back to indexWojciech Zaremba: OpenAI Codex, GPT-3, Robotics, and the Future of AI | Lex Fridman Podcast #215
link |
The following is a conversation with Wojciech Zaremba, cofounder of OpenAI,
link |
which is one of the top organizations in the world doing artificial intelligence
link |
research and development.
link |
Wojciech is the head of language and cogeneration teams, building and doing
link |
research on GitHub Copilot, OpenAI Codex, and GPT 3, and who knows, 4, 5, 6,
link |
and, and, and plus one, and he also previously led OpenAI's robotic efforts.
link |
These are incredibly exciting projects to me that deeply challenge and expand
link |
our understanding of the structure and nature of intelligence.
link |
The 21st century, I think, may very well be remembered for a handful of
link |
revolutionary AI systems and their implementations.
link |
GPT, Codex, and applications of language models and transformers in general
link |
to the language and visual domains may very well be at the core of these AI
link |
systems. To support this podcast, please check out our sponsors.
link |
They're listed in the description.
link |
This is the Lex Friedman podcast, and here is my conversation
link |
with Wojciech Zaremba.
link |
You mentioned that Sam Altman asked about the Fermi Paradox, and the people
link |
at OpenAI had really sophisticated, interesting answers, so that's when you
link |
knew this is the right team to be working with.
link |
So let me ask you about the Fermi Paradox, about aliens.
link |
Why have we not found overwhelming evidence for aliens visiting Earth?
link |
I don't have a conviction in the answer, but rather kind of probabilistic
link |
perspective on what might be, let's say, possible answers.
link |
It's also interesting that the question itself even can touch on the, you
link |
know, your typical question of what's the meaning of life, because if you
link |
assume that, like, we don't see aliens because they destroy themselves, that
link |
kind of upweights the focus on making sure that we won't destroy ourselves.
link |
At the moment, the place where I am actually with my belief, and these
link |
things also change over the time, is I think that we might be alone in
link |
the universe, which actually makes life more, or let's say, consciousness
link |
life, more kind of valuable, and that means that we should more appreciate it.
link |
Have we always been alone?
link |
So what's your intuition about our galaxy, our universe?
link |
Is it just sprinkled with graveyards of intelligent civilizations, or are
link |
we truly, is life, intelligent life, truly unique?
link |
At the moment, my belief that it is unique, but I would say I could also,
link |
you know, there was like some footage released with UFO objects, which makes
link |
me actually doubt my own belief.
link |
Yeah, I can tell you one crazy answer that I have heard.
link |
So, apparently, when you look actually at the limits of computation, you
link |
can compute more if the temperature of the universe would drop.
link |
Temperature of the universe would drop down.
link |
So one of the things that aliens might want to do if they are truly optimizing
link |
to maximize amount of compute, which, you know, maybe can lead to, or let's
link |
say simulations or so, it's instead of wasting current entropy of the
link |
universe, because, you know, we, by living, we are actually somewhat
link |
wasting entropy, then you can wait for the universe to cool down such that
link |
you have more computation.
link |
So that's kind of a funny answer.
link |
I'm not sure if I believe in it, but that would be one of the
link |
reasons why you don't see aliens.
link |
It's also possible to some people say that maybe there is not that much
link |
point in actually going to other galaxies if you can go inwards.
link |
So there is no limits of what could be an experience if we could, you
link |
know, connect machines to our brains while there are still some limits
link |
if we want to explore the universe.
link |
Yeah, there could be a lot of ways to go inwards too.
link |
Once you figure out some aspect of physics, we haven't figured out yet.
link |
Maybe you can travel to different dimensions.
link |
I mean, travel in three dimensional space may not be the most fun kind of travel.
link |
There may be like just a huge amount of different ways to travel and it
link |
doesn't require a spaceship going slowly in 3d space to space time.
link |
It also feels, you know, one of the problems is that speed of light
link |
is low and the universe is vast.
link |
And it seems that actually most likely if we want to travel very far, then
link |
we would, instead of actually sending spaceships with humans that weight a
link |
lot, we would send something similar to what Yuri Miller is working on.
link |
These are like a huge sail, which is at first powered or there is a shot of
link |
laser from an air and it can propel it to quarter of speed of light and sail
link |
itself contains a few grams of equipment.
