And not knowing a whole lot about how to build
link |
these 200 pound fighting robots.
link |
So I sort of obsessively poured over the internet forums
link |
where all the creators for BattleBots would sort of hang out
link |
and talk about, you know, document their build progress
link |
And I think I read, I must have read like, you know,
link |
tens of thousands of forum posts from basically everything
link |
that was out there on what these people were doing.
link |
And eventually, like sort of triangulated how to put
link |
some of these things together and ended up doing BattleBots,
link |
which was, you know, it was like 13 or 14,
link |
which was pretty awesome.
link |
I'm not sure if the show's still running,
link |
but so BattleBots is, there's not an artificial intelligence
link |
component, it's remotely controlled.
link |
And it's almost like a mechanical engineering challenge
link |
of building things that can be broken.
link |
They're radio controlled.
link |
So, and I think that they allowed some limited form
link |
of autonomy, but, you know, in a two minute match,
link |
you're, in the way these things ran,
link |
you're really doing yourself a disservice by trying
link |
to automate it versus just, you know,
link |
do the practical thing, which is drive it yourself.
link |
And there's an entertainment aspect,
link |
just going on YouTube.
link |
There's like some of them wield an axe, some of them,
link |
I mean, there's that fun.
link |
So what drew you to that aspect?
link |
Was it the mechanical engineering?
link |
Was it the dream to create like Frankenstein
link |
and sentient being?
link |
Or was it just like the Lego, you like tinkering stuff?
link |
I mean, that was just building something.
link |
I think the idea of, you know,
link |
this radio controlled machine that can do various things.
link |
If it has like a weapon or something was pretty interesting.
link |
I agree, it doesn't have the same appeal as, you know,
link |
autonomous robots, which I, which I, you know,
link |
sort of gravitated towards later on,
link |
but it was definitely an engineering challenge
link |
because everything you did in that competition
link |
was pushing components to their limits.
link |
So we would buy like these $40 DC motors
link |
that came out of a winch,
link |
like on the front of a pickup truck or something.
link |
And we'd power the car with those
link |
and we'd run them at like double or triple
link |
their rated voltage.
link |
So they immediately start overheating,
link |
but for that two minute match, you can get, you know,
link |
a significant increase in the power output
link |
of those motors before they burn out.
link |
And so you're doing the same thing for your battery packs,
link |
all the materials in the system.
link |
And I think there was something,
link |
something intrinsically interesting
link |
about just seeing like where things break.
link |
And did you offline see where they break?
link |
Did you take it to the testing point?
link |
Like, how did you know two minutes?
link |
Or was there a reckless, let's just go with it and see.
link |
We weren't very good at battle bots.
link |
We lost all of our matches the first round.
link |
The one I built first,
link |
both of them were these wedge shaped robots
link |
because the wedge, even though it's sort of boring
link |
to look at is extremely effective.
link |
You drive towards another robot
link |
and the front edge of it gets under them
link |
and then they sort of flip over,
link |
it's kind of like a door stopper.
link |
And the first one had a pneumatic polished stainless steel
link |
spike on the front that would shoot out about eight inches.
link |
The purpose of which is what?
link |
Pretty ineffective actually, but it looked cool.
link |
And was it to help with the lift?
link |
No, it was just to try to poke holes in the other robot.
link |
And then the second time I did it,
link |
which is the following, I think maybe 18 months later,
link |
we had a titanium axe with a hardened steel tip on it
link |
that was powered by a hydraulic cylinder,
link |
which we were activating with liquid CO2,
link |
which had its own set of problems.
link |
So great, so that's kind of on the hardware side.
link |
I mean, at a certain point,
link |
there must have been born a fascination
link |
on the software side.
link |
So what was the first piece of code you've written?
link |
If you didn't go back there, see what language was it?
link |
What was it, was it EMAX, VAM?
link |
Was it a more respectable, modern ID?
link |
Do you remember any of this?
link |
Yeah, well, I remember, I think maybe when I was in
link |
third or fourth grade, I was at elementary school,
link |
had a bunch of Apple II computers,
link |
and we'd play games on those.
link |
And I remember every once in a while,
link |
something would crash or wouldn't start up correctly,
link |
and it would dump you out to what I later learned
link |
was like sort of a command prompt.
link |
And my teacher would come over and type,
link |
I actually remember this to this day for some reason,
link |
like PR number six, or PR pound six,
link |
which is peripheral six, which is the disk drive,
link |
which would fire up the disk and load the program.
link |
And I just remember thinking, wow, she's like a hacker,
link |
like teach me these codes, these error codes,
link |
that is what I called them at the time.
link |
But she had no interest in that.
link |
So it wasn't until I think about fifth grade
link |
that I had a school where you could actually
link |
go on these Apple II's and learn to program.
link |
And so it was all in basic, you know,
link |
where every line, you know, the line numbers are all,
link |
or that every line is numbered,
link |
and you have to like leave enough space
link |
between the numbers so that if you want to tweak your code,
link |
you go back and if the first line was 10
link |
and the second line is 20, now you have to go back
link |
And if you need to add code in front of that,
link |
you know, 11 or 12, and you hope you don't run out
link |
of line numbers and have to redo the whole thing.
link |
And there's go to statements?
link |
Yeah, go to and is very basic, maybe hence the name,
link |
And that was like, that was, you know,
link |
that's when, you know, when you first program,
link |
you see the magic of it.
link |
It's like, just like this world opens up with,
link |
you know, endless possibilities for the things
link |
you could build or accomplish with that computer.
link |
So you got the bug then, so even starting with basic
link |
and then what, C++ throughout, what did you,
link |
was there a computer programming,
link |
computer science classes in high school?
link |
Not, not where I went, so it was self taught,
link |
but I did a lot of programming.
link |
The thing that, you know, sort of pushed me in the path
link |
of eventually working on self driving cars
link |
is actually one of these really long trips
link |
driving from my house in Kansas to, I think, Las Vegas,
link |
where we did the BattleBots competition.
link |
And I had just gotten my, I think my learners permit
link |
or early drivers permit.
link |
And so I was driving this, you know, 10 hour stretch
link |
across Western Kansas where it's just,
link |
you're going straight on a highway
link |
and it is mind numbingly boring.
link |
And I remember thinking even then
link |
with my sort of mediocre programming background
link |
that this is something that a computer can do, right?
link |
Let's take a picture of the road,
link |
let's find the yellow lane markers
link |
and, you know, steer the wheel.