link |
And that might be the way to actually transport matter through universe.
link |
But then when you think what would it mean for humans, it means that we would
link |
need to actually put their 3d printer and, you know, 3d print the human on
link |
other planet, I don't know, play them YouTube or let's say, or like a 3d
link |
print like huge human right away, or maybe a womb or so, um, yeah.
link |
With our current techniques of archeology, if, if, if a civilization
link |
was born and died, uh, long, long enough ago on earth, we wouldn't be able to
link |
tell, and so that makes me really sad.
link |
And so I think about earth in that same way.
link |
How can we leave some remnants if we do destroy ourselves?
link |
How can we leave remnants for aliens in the future to discover?
link |
Like, here's some nice stuff we've done, like Wikipedia and YouTube.
link |
Do we have it like in a satellite orbiting earth with a hard drive?
link |
Like, how, how do we say, how do we back up human civilization?
link |
Uh, the good parts or all of it is good parts so that, uh, it can be
link |
preserved longer than our bodies can.
link |
That's a, that's kind of, um, it's a difficult question.
link |
It also requires the difficult acceptance of the fact that we may die.
link |
And if we die, we may die suddenly as a civilization.
link |
So let's see, I think it kind of depends on the cataclysm.
link |
We have observed in other parts of the universe that birds of gamma rays, uh,
link |
these are, uh, high energy, uh, rays of light that actually can
link |
apparently kill entire galaxy.
link |
So there might be actually nothing, even to, nothing to protect us from it.
link |
I'm also, and I'm looking actually at the past civilizations.
link |
So it's like Aztecs or so they disappear from the surface of the earth.
link |
And one can ask, why is it the case?
link |
And the way I'm thinking about it is, you know, that definitely they had some
link |
problem that they couldn't solve and maybe there was a flood and all of a
link |
sudden they couldn't drink, uh, there was no potable water and they all died.
link |
And, um, I think that, uh, so far the best solution to such a problems is I
link |
guess, technology, so, I mean, if they would know that you can just boil
link |
water and then drink it after, then that would save their civilization.
link |
And even now, when we look actually at the current pandemic, it seems
link |
that there, once again, actually science comes to rest.
link |
And somehow science increases size of the action space.
link |
And I think that's a good thing.
link |
But nature has a vastly larger action space, but still it might be a good thing
link |
for us to keep on increasing action space.
link |
Uh, looking at past civilizations.
link |
But looking at the destruction of human civilization, perhaps expanding the
link |
action space will add, um, actions that are easily acted upon, easily executed
link |
and as a result, destroy us.
link |
So let's see, I was pondering, uh, why actually even, uh, we have
link |
negative impact on the, uh, globe.
link |
Because, you know, if you ask every single individual, they
link |
would like to have clean air.
link |
They would like healthy planet, but somehow it's not.
link |
It's not the case that as a collective, we are not going in this direction.
link |
I think that there exists very powerful system to describe what we value.
link |
That's capitalism.
link |
It assigns actually monetary values to various activities.
link |
At the moment, the problem in the current system is that there's
link |
some things which we value.
link |
There is no cost assigned to it.
link |
So even though we value clean air, or maybe we also, uh, value, uh,
link |
value lack of destruction on, let's say internet or so at the moment, these
link |
quantities, you know, companies, corporations can pollute them, uh, for free.
link |
So in some sense, I wished or like, and that's, I guess, purpose of politics
link |
to, to align the incentive systems.
link |
And we are kind of maybe even moving in this direction.
link |
The first issue is even to be able to measure the things that we value.
link |
Then we can actually assign the monetary value to them.
link |
And that's, so it's getting the data and also probably through technology,
link |
enabling people to vote and to move money around in a way that is aligned
link |
with their values, and that's very much a technology question.
link |
So like having one president and Congress and voting that happens every four years
link |
or something like that, that's a very outdated idea that could be some
link |
technological improvements to that kind of idea.
link |
So I'm thinking from time to time about these topics, but it's also feels to me
link |
that it's, it's a little bit like, uh, it's hard for me to actually make
link |
correct predictions.
link |
What is the appropriate thing to do?
link |
I extremely trust, uh, Sam Altman, our CEO on these topics here, um, like, uh,
link |
I'm more on the side of being, I guess, naive hippie.
link |
That, uh, yeah, that's your life philosophy.
link |
Um, well, like I think self doubt and, uh, I think hippie implies optimism.
link |
Those, those two things are pretty, pretty good way to operate.
link |
I mean, still, it is hard for me to actually understand how the politics
link |
works or like, uh, how this, like, uh, exactly how the things would play out.
link |
And Sam is, uh, really excellent with it.
link |
What do you think is rarest in the universe?
link |
You said we might be alone.
link |
What's hardest to build is another engineering way to ask that life,
link |
intelligence or consciousness.
link |
So like you said that we might be alone, which is the thing that's hardest to get
link |
to, is it just the origin of life?
link |
Is it the origin of intelligence?
link |
Is it the origin of consciousness?
link |
So, um, let me at first explain to you my kind of mental model, what I think
link |
is needed for life to appear.
link |
Um, so I imagine that at some point there was this primordial, uh, soup of, uh,
link |
amino acids and maybe some proteins in the ocean and, uh, you know, some
link |
proteins were turning into some other proteins through reaction and, uh, you
link |
can also, uh, you know, you can, you know, you can, you know, you can
link |
and, uh, you can almost think about this, uh, cycle of what, uh, turns into what
link |
as there is a graph essentially describing which substance turns into
link |
some other substance and essentially life means that all of a sudden in the graph
link |
has been created that cycle such that the same thing keeps on happening over
link |
and over again, that's what is needed for life to happen.
link |
And in some sense, you can think almost that you have this gigantic graph and it
link |
needs like a sufficient number of edges for the cycle to appear.
link |
Um, then, um, from perspective of intelligence and consciousness, uh, my
link |
current intuition is that they might be quite intertwined.
link |
First of all, it might not be that it's like a binary thing that you
link |
have intelligence or consciousness.
link |
It seems to be, uh, uh, more, uh, continuous component.
link |
Let's see, if we look for instance on the event networks, uh, recognizing
link |
images and people are able to show that the activations of these networks
link |
correlate very strongly, uh, with activations in visual cortex, uh, of
link |
some monkeys, the same seems to be true about language models.
link |
Um, also if you, for instance, um, look, um, if you train agent in, um, 3d
link |
world, um, at first, you know, it, it, it, it barely recognizes what is going
link |
on over the time, it kind of recognizes foreground from a background over the
link |
time, it kind of knows where there is a foot, uh, and it just follows it.
link |
Um, over the time it actually starts having a 3d perception.
link |
So it is possible for instance, to look inside of the head of an agent and ask,
link |
what would it see if it looks to the right?
link |
And the crazy thing is, you know, initially when the agents are barely
link |
trained, that these predictions are pretty bad over the time they become
link |
better and better, you can still see that if you ask what happens when the
link |
head is turned by 360 degrees for some time, they think that the different
link |
thing appears and then at some stage they understand actually that the same
link |
thing supposed to appear.
link |
So they get that understanding of 3d structure.
link |
It's also, you know, very likely that they have inside some level of, of like
link |
a symbolic reasoning, like a particular, these symbols for other agents.
link |
So when you look at DOTA agents, they collaborate together and, uh, and, uh,
link |
no, they, they, they, they have some anticipation of, uh, if, if they would
link |
win battle, they have some, some expectations with respect to other
link |
I might be, you know, too much anthropomorphizing, um, the, the, the,
link |
how the things look, look, look for me, but then the fact that they have a
link |
symbol for other agents, uh, makes me believe that, uh, at some stage as the,
link |
uh, you know, as they are optimizing for skills, they would have also symbol to
link |
describe themselves.
link |
Uh, this is like a very useful symbol to have.
link |
And this particularity, I would call it like a self consciousness or self
link |
awareness, uh, and, uh, still it might be different from the consciousness.
link |
So I guess the, the way how I'm understanding the word consciousness,
link |
I'd say the experience of drinking a coffee or let's say experience of being
link |
a bat, that's the meaning of the word consciousness.
link |
It doesn't mean to be awake.
link |
Uh, yeah, it feels, it might be also somewhat related to memory and
link |
recurrent connections.
link |
So, um, it's kind of like, if you look at anesthetic drugs, they might be, uh,
link |
uh, like, uh, that they essentially, they, they disturb, uh, uh, brainwaves, uh, such
link |
that, um, maybe memories, not, not form.