link |
And, you know, later I'd come to realize
link |
this had been done, you know, since the 80s
link |
or the 70s or even earlier, but I still wanted to do it.
link |
And sort of immediately after that trip,
link |
switched from sort of BattleBots,
link |
which is more radio controlled machines
link |
to thinking about building, you know,
link |
autonomous vehicles of some scale,
link |
start off with really small electric ones
link |
and then, you know, progress to what we're doing now.
link |
So what was your view of artificial intelligence
link |
What did you think?
link |
So this is before there's been waves
link |
in artificial intelligence, right?
link |
The current wave with deep learning
link |
makes people believe that you can solve
link |
in a really rich, deep way,
link |
the computer vision perception problem.
link |
But like before the deep learning craze,
link |
you know, how do you think about
link |
how would you even go about building a thing
link |
that perceives itself in the world,
link |
localize itself in the world, moves around the world?
link |
Like when you were younger, I mean,
link |
as what was your thinking about it?
link |
Well, prior to deep neural networks
link |
or convolutional neural nets,
link |
these modern techniques we have,
link |
or at least ones that are in use today,
link |
it was all heuristic space.
link |
And so like old school image processing,
link |
and I think extracting, you know,
link |
yellow lane markers out of an image of a road
link |
is one of the problems that lends itself
link |
reasonably well to those heuristic base methods, you know,
link |
like just do a threshold on the color yellow
link |
and then try to fit some lines to that
link |
using a huff transform or something
link |
and then go from there.
link |
Traffic light detection and stop sign detection,
link |
red, yellow, green.
link |
And I think you can, you could,
link |
I mean, if you wanted to do a full,
link |
I was just trying to make something that would stay
link |
in between the lanes on a highway,
link |
but if you wanted to do the full,
link |
the full, you know, set of capabilities
link |
needed for a driverless car,
link |
I think you could, and we've done this at cruise,
link |
you know, in the very first days,
link |
you can start off with a really simple,
link |
you know, human written heuristic
link |
just to get the scaffolding in place
link |
for your system, traffic light detection,
link |
probably a really simple, you know,
link |
color thresholding on day one
link |
just to get the system up and running
link |
before you migrate to, you know,
link |
a deep learning based technique or something else.
link |
And, you know, back in, when I was doing this,
link |
my first one, it was on a Pentium 203,
link |
233 megahertz computer in it.
link |
And I think I wrote the first version in basic,
link |
which is like an interpreted language.
link |
It's extremely slow because that's the thing
link |
I knew at the time.
link |
And so there was no, no chance at all of using,
link |
there's no computational power to do
link |
any sort of reasonable deep nets like you have today.
link |
So I don't know what kids these days are doing.
link |
Are kids these days, you know, at age 13
link |
using neural networks in their garage?
link |
I mean, that would be awesome.
link |
I get emails all the time from, you know,
link |
like 11, 12 year olds saying, I'm having, you know,
link |
I'm trying to follow this TensorFlow tutorial
link |
and I'm having this problem.
link |
And the general approach in the deep learning community
link |
is of extreme optimism of, as opposed to,
link |
you mentioned like heuristics, you can,
link |
you can, you can separate the autonomous driving problem
link |
into modules and try to solve it sort of rigorously,
link |
where you can just do it end to end.
link |
And most people just kind of love the idea that,
link |
you know, us humans do it end to end,
link |
we just perceive and act.
link |
We should be able to use that,
link |
do the same kind of thing with your own nets.
link |
And that, that kind of thinking,
link |
you don't want to criticize that kind of thinking
link |
because eventually they will be right.
link |
Yeah. And so it's exciting.
link |
And especially when they're younger to explore that
link |
is a really exciting approach.
link |
But yeah, it's, it's changed the, the language,
link |
the kind of stuff you're tinkering with.
link |
It's kind of exciting to see when these teenagers grow up.
link |
Yeah, I can only imagine if you, if your starting point
link |
is, you know, Python and TensorFlow at age 13,
link |
where you end up, you know,
link |
after 10 or 15 years of that, that's, that's pretty cool.
link |
Because of GitHub, because the state tools
link |
for solving most of the major problems
link |
that are artificial intelligence
link |
are within a few lines of code for most kids.
link |
And that's incredible to think about,
link |
also on the entrepreneurial side.
link |
And, and, and at that point, was there any thought
link |
about entrepreneurship before you came to college
link |
is sort of doing your building this into a thing
link |
that impacts the world on a large scale?
link |
Yeah, I've always wanted to start a company.
link |
I think that's, you know, just a cool concept
link |
of creating something and exchanging it
link |
for value or creating value, I guess.
link |
So in high school, I was, I was trying to build like,
link |
you know, servo motor drivers, little circuit boards
link |
and sell them online or other, other things like that.
link |
And certainly knew at some point I wanted to do a startup,
link |
but it wasn't really, I'd say until college until I felt
link |
like I had the, I guess the right combination
link |
of the environment, the smart people around you
link |
and some free time and a lot of free time at MIT.
link |
So you came to MIT as an undergrad 2004.
link |
And that's when the first DARPA Grand Challenge
link |
The timing of that is beautifully poetic.
link |
So how'd you get yourself involved in that one?
link |
Originally there wasn't a
link |
Yeah, faculty sponsored thing.
link |
And so a bunch of undergrads, myself included,
link |
started meeting and got together
link |
and tried to, to haggle together some sponsorships.
link |
We got a vehicle donated, a bunch of sensors
link |
and tried to put something together.
link |
And so we had, our team was probably mostly freshmen
link |
and sophomores, you know, which, which was not really
link |
a fair, fair fight against maybe the, you know, postdoc
link |
and faculty led teams from other schools.
link |
But we, we got something up and running.
link |
We had our vehicle drive by wire and, you know,
link |
very, very basic control and things, but on the day
link |
of the qualifying, sort of pre qualifying round,
link |
the one and only steering motor that we had purchased,
link |
the thing that we had, you know, retrofitted to turn
link |
the steering wheel on the truck died.
link |
And so our vehicle was just dead in the water, couldn't steer.
link |
So we didn't make it very far.
link |
On the hardware side.
link |
So was there a software component?
link |
Was there, like, how did your view of autonomous vehicles
link |
in terms of artificial intelligence
link |
evolve in this moment?
link |
I mean, you know, like you said,
link |
from the 80s has been autonomous vehicles,
link |
but really that was the birth of the modern wave.
link |
The, the thing that captivated everyone's imagination
link |
that we can actually do this.
link |
So how, were you captivated in that way?
link |
So how did your view of autonomous vehicles
link |
change at that point?