link |
And so there's a lessening of consciousness when you do that.
link |
And so that's the one way to intuit what is consciousness.
link |
There's also kind of another element here.
link |
It could be that it's, you know, this kind of self awareness
link |
module that you described, plus the actual subjective experience is a
link |
storytelling module that tells us a story about, uh, what we're experiencing.
link |
So let's say, I mean, in meditation, they teach people not to speak
link |
story inside of their head.
link |
And there is also some fraction of population who doesn't have actually
link |
a narrator, I know people who don't have a narrator and, you know, they have
link |
to use external people in order to, um, kind of solve tasks that
link |
require internal narrator.
link |
Um, so it seems that it's possible to have the experience without the talk.
link |
What are we talking about when we talk about the internal narrator?
link |
Is that the voice when you're like, yeah, I thought that that's what you are
link |
referring to while I was referring more on the, like, not an actual voice.
link |
I meant like, there's some kind of like subjective experience feels like it's.
link |
It's fundamentally about storytelling to ourselves.
link |
It feels like, like the feeling is a story that is much, uh, much
link |
simpler abstraction than the raw sensory information.
link |
So there feels like it's a very high level of abstraction that, uh, is useful
link |
for me to feel like entity in this world.
link |
M most useful aspect of it is that because I'm conscious, I think there's
link |
an intricate connection to me, not wanting to die.
link |
So like, it's a useful hack to really prioritize not dying, like those
link |
seem to be somehow connected.
link |
So I'm telling the story of like, it's rich.
link |
He feels like something to be me and the fact that me exists in this world.
link |
I want to preserve me.
link |
And so that makes it a useful agent hack.
link |
So I will just refer maybe to that first part, as you said, about that kind
link |
of story of describing who you are.
link |
Um, I was, uh, thinking about that even, so, you know, obviously I'm, I, I like
link |
thinking about consciousness, uh, I like thinking about AI as well, and I'm trying
link |
to see analogies of these things in AI, what would it correspond to?
link |
So, um, um, you know, open AI train, uh, uh, a model called GPT, uh, which, uh,
link |
can generate, uh, pretty, I'm using texts on arbitrary topic and, um, um, and one
link |
way to control GPT is, uh, by putting into prefix at the beginning of the text, some
link |
information, what would be the story about, uh, you can have even chat with, uh, uh,
link |
you know, with GPT by saying that the chat is with Lex or Elon Musk or so, and, uh,
link |
GPT would just pretend to be you or Elon Musk or so, and, uh, uh, it almost feels
link |
that this, uh, story that we give ourselves to describe our life, it's almost like, uh,
link |
things that you put into context of GPT.
link |
The primary, it's the, and so, but the context we provide to GPT is, uh, is multimodal.
link |
It's more so GPT itself is multimodal.
link |
GPT itself, uh, hasn't learned actually from experience of single human, but from the
link |
experience of humanity, it's a chameleon.
link |
You can turn it into anything and in some sense, by providing context, um, it, you
link |
know, behaves as the thing that you wanted it to be.
link |
Um, it's interesting that the, you know, people have a stories of who they are.
link |
And, uh, as you said, these stories, they help them to operate in the world.
link |
Um, but it's also, you know, interesting, I guess, various people find it out through
link |
meditation or so that, uh, there might be some patterns that you have learned when
link |
you were a kid that actually are not serving you anymore.
link |
And you also might be thinking that that's who you are and that's actually just a story.
link |
So it's a useful hack, but sometimes it gets us into trouble.
link |
It's a local optima.
link |
It's a local optima.
link |
You wrote that Stephen Hawking, he tweeted, Stephen Hawking asked what
link |
breathes fire into equations, which meant what makes given mathematical
link |
equations realize the physics of a universe.
link |
Similarly, I wonder what breathes fire into computation.
link |
What makes given computation conscious?
link |
So how do we engineer consciousness?
link |
How do you breathe fire and magic?
link |
How do you breathe fire and magic into the machine?
link |
So, um, it seems clear to me that not every computation is conscious.
link |
I mean, you can, let's say, just keep on multiplying one matrix over and over
link |
again and might be gigantic matrix.
link |
You can put a lot of computation.
link |
I don't think it would be conscious.