link |
I'd say at that point in time, it was, it was a curiosity
link |
as in like, is this really possible?
link |
And I think that was generally the spirit
link |
and the purpose of that original DARPA Grand Challenge,
link |
which was to just get a whole bunch
link |
of really brilliant people exploring the space
link |
and pushing the limits.
link |
And, and I think like to this day,
link |
that DARPA challenge with its, you know,
link |
million dollar prize pool was probably one
link |
of the most effective, you know, uses of taxpayer money,
link |
dollar for dollar that I've seen, you know,
link |
because that, that small sort of initiative
link |
that DARPA put put out sort of, in my view,
link |
was the catalyst or the tipping point
link |
for this, this whole next wave
link |
of autonomous vehicle development.
link |
So that was pretty cool.
link |
So let me jump around a little bit on that point.
link |
They also did the urban challenge where it was in the city,
link |
but it was very artificial and there's no pedestrians
link |
and there's very little human involvement
link |
except a few professional drivers.
link |
Do you think there's room, and then there was
link |
the robotics challenge with human robots?
link |
So in your now role as looking at this,
link |
you're trying to solve one of the, you know,
link |
autonomous driving, one of the harder,
link |
more difficult places in San Francisco.
link |
Is there a role for DARPA to step in
link |
to also kind of help out, like,
link |
challenge with new ideas, specifically pedestrians
link |
and so on, all these kinds of interesting things?
link |
Well, I haven't thought about it from that perspective.
link |
Is there anything DARPA could do today
link |
to further accelerate things?
link |
And I would say my instinct is that that's maybe not
link |
the highest and best use of their resources in time
link |
because, like, kick starting and spinning up the flywheel
link |
is I think what they did in this case
link |
for very, very little money.
link |
But today this has become,
link |
this has become, like, commercially interesting
link |
to very large companies and the amount of money
link |
going into it and the amount of people,
link |
like, going through your class and learning
link |
about these things and developing these skills
link |
is just, you know, orders of magnitude
link |
more than it was back then.
link |
And so there's enough momentum and inertia
link |
and energy and investment dollars into this space right now
link |
that I don't, I don't, I think they're,
link |
I think they're, they can just say mission accomplished
link |
and move on to the next area of technology
link |
So then stepping back to MIT,
link |
you left MIT Junior Junior year,
link |
what was that decision like?
link |
As I said, I always wanted to do a company
link |
or start a company and this opportunity landed in my lap
link |
which was a couple of guys from Yale
link |
were starting a new company and I Googled them
link |
and found that they had started a company previously
link |
and sold it actually on eBay for about a quarter million bucks
link |
which was a pretty interesting story.
link |
But so I thought to myself, these guys are, you know,
link |
rock star entrepreneurs, they've done this before,
link |
they must be driving around in Ferraris
link |
because they sold their company.
link |
And, you know, I thought I could learn a lot from them.
link |
So I teamed up with those guys and, you know,
link |
went out during, went out to California during IAP
link |
which is MIT's month off on one way ticket
link |
and basically never went back.
link |
We were having so much fun,
link |
we felt like we were building something and creating something
link |
and it was gonna be interesting that, you know,
link |
I was just all in and got completely hooked
link |
and that business was Justin TV
link |
which is originally a reality show about a guy named Justin
link |
which morphed into a live video streaming platform
link |
which then morphed into what is Twitch today.
link |
So that was quite an unexpected journey.
link |
Looking back, it was just an obvious,
link |
I mean, one way ticket.
link |
I mean, if we just pause on that for a second,
link |
there was no, how did you know these were the right guys?
link |
This is the right decision.
link |
You didn't think it was just follow the heart kind of thing?
link |
Well, I didn't know, but, you know,
link |
just trying something for a month during IAP
link |
seems pretty low risk, right?
link |
And then, you know, well, maybe I'll take a semester off.
link |
MIT's pretty flexible about that.
link |
You can always go back, right?
link |
And then after two or three cycles of that,
link |
I eventually threw in the towel.
link |
But, you know, I think it's, I guess in that case,
link |
I felt like I could always hit the undo button if I had to.
link |
But nevertheless, from when you look in retrospect,
link |
I mean, it seems like a brave decision.
link |
You know, it would be difficult
link |
for a lot of people to make.
link |
It wasn't as popular.
link |
I'd say that the general, you know,
link |
flux of people out of MIT at the time was mostly
link |
into, you know, finance or consulting jobs
link |
in Boston or New York.
link |
And very few people were going to California
link |
to start companies.
link |
But today, I'd say that's probably inverted,
link |
which is just a sign of a sign of the times, I guess.
link |
So there's a story about midnight of March 18, 2007,
link |
where TechCrunch, I guess, announced Justin TV earlier
link |
than it was supposed to a few hours.
link |
The site didn't work.
link |
I don't know if any of this is true, you can tell me.
link |
And you and one of the folks at Justin TV,
link |
Emma Shear, coded through the night.
link |
Can you take me through that experience?
link |
So let me say a few nice things that the article I read quoted
link |
Justin Khan said that you were known for bureau coding
link |
through problems and being a creative genius.
link |
So on that night, what was going through your head?
link |
Or maybe I put another way, how do you
link |
solve these problems?
link |
What's your approach to solving these kind of problems
link |
where the line between success and failure
link |
seems to be pretty thin?
link |
That's a good question.
link |
Well, first of all, that's nice of Justin to say that.
link |
I think I would have been maybe 21 years old then
link |
and not very experienced at programming.
link |
But as with everything in a startup,
link |
you're sort of racing against the clock.
link |
And so our plan was the second we
link |
had this live streaming camera backpack up and running
link |
where Justin could wear it.
link |
And no matter where he went in the city,
link |
it would be streaming live video.
link |
And this is even before the iPhones,
link |
this is like hard to do back then.
link |
And so we thought we were there and the backpack was working.
link |
And then we sent out all the emails
link |
to launch the company and do the press thing.
link |
And then we weren't quite actually there.
link |
And then we thought, oh, well, they're
link |
not going to announce it until maybe 10 AM the next morning.
link |
And it's, I don't know, it's 5 PM now.
link |
So how many hours do we have left?
link |
What is that, like 17 hours to go?
link |
And that was going to be fine.
link |
Was the problem obvious?
link |
Did you understand what could possibly be?
link |
Like how complicated was the system at that point?
link |
It was pretty messy.
link |
So to get a live video feed that looked decent working
link |
from anywhere in San Francisco, I
link |
put together this system where we had like three or four
link |
cell phone data modems.
link |
And they were like, we take the video stream
link |
and sort of spray it across these three or four modems
link |
and then try to catch all the packets on the other side
link |
with unreliable cell phone networks.
link |
Pretty low level networking.
link |
And putting these sort of protocols
link |
on top of all that to reassemble and reorder the packets
link |
and have time buffers and error correction
link |
and all that kind of stuff.
link |
And the night before, it was just
link |
staticky. Every once in a while, the image would go
link |
staticky and there would be this horrible like screeching
link |
audio noise because the audio was also corrupted.
link |
And this would happen like every five to 10 minutes or so.
link |
And it was a really, you know, off of putting to the viewers.
link |
How do you tackle that problem?
link |
What was the, you're just freaking out behind a computer.
link |
There's the word, are there other folks working on this problem?
link |
Like were you behind a whiteboard?
link |
Were you doing a hair coding?
link |
Yeah, it's a little lonely because there's four of us
link |
working on the company and only two people really wrote code.
link |
And Emmett wrote the website in the chat system
link |
and I wrote the software for this video streaming device
link |
And so, you know, it was my sole responsibility
link |
to figure that out.
link |
And I think it's those, you know,
link |
setting deadlines, trying to move quickly and everything
link |
where you're in that moment of intense pressure
link |
that sometimes people do their best and most interesting work.
link |
And so even though that was a terrible moment,
link |
I look back on it fondly because that's like, you know,
link |
that's one of those character defining moments, I think.
link |
So in 2013, October, you founded Cruise Automation.
link |
So progressing forward, another exceptionally successful
link |
company was acquired by GM in 2016 for $1 billion.
link |
But in October 2013, what was on your mind?
link |
What was the plan?
link |
How does one seriously start to tackle
link |
one of the hardest robotics, most important impact
link |
for robotics problems of our age?
link |
After going through Twitch, Twitch was,
link |
and is today pretty successful.
link |
But the work was, the result was entertainment mostly.
link |
Like the better the product was, the more we would entertain
link |
people and then, you know, make money on the ad revenues
link |
And that was a good thing.
link |
It felt good to entertain people.
link |
But I figured like, you know, what is really the point
link |
of becoming a really good engineer
link |
and developing these skills other than, you know,
link |
And I realized I wanted something that scratched
link |
more of an existential itch, like something
link |
that truly matters.
link |
And so I basically made this list of requirements
link |
for a new, if I was going to do another company.
link |
And the one thing I knew in the back of my head
link |
that Twitch took like eight years to become successful.
link |
And so whatever I do, I better be willing to commit,
link |
you know, at least 10 years to something.
link |
And when you think about things from that perspective,
link |
you certainly, I think, raise the bar
link |
on what you choose to work on.
link |
So for me, the three things where
link |
it had to be something where the technology itself
link |
determines the success of the product,
link |
like hard, really juicy technology problems,
link |
because that's what motivates me.
link |
And then it had to have a direct and positive impact
link |
on society in some way.
link |
So an example would be like, you know,
link |
health care, self driving cars because they save lives,
link |
other things where there's a clear connection to somehow
link |
improving other people's lives.
link |
And the last one is it had to be a big business
link |
because for the positive impact to matter,
link |
it's got to be a large scale.
link |
And I was thinking about that for a while
link |
and I made like a, I tried writing a Gmail clone
link |
and looked at some other ideas.
link |
And then it just sort of light bulb went off
link |
like self driving cars.
link |
Like that was the most fun I had ever had
link |
in college working on that.
link |
And like, well, what's the state of the technology
link |
has been 10 years, maybe times have changed
link |
and maybe now is the time to make this work.
link |
And I poked around and looked at the only other thing
link |
out there really at the time was the Google self driving
link |
And I thought surely there's a way to, you know,
link |
have an entrepreneur mindset and sort of solve
link |
the minimum viable product here.
link |
And so I just took the plunge right then and there
link |
and said, this, this is something I know
link |
I can commit 10 years to.
link |
It's probably the greatest applied AI problem
link |
of our generation.
link |
And if it works, it's going to be both a huge business
link |
and therefore like probably the most positive impact
link |
I can possibly have on the world.
link |
So after that light bulb went off,
link |
I went all in on cruise immediately
link |
Did you have an idea how to solve this problem?
link |
Which aspect of the problem to solve?
link |
You know, slow, like we just had Oliver from voyage here
link |
slow moving retirement communities,
link |
urban driving, highway driving.
link |
Did you have like, did you have a vision
link |
of the city of the future or, you know,
link |
the transportation is largely automated,
link |
that kind of thing.
link |
Or was it sort of more fuzzy and gray area than that?
link |
My analysis of the situation is that Google's putting a lot,
link |
had been putting a lot of money into that project.
link |
They had a lot more resources.
link |
And so, and they still hadn't cracked
link |
the fully driverless car.
link |
You know, this is 2013, I guess.
link |
So I thought, what can I do to sort of go from zero
link |
to, you know, significant scale
link |
so I can actually solve the real problem,
link |
which is the driverless cars.
link |
And I thought, here's the strategy.
link |
We'll start by doing a really simple problem
link |
or solving a really simple problem
link |
that creates value for people.
link |
So it eventually ended up deciding
link |
on automating highway driving,
link |
which is relatively more straightforward
link |
as long as there's a backup driver there.
link |
And, you know, the go to market
link |
will be able to retrofit people's cars
link |
and just sell these products directly.
link |
And the idea was, we'll take all the revenue
link |
and profits from that and use it to do the,
link |
to sort of reinvest that in research for doing
link |
fully driverless cars.
link |
And that was the plan.
link |
The only thing that really changed along the way
link |
between then and now is,
link |
we never really launched the first product.
link |
We had enough interest from investors
link |
and enough of a signal that this was something
link |
that we should be working on,
link |
that after about a year of working on the highway autopilot,
link |
we had it working, you know, at a prototype stage,
link |
but we just completely abandoned that
link |
and said, we're gonna go all in on driverless cars
link |
Can't think of anything that's more exciting.
link |
And if it works more impactful,
link |
so we're just gonna go for it.
link |
The idea of retrofit is kind of interesting.
link |
Being able to, it's how you achieve scale.
link |
It's a really interesting idea,
link |
is it's something that's still in the back of your mind
link |
I've come full circle on that one after trying
link |
to build a retrofit product.
link |
And I'll touch on some of the complexities of that.
link |
And then also having been inside an OEM
link |
and seeing how things work
link |
and how a vehicle is developed and validated.