link |
So in some sense, the question is, uh, what are the computations which could be
link |
conscious, uh, I mean, so, so one assumption is that it has to do purely
link |
with computation that you can abstract away matter and other possibilities
link |
that it's very important was the realization of computation that it has
link |
to do with some, uh, uh, force fields or so, and they bring consciousness.
link |
At the moment, my intuition is that it can be fully abstracted away.
link |
So in case of computation, you can ask yourself, what are the mathematical
link |
objects or so that could bring such a properties?
link |
So for instance, if we think about the models, uh, AI models, the, what they
link |
truly try to do, uh, or like a models like GPT is, uh, uh, you know, they try
link |
to predict, uh, next word or so.
link |
And this turns out to be equivalent to, uh, compressing, uh, text.
link |
Um, and, uh, because in some sense, compression means that, uh, you learn
link |
the model of reality and you have just to, uh, remember where are your mistakes.
link |
The better you are in predicting the, and, and, and in some sense, when we
link |
look at our experience, also, when you look, for instance, at the car driving,
link |
you know, in which direction it will go, you are good like in prediction.
link |
And, um, you know, it might be the case that the consciousness is intertwined
link |
with, uh, compression, it might be also the case that self consciousness, uh,
link |
has to do with compress or trying to compress itself.
link |
I was just wondering, what are the objects in, you know, mathematics or
link |
computer science, which are mysterious that could, uh, that, that, that could
link |
have to do with consciousness.
link |
And then I thought, um, you know, you, you see in mathematics, there is
link |
something called Gadel theorem, uh, which means, okay, you have, if you have
link |
sufficiently complicated mathematical system, it is possible to point the
link |
mathematical system back on itself.
link |
In computer science, there is, uh, something called helping problem.
link |
It's, it's somewhat similar construction.
link |
So I thought that, you know, if we believe that, uh, that, uh, that under
link |
assumption that consciousness has to do with, uh, with compression, uh, then
link |
you could imagine that the, that the, as you keep on compressing things, then
link |
at some point, it actually makes sense for the compressor to compress itself.
link |
Metacompression consciousness is metacompression.
link |
That's a, that's an I, an, an, an idea.
link |
And in some sense, you know, the crazy, thank you.
link |
So, uh, but do you think if we think of a Turing machine, a universal
link |
Turing machine, can that achieve consciousness?
link |
So is there some thing beyond our traditional definition
link |
of computation that's required?
link |
So it's a specific computation.
link |
And I said, this computation has to do with compression and, uh, the compression
link |
itself, maybe other way of putting it is like, uh, you are internally creating
link |
the model of reality in order, like, uh, it's like a, you try inside to simplify
link |
reality in order to predict what's going to happen.
link |
And, um, that also feels somewhat similar to how I think actually about my own
link |
conscious experience, though clearly I don't have access to reality.
link |
The only access to reality is through, you know, cable going to my brain and my
link |
brain is creating a simulation of reality and I have access to the simulation of
link |
Are you by any chance, uh, aware of, uh, the Hutter prize, Marcus Hutter?
link |
He, uh, he made this prize for compression.
link |
Uh, Wikipedia pages, and, uh, there's a few qualities to it.
link |
One, I think has to be perfect compression, which makes, I think that
link |
little cork makes it much less, um, applicable to the general task of
link |
intelligence, because it feels like intelligence is always going to be messy.
link |
Uh, like perfect compression is feels like it's not the right goal, but
link |
it's nevertheless a very interesting goal.
link |
So for him, intelligence equals compression.
link |
And so the smaller you make the file, given a large Wikipedia page, the
link |
more intelligent the system has to be.
link |
Yeah, that makes sense.
link |
So you can make perfect compression if you store errors.
link |
And I think that actually what he meant is you have algorithm plus errors.
link |
Uh, by the way, Hutter, Hutter is, uh, he was a PhD advisor of Sean
link |
Leck, who is a DeepMind, uh, uh, DeepMind cofounder.
link |
So there's an interesting, uh, and now he's a DeepMind, there's an
link |
interesting, uh, network of people.
link |
And he's one of the people that I think seriously took on the task of
link |
what would an AGI system look like?
link |
Uh, I think for a longest time, the question of AGI was not taken
link |
seriously or rather rigorously.