link |
When it comes to something
link |
that has safety critical implications,
link |
like controlling the steering
link |
and other control inputs on your car,
link |
it's pretty hard to get there with a retrofit.
link |
Or if you did, even if you did,
link |
it creates a whole bunch of new complications around
link |
liability or how did you truly validate that?
link |
Or, you know, something in the base vehicle fails
link |
and causes your system to fail, whose fault is it?
link |
Or if the car's anti lock brake systems
link |
or other things kick in or the software has been,
link |
it's different in one version of the car.
link |
You retrofit versus another and you don't know
link |
because the manufacturer has updated it behind the scenes.
link |
There's basically an infinite list of long tail issues
link |
And if you're dealing with a safety critical product,
link |
that's not really acceptable.
link |
That's a really convincing summary of why
link |
it's really challenging.
link |
But I didn't know all that at the time.
link |
So we tried it anyway.
link |
But as a pitch also at the time,
link |
it's a really strong one.
link |
That's how you achieve scale and that's how you beat
link |
the current, the leader at the time of Google
link |
or the only one in the market.
link |
The other big problem we ran into,
link |
which is perhaps the biggest problem
link |
from a business model perspective,
link |
is we had kind of assumed that we started with an Audi S4
link |
as the vehicle we retrofitted
link |
with this highway driving capability.
link |
And we had kind of assumed that if we just knock out
link |
like three make and models of vehicle,
link |
that'll cover like 80% of the San Francisco market.
link |
Doesn't everyone there drive, I don't know,
link |
a BMW or a Honda Civic or one of these three cars?
link |
And then we surveyed our users and we found out
link |
that it's all over the place.
link |
We would, to get even a decent number of units sold,
link |
we'd have to support like 20 or 50 different models.
link |
And each one is a little butterfly that takes time
link |
and effort to maintain that retrofit integration
link |
and custom hardware and all this.
link |
So it was a tough business.
link |
So GM manufactures and sells over nine million cars a year.
link |
And what you with crews are trying to do
link |
some of the most cutting edge innovation
link |
in terms of applying AI.
link |
And so how do those, you've talked about it a little bit
link |
before, but it's also just fascinating to me,
link |
we work a lot of automakers.
link |
The difference between the gap between Detroit
link |
and Silicon Valley, let's say,
link |
just to be sort of poetic about it, I guess.
link |
How do you close that gap?
link |
How do you take GM into the future
link |
where a large part of the fleet would be autonomous perhaps?
link |
I wanna start by acknowledging that GM is made up of
link |
tens of thousands of really brilliant,
link |
motivated people who wanna be a part of the future.
link |
And so it's pretty fun to work with them.
link |
The attitude inside a car company like that
link |
is embracing this transformation and change
link |
rather than fearing it.
link |
And I think that's a testament to the leadership at GM
link |
and that's flown all the way through to everyone
link |
you talk to, even the people in the assembly plants
link |
working on these cars.
link |
So that's really great.
link |
So starting from that position makes it a lot easier.
link |
So then when the people in San Francisco
link |
but cruise interact with the people at GM,
link |
at least we have this common set of values,
link |
which is that we really want this stuff to work
link |
because we think it's important
link |
and we think it's the future.
link |
That's not to say those two cultures don't clash.
link |
They absolutely do.
link |
There's different sort of value systems.
link |
Like in a car company, the thing that gets you promoted
link |
and sort of the reward system is following the processes,
link |
delivering the program on time and on budget.
link |
So any sort of risk taking is discouraged in many ways
link |
because if a program is late
link |
or if you shut down the plant for a day,
link |
you can count the millions of dollars
link |
that burn by pretty quickly.
link |
Whereas I think most Silicon Valley companies
link |
and in cruise and the methodology we were employing,
link |
especially around the time of the acquisition,
link |
the reward structure is about trying to solve
link |
these complex problems in any way, shape or form
link |
or coming up with crazy ideas that 90% of them won't work.
link |
And so meshing that culture
link |
of sort of continuous improvement and experimentation
link |
with one where everything needs to be
link |
rigorously defined up front
link |
so that you never slip a deadline or miss a budget
link |
was a pretty big challenge
link |
and that we're over three years in now
link |
after the acquisition.
link |
And I'd say like the investment we made
link |
in figuring out how to work together successfully
link |
and who should do what
link |
and how we bridge the gaps
link |
between these very different systems
link |
and way of doing engineering work
link |
is now one of our greatest assets
link |
because I think we have this really powerful thing
link |
but for a while it was both GM and cruise
link |
were very steep on the learning curve.
link |
Yeah, so I'm sure it was very stressful.
link |
It's really important work
link |
because that's how to revolutionize the transportation.
link |
Really to revolutionize any system,
link |
you look at the healthcare system
link |
or you look at the legal system.
link |
I have people like Laura's come up to me all the time
link |
like everything they're working on
link |
can easily be automated.
link |
But then that's not a good feeling.
link |
Well, it's not a good feeling,
link |
but also there's no way to automate
link |
because the entire infrastructure is really based
link |
is older and it moves very slowly.
link |
And so how do you close the gap between?
link |
I haven't, how can I replace?
link |
Of course, Laura's the one be replaced with an app
link |
but you could replace a lot of aspect
link |
when most of the data is still on paper.
link |
And so the same thing with automotive.
link |
I mean, it's fundamentally software.
link |
So it's basically hiring software engineers.
link |
It's thinking of software world.
link |
I mean, I'm pretty sure nobody in Silicon Valley
link |
has ever hit a deadline.
link |
So and then on GM.
link |
That's probably true, yeah.
link |
And GM side is probably the opposite.
link |
So that's that culture gap is really fascinating.
link |
So you're optimistic about the future of that.
link |
Yeah, I mean, from what I've seen, it's impressive.
link |
And I think like, especially in Silicon Valley,
link |
it's easy to write off building cars
link |
because people have been doing that
link |
for over a hundred years now in this country.
link |
And so it seems like that's a solved problem,
link |
but that doesn't mean it's an easy problem.
link |
And I think it would be easy to sort of overlook that
link |
and think that we're Silicon Valley engineers,
link |
we can solve any problem, building a car,
link |
it's been done, therefore it's not a real engineering
link |
But after having seen just the sheer scale
link |
and magnitude and industrialization that occurs
link |
inside of an automotive assembly plant,
link |
that is a lot of work that I am very glad
link |
that we don't have to reinvent
link |
to make self driving cars work.