link |
And he did just that, like mathematically speaking, what
link |
would the model look like if you remove the constraints of it, having to be,
link |
uh, um, having to have a reasonable amount of memory, reasonable amount
link |
of, uh, running time, complexity, uh, computation time, what would it look
link |
like and essentially it's, it's a half math, half philosophical discussion
link |
of, uh, how would it like a reinforcement learning type of
link |
framework look like for an AGI?
link |
So he developed the framework even to describe what's optimal with
link |
respect to reinforcement learning.
link |
Like there is a theoretical framework, which is, as you said, under assumption,
link |
there is infinite amount of memory and compute.
link |
Um, there was actually one person before his name is Solomonov, who
link |
there extended, uh, Solomonov work to reinforcement learning, but there
link |
exists the, uh, theoretical algorithm, which is optimal algorithm to build
link |
intelligence and I can actually explain you the algorithm.
link |
So the task itself, can I just pause how absurd it is for brain in a
link |
skull, trying to explain the algorithm for intelligence, just go ahead.
link |
It is pretty crazy.
link |
It is pretty crazy that, you know, the brain itself is actually so
link |
small and it can ponder, uh, how to design algorithms that optimally
link |
solve the problem of intelligence.
link |
So what's the algorithm?
link |
So first of all, the task itself is, uh, described as, uh, you have infinite
link |
sequence of zeros and ones.
link |
Okay. You read, uh, N bits and they are about to predict N plus one bit.
link |
So that's the task.
link |
And you could imagine that every task could be casted as such a task.
link |
So if for instance, you have images and labels, you can just turn every image
link |
into a sequence of zeros and ones, then label, you concatenate labels and
link |
you, and that that's actually the, the, and you could, you could start by
link |
having training data first, and then afterwards you have test data.
link |
So theoretically any problem could be casted as a problem of predicting
link |
zeros and ones on this, uh, infinite tape.
link |
So, um, so let's say you read already N bits and you want to predict N plus
link |
one bit, and I will ask you to write every possible program that generates
link |
So, um, and you can have, you, you choose programming language.
link |
It can be Python or C plus plus.
link |
And the difference between programming languages, uh, might be, there is
link |
a difference by constant asymptotically, your predictions will be equivalent.
link |
So you read N bits, you enumerate all the programs that produce
link |
these N bits in their output.
link |
And then in order to predict N plus one bit, you actually weight the programs
link |
according to their length.
link |
And there is like a, some specific formula, how you weight them.
link |
And then the N plus, uh, one bit prediction is the prediction, uh, from each
link |
of these program, according to that weight.
link |
Like statistically, you pick, so the smaller the program, the more likely
link |
you, you are to pick the, its output.
link |
So, uh, that's, that algorithm is grounded in the hope or the intuition
link |
that the simple answer is the right one.
link |
It's a formalization of it.
link |
Um, it also means like, if you would ask the question after how many years
link |
would, you know, sun explode, uh, you can say, hmm, it's more likely
link |
the answer is due to some power because they're shorter program.
link |
Um, then other, well, I don't have a good intuition about, uh, how different
link |
the space of short programs are from the space of large programs.
link |
Like, what is the universe where short programs, uh, like run things?
link |
Uh, so, so I said, the things have to agree with N bits.
link |
So even if you have, you, you need to start, okay.
link |
If, if you have very short program and they're like a steel, some has, if, if
link |
it's not perfectly prediction of N bits, you have to start errors.
link |
What are the errors?
link |
And that gives you the full program that agrees on N bits.
link |
Oh, so you don't agree with the N bits.
link |
And you store, that's like a longer, a longer program, slightly longer program
link |
because it can take these extra bits of errors.
link |
That's fascinating.
link |
What's what's your intuition about the, the programs that are able to do cool
link |
stuff like intelligence and consciousness, are they, uh, perfectly like, is, is it,
link |
uh, is there if then statements in them?
link |
So like, is there a lot of a good, uh, if then statements in them?
link |
So like, is there a lot of exceptions that they're storing?
link |
So, um, you could imagine if there would be tremendous amount of if statements,
link |
then they wouldn't be that short.
link |
In case of neural networks, you could imagine that, um, what happens is, uh,
link |
they, uh, when you start with an initialized neural network, uh, it stores
link |
internally many possibilities, how the, uh, how the problem can be solved.
link |
And SGD is kind of magnifying some, some, uh, some, uh, paths, which are slightly
link |
similar to the correct answer.