link |
And so to have partners who have done that for a hundred
link |
years and have these great processes
link |
and this huge infrastructure and supply base
link |
that we can tap into is just remarkable
link |
because the scope and surface area of the problem
link |
of deploying fleets of self driving cars is so large
link |
that we're constantly looking for ways to do less
link |
so we can focus on the things that really matter more.
link |
And if we had to figure out how to build and assemble
link |
and test and build the cars themselves,
link |
I mean, we work closely with GM on that,
link |
but if we had to develop all that capability
link |
in house as well, that would just make the problem
link |
really intractable, I think.
link |
So yeah, just like your first entry at the MIT DARPA
link |
challenge when it was what the motor that failed
link |
and somebody that knows what they're doing
link |
with the motor did it.
link |
It would have been nice if we could focus on the software
link |
and not the hardware platform.
link |
So from your perspective now,
link |
there's so many ways that autonomous vehicles
link |
can impact society in the next year, five years, 10 years.
link |
What do you think is the biggest opportunity
link |
to make money in autonomous driving,
link |
sort of make it a financially viable thing in the near term?
link |
What do you think would be the biggest impact there?
link |
Well, the things that drive the economics
link |
for fleets of self driving cars
link |
are there's sort of a handful of variables.
link |
One is the cost to build the vehicle itself.
link |
So the material cost, what's the cost of all your sensors,
link |
plus the cost of the vehicle
link |
and all the other components on it.
link |
Another one is the lifetime of the vehicle.
link |
It's very different if your vehicle drives 100,000 miles
link |
and then it falls apart versus 2 million.
link |
And then if you have a fleet,
link |
it's kind of like an airplane or an airline
link |
where once you produce the vehicle,
link |
you want it to be in operation
link |
as many hours a day as possible producing revenue.
link |
And then the other piece of that
link |
is how are you generating revenue?
link |
I think that's kind of what you're asking in.
link |
I think the obvious things today
link |
are the ride sharing business
link |
because that's pretty clear that there's demand for that.
link |
There's existing markets you can tap into and...
link |
Large urban areas, that kind of thing.
link |
And I think that there are some real benefits
link |
to having cars without drivers
link |
compared to sort of the status quo
link |
for people who use ride share services today.
link |
You know, your privacy, consistency,
link |
hopefully significantly improve safety,
link |
all these benefits versus the current product.
link |
But it's a crowded market.
link |
And then other opportunities
link |
which you've seen a lot of activity in the last,
link |
really in the last six or 12 months is delivery,
link |
whether that's parcels and packages, food or groceries.
link |
Those are all sort of, I think, opportunities
link |
that are pretty ripe for these.
link |
Once you have this core technology,
link |
which is the fleet of autonomous vehicles,
link |
there's all sorts of different business opportunities
link |
you can build on top of that.
link |
But I think the important thing, of course,
link |
is that there's zero monetization opportunity
link |
until you actually have that fleet
link |
of very capable driverless cars
link |
that are as good or better than humans.
link |
And that's sort of where the entire industry
link |
is sort of in this holding pattern right now.
link |
Yeah, they're trying to achieve that baseline.
link |
But you said sort of not reliability consistency.
link |
It's kind of interesting.
link |
I think I heard you say somewhere,
link |
not sure if that's what you meant,
link |
but I can imagine a situation
link |
where you would get an autonomous vehicle.
link |
And when you get into an Uber or Lyft,
link |
you don't get to choose the driver
link |
in a sense that you don't get to choose
link |
the personality of the driving.
link |
Do you think there's room
link |
to define the personality of the car
link |
the way it drives you,
link |
in terms of aggressiveness, for example,
link |
in terms of sort of pushing the boundaries.
link |
One of the biggest challenges in autonomous driving
link |
is the trade off between sort of safety and assertiveness.
link |
And do you think there's any room
link |
for the human to take a role in that decision?
link |
Sort of accept some of the liability, I guess.
link |
I wouldn't say, no, I'd say within reasonable bounds,
link |
as in we're not gonna,
link |
I think it'd be higher than likely
link |
we'd expose any knob that would let you
link |
significantly increase safety risk.
link |
I think that's just not something we'd be willing to do.
link |
But I think driving style or like,
link |
are you gonna relax the comfort constraints slightly
link |
or things like that?
link |
All of those things make sense and are plausible.
link |
I see all those as nice optimizations.
link |
Once again, we get the core problem solved
link |
in these fleets out there.
link |
But the other thing we've sort of observed
link |
is that you have this intuition
link |
that if you sort of slam your foot on the gas
link |
right after the light turns green
link |
and aggressively accelerate,
link |
you're gonna get there faster.
link |
But the actual impact of doing that is pretty small.
link |
You feel like you're getting there faster,
link |
but so the same would be true for AVs.
link |
Even if they don't slam the pedal to the floor
link |
when the light turns green,
link |
they're gonna get you there within,
link |
if it's a 15 minute trip,
link |
within 30 seconds of what you would have done otherwise
link |
if you were going really aggressively.
link |
So I think there's this sort of self deception
link |
that my aggressive driving style is getting me there faster.
link |
Well, so that's, you know, some of the things I study,
link |
some of the things I'm fascinated by the psychology of that.
link |
And I don't think it matters
link |
that it doesn't get you there faster.
link |
It's the emotional release.
link |
Driving is a place, being inside our car,
link |
somebody said it's like the real world version
link |
So you have this protection, this mental protection,
link |
and you're able to sort of yell at the world,
link |
like release your anger, whatever it is.
link |
But so there's an element of that
link |
that I think autonomous vehicles
link |
would also have to, you know, giving an outlet to people,
link |
but it doesn't have to be through driving or honking
link |
or so on, there might be other outlets.
link |
But I think to just sort of even just put that aside,
link |
the baseline is really, you know, that's the focus,
link |
that's the thing you need to solve,
link |
and then the fun human things can be solved after.
link |
But so from the baseline of just solving autonomous driving,
link |
you're working in San Francisco,
link |
one of the more difficult cities to operate in,
link |
what is the, in your view currently,
link |
the hardest aspect of autonomous driving?
link |
Negotiating with pedestrians,
link |
is it edge cases of perception?
link |
Is there a mechanical engineering?
link |
Is it data, fleet stuff?
link |
What are your thoughts on the more challenging aspects there?
link |
That's a good question.
link |
I think before we go to that though,
link |
I just want to, I like what you said
link |
about the psychology aspect of this,
link |
because I think one observation I've made is,
link |
I think I read somewhere that I think it's,
link |
maybe Americans on average spend, you know,
link |
over an hour a day on social media,
link |
like staring at Facebook.
link |
And so that's just, you know,
link |
60 minutes of your life, you're not getting back.
link |
It's probably not super productive.
link |
And so that's 3,600 seconds, right?
link |
And that's, that's time, you know,
link |
it's a lot of time you're giving up.
link |
And if you compare that to people being on the road,
link |
if another vehicle,
link |
whether it's a human driver or autonomous vehicle,
link |
delays them by even three seconds,
link |
they're laying in on the horn, you know,
link |
even though that's, that's, you know,
link |
one 1,000th of the time they waste
link |
looking at Facebook every day.
link |
So there's, there's definitely some,
link |
you know, psychology aspects of this,
link |
I think that are pretty interesting.
link |
Road rage in general.
link |
And then the question, of course,
link |
is if everyone is in self driving cars,
link |
do they even notice these three second delays anymore?
link |
Because they're doing other things
link |
or reading or working or just talking to each other.
link |
So it'll be interesting to see where that goes.
link |
In a certain aspect, people,
link |
people need to be distracted
link |
by something entertaining,
link |
something useful inside the car
link |
so they don't pay attention to the external world.
link |
And then, and then they can take whatever psychology
link |
and bring it back to Twitter and then focus on that
link |
as opposed to sort of interacting,
link |
sort of putting the emotion out there into the world.
link |
So it's an interesting problem,
link |
but baseline autonomy.
link |
I guess you could say self driving cars,
link |
you know, at scale will lower the collective blood pressure
link |
of society probably by a couple of points
link |
without all that road rage and stress.
link |
So that's a good, good externality.
link |
So back to your question about the technology
link |
and the, I guess the biggest problems.
link |
And I have a hard time answering that question
link |
because, you know, we've been at this,
link |
like specifically focusing on driverless cars
link |
and all the technology needed to enable that
link |
for a little over four and a half years now.
link |
And even a year or two in,
link |
I felt like we had completed the functionality needed
link |
to get someone from point A to point B.
link |
As in, if we need to do a left turn maneuver
link |
or if we need to drive around a, you know,
link |
a double parked vehicle into oncoming traffic
link |
or navigate through construction zones,
link |
the scaffolding and the building blocks
link |
was there pretty early on.
link |
And so the challenge is not any one scenario or situation
link |
for which, you know, we fail at 100% of those.
link |
It's more, you know, we're benchmarking against a pretty good
link |
or pretty high standard, which is human driving.
link |
All things considered, humans are excellent
link |
at handling edge cases and unexpected scenarios
link |
where it's computers are the opposite.
link |
And so beating that baseline set by humans is the challenge.
link |
And so what we've been doing for quite some time now
link |
is basically it's this continuous improvement process
link |
where we find sort of the most, you know, uncomfortable
link |
or the things that could lead to a safety issue
link |
or other things, all these events.
link |
And then we sort of categorize them
link |
and rework parts of our system
link |
to make incremental improvements
link |
and do that over and over and over again.
link |
And we just see sort of the overall performance
link |
of the system, you know,
link |
actually increasing in a pretty steady clip.
link |
But there's no one thing.
link |
There's actually like thousands of little things
link |
and just like polishing functionality
link |
and making sure that it handles, you know,
link |
every version and possible permutation of a situation
link |
by either applying more deep learning systems
link |
or just by, you know, adding more test coverage
link |
or new scenarios that we develop against
link |
and just grinding on that.
link |
We're sort of in the unsexy phase of development right now
link |
which is doing the real engineering work
link |
that it takes to go from prototype to production.
link |
You're basically scaling the grinding.
link |
So sort of taking seriously the process
link |
of all those edge cases, both with human experts
link |
and machine learning methods to cover,
link |
to cover all those situations.
link |
Yeah, and the exciting thing for me is
link |
I don't think that grinding ever stops
link |
because there's a moment in time
link |
where you've crossed that threshold of human performance
link |
and become superhuman.
link |
But there's no reason, there's no first principles reason
link |
that AV capability will tap out anywhere near humans.
link |
Like there's no reason it couldn't be 20 times better
link |
whether that's, you know, just better driving
link |
or safer driving or more comfortable driving
link |
or even a thousand times better given enough time.
link |
And we intend to basically chase that, you know, forever
link |
to build the best possible product.
link |
Better and better and better
link |
and always new edge cases come up and new experiences.
link |
So, and you want to automate that process
link |
as much as possible.
link |
So what do you think in general in society
link |
when do you think we may have hundreds of thousands
link |
of fully autonomous vehicles driving around?
link |
So first of all, predictions, nobody knows the future.
link |
You're a part of the leading people
link |
trying to define that future,
link |
but even then you still don't know.
link |
But if you think about hundreds of thousands of vehicles,
link |
so a significant fraction of vehicles
link |
in major cities are autonomous.
link |
Do you think, are you with Rodney Brooks
link |
who is 2050 and beyond?
link |
Or are you more with Elon Musk
link |
who is, we should have had that two years ago?
link |
Well, I mean, I'd love to have it two years ago,
link |
but we're not there yet.
link |
So I guess the way I would think about that
link |
is let's flip that question around.
link |
So what would prevent you to reach hundreds
link |
of thousands of vehicles and...
link |
That's a good rephrasing.
link |
Yeah, so the, I'd say that it seems the consensus
link |
among the people developing self driving cars today
link |
is to sort of start with some form of an easier environment,
link |
whether it means lacking, inclement weather,
link |
or mostly sunny or whatever it is.
link |
And then add capability for more complex situations
link |
And so if you're only able to deploy in areas
link |
that meet sort of your criteria
link |
or that the current don't meet,
link |
operating domain of the software you developed,
link |
that may put a cap on how many cities you could deploy in.
link |
But then as those restrictions start to fall away,
link |
like maybe you add capability to drive really well
link |
and safely and have you rain or snow,
link |
that probably opens up the market by two or three fold
link |
in terms of the cities you can expand into and so on.
link |
And so the real question is,
link |
I know today if we wanted to,
link |
we could produce that many autonomous vehicles,
link |
but we wouldn't be able to make use of all of them yet
link |
because we would sort of saturate the demand in the cities
link |
in which we would want to operate initially.
link |
So if I were to guess what the timeline is
link |
for those things falling away
link |
and reaching hundreds, thousands of vehicles.
link |
Maybe a range is better.
link |
I would say less than five years.
link |
Less than five years.
link |
And of course you're working hard to make that happen.
link |
So you started two companies that were eventually acquired
link |
for each $4 billion.
link |
So you're a pretty good person to ask,
link |
what does it take to build a successful startup?
link |
I think there's sort of survivor bias here a little bit,
link |
but I can try to find some common threads
link |
for the things that worked for me, which is...
link |
In both of these companies,
link |
I was really passionate about the core technology.
link |
I actually lay awake at night thinking about these problems
link |
and how to solve them.
link |
And I think that's helpful because when you start a business,
link |
To this day, there are these crazy ups and downs.
link |
One day you think the business is just on top of the world
link |
and unstoppable and the next day you think,
link |
okay, this is all going to end.
link |
It's just going south and it's going to be over tomorrow.
link |
And so I think having a true passion that you can fall back on
link |
and knowing that you would be doing it
link |
even if you weren't getting paid for it
link |
helps you weather those tough times.
link |
So that's one thing.
link |
I think the other one is really good people.
link |
So I've always been surrounded by really good cofounders
link |
that are logical thinkers,
link |
are always pushing their limits
link |
and have very high levels of integrity.
link |
So that's Dan Kahn in my current company
link |
and actually his brother and a couple other guys
link |
for Justin TV and Twitch.
link |
And then I think the last thing is just,
link |
I guess, persistence or perseverance.
link |
And that can apply to sticking to
link |
having conviction around the original premise of your idea
link |
and sticking around to do all the unsexy work
link |
to actually make it come to fruition,
link |
including dealing with whatever it is
link |
that you're not passionate about,
link |
whether that's finance or HR or operations or those things.
link |
As long as you are grinding away
link |
and working towards that North Star for your business,
link |
whatever it is and you don't give up
link |
and you're making progress every day,
link |
it seems like eventually you'll end up in a good place.
link |
And the only things that can slow you down
link |
are running out of money
link |
or I suppose your competitor is destroying you,
link |
but I think most of the time it's people giving up
link |
or somehow destroying things themselves
link |
rather than being beaten by their competition
link |
or running out of money.
link |
Yeah, if you never quit, eventually you'll arrive.
link |
It's a much more concise version
link |
of what I was trying to say.
link |
So you went the Y Combinator out twice.
link |
What do you think, in a quick question,
link |
do you think is the best way to raise funds
link |
in the early days?
link |
Or not just funds, but just community,
link |
develop your idea and so on.
link |
Can you do it solo or maybe with a cofounder
link |
Do you think Y Combinator is good?
link |
Is it good to do VC route?
link |
Is there no right answer or is there,
link |
from the Y Combinator experience,
link |
something that you could take away
link |
that that was the right path to take?
link |
There's no one size fits all answer,
link |
but if your ambition I think is to see how big
link |
you can make something or rapidly expand
link |
and capture a market or solve a problem
link |
or whatever it is, then going the venture
link |
back route is probably a good approach
link |
so that capital doesn't become your primary constraint.
link |
Y Combinator, I love because it puts you
link |
in this sort of competitive environment
link |
where you're surrounded by the top,
link |
maybe 1% of other really highly motivated
link |
peers who are in the same place.
link |
In that environment I think just breeds success.
link |
If you're surrounded by really brilliant
link |
hardworking people, you're going to feel
link |
sort of compelled or inspired to try
link |
to emulate them or beat them.
link |
So even though I had done it once before
link |
and I felt like I'm pretty self motivated,
link |
I thought this is going to be a hard problem,
link |
I can use all the help I can get.
link |
So surrounding myself with other entrepreneurs
link |
is going to make me work a little bit harder
link |
or push a little harder then it's worth it.
link |
That's why I did it, for example, the second time.
link |
Let's go full soft, go existential.
link |
If you go back and do something differently in your life,
link |
starting in high school and MIT, leaving MIT,
link |
you could have gone to the PhD route,
link |
doing startup, going to see about a startup in California
link |
or maybe some aspects of fundraising.
link |
Is there something you regret,
link |
not necessarily regret, but if you go back,
link |
you could do differently?
link |
I think I've made a lot of mistakes,
link |
pretty much everything you can screw up,
link |
I think I've screwed up at least once.
link |
But I don't regret those things.
link |
I think it's hard to look back on things,
link |
even if they didn't go well and call it a regret,
link |
because hopefully it took away some new knowledge
link |
or learning from that.
link |
I would say there's a period,
link |
the closest I can come to this,
link |
there's a period in just in TV,
link |
I think after seven years where the company was going
link |
one direction, which is towards Twitch and video gaming.
link |
I'm not a video gamer.
link |
I don't really even use Twitch at all.
link |
I was still working on the core technology there,
link |
but my heart was no longer in it,
link |
because the business that we were creating
link |
was not something that I was personally passionate about.
link |
It didn't meet your bar of existential impact.
link |
Yeah, and I'd say I probably spent an extra year or two
link |
working on that, and I'd say I would have just tried
link |
to do something different sooner.
link |
Because those were two years where I felt like,
link |
from this philosophical or existential thing,
link |
I just felt that something was missing.
link |
If I could look back now and tell myself,
link |
I would have said exactly that.
link |
You're not getting any meaning out of your work personally
link |
You should find a way to change that.
link |
And that's part of the pitch I used
link |
to basically everyone who joins Cruise today.
link |
It's like, hey, you've got that now by coming here.
link |
Well, maybe you needed the two years of that existential dread
link |
to develop the feeling that ultimately
link |
it was the fire that created Cruise.
link |
So you never know.
link |
Good theory, yeah.
link |
What does 2019 hold for Cruise?
link |
After this, I guess we're going to go and talk to your class.
link |
But one of the big things is going from prototype to production
link |
for autonomous cars.
link |
And what does that mean?
link |
What does that look like?
link |
2019 for us is the year that we try to cross over
link |
that threshold and reach superhuman level of performance
link |
to some degree with the software and have all the other
link |
of the thousands of little building blocks in place
link |
to launch our first commercial product.
link |
So that's what's in store for us.
link |
And we've got a lot of work to do.
link |
We've got a lot of brilliant people working on it.
link |
So it's all up to us now.
link |
So Charlie Miller and Chris Vell is like the people I've
link |
crossed paths with.
link |
It sounds like you have an amazing team.
link |
So like I said, it's one of the most, I think, one of the most
link |
important problems in artificial intelligence of this century.
link |
It'll be one of the most defining.
link |
It's super exciting that you work on it.
link |
And the best of luck in 2019.
link |
I'm really excited to see what Cruise comes up with.
link |
Thanks for having me today.