link |
So it's kind of magnifying correct programs.
link |
And in some sense, SGD is a search algorithm in the program space and the
link |
program space is represented by, uh, you know, kind of the wiring inside of the
link |
neural network and there's like an insane number of ways how the features can be
link |
Let me ask you the high level, basic question that's not so basic.
link |
What is deep learning?
link |
Is there a way you'd like to think of it that is different than like
link |
a generic textbook definition?
link |
The thing that I hinted just a second ago is maybe that, uh, closest to how I'm
link |
thinking these days about deep learning.
link |
So, uh, now the statement is, uh, neural networks can represent some programs.
link |
Uh, it seems that various modules that we are actually adding up to, or like, uh,
link |
you know, we, we want networks to be deep because we, we want multiple
link |
steps of the computation and, uh, uh, and deep learning provides the way to
link |
represent space of programs, which is searchable and it's searchable with,
link |
uh, stochastic gradient descent.
link |
So we have an algorithm to search over humongous number of programs and
link |
gradient descent kind of bubbles up the things that are, uh, tend to give correct
link |
So a neural network with a, with fixed weights that's optimized, do you think
link |
of that as a single program?
link |
Um, so there is a, uh, work by Christopher Olaj where he, uh, so he works on
link |
interpretability of neural networks and he was able to, uh, to identify the
link |
neural network, for instance, a detector of a wheel for a car, or the detector of
link |
a mask for a car, and then he was able to separate them out and assemble them, uh,
link |
together using a simple program, uh, for the detector, for a car detector.
link |
That's like, uh, if you think of traditionally defined programs, that's
link |
like a function within a program that this particular neural network was able
link |
to find and you can tear that out, just like you can copy and paste it into a
link |
stack overflow that, so, uh, any program is a composition of smaller programs.
link |
I mean, the nice thing about the neural networks is that it allows the things
link |
to be more fuzzy than in case of programs.
link |
Uh, in case of programs, you have this, like a branching this way or that way.
link |
And the neural networks, they, they have an easier way to, to be somewhere in
link |
between or to share things.
link |
What is the most beautiful or surprising idea in deep learning and the utilization
link |
of these neural networks, which by the way, for people who are not familiar,
link |
neural networks is a bunch of, uh, what would you say it's inspired by the human
link |
brain, there's neurons, there's connection between those neurons, there's inputs and
link |
there's outputs and there's millions or billions of those neurons and the
link |
learning happens in the neural network.
link |
Neurons and the learning happens, uh, by adjusting the weights on the
link |
edges that connect these neurons.
link |
Thank you for giving definition that I supposed to do it, but I guess you have
link |
enough empathy to listeners to actually know that the, that that might be useful.
link |
No, that's like, so I'm asking Plato of like, what is the meaning of life?
link |
He's not going to answer.
link |
You're being philosophical and deep and quite profound talking about the space
link |
of programs, which is, which is very interesting, but also for people who
link |
just not familiar with the hell we're talking about when we talk about deep
link |
learning anyway, sorry, what is the most beautiful or surprising idea to you in,
link |
in, um, in all the time you've worked at deep learning and you worked on a lot of.
link |
Fascinating projects, applications of neural networks.
link |
It doesn't have to be big and profound.
link |
It can be a cool trick.
link |
I mean, I'm thinking about the trick, but like, uh, it's still, uh, I'm using
link |
to me that it works at all that let's say that the extremely simple algorithm
link |
stochastic gradient descent, which is something that I would be able to derive
link |
on the piece of paper to high school student, uh, when put at the, at the
link |
scale of, you know, thousands of machines actually, uh, can create the.
link |
Behaviors we, which we called kind of human like behaviors.
link |
So in general, any application is stochastic gradient descent
link |
to neural networks is, is amazing to you.
link |
So that, or is there a particular application in natural language
link |
reinforcement learning, uh, and also what do you attribute that success to?
link |
What profound insight can we take from the fact that the thing works
link |
for gigantic, uh, sets of variables?
link |
I mean, the interesting thing is this algorithms, they were invented decades
link |
ago and, uh, people actually, uh, gave up on the idea and, um, you know, back
link |
then they thought that we need profoundly different algorithms and they spent a lot
link |
of cycles on very different algorithms.
link |
And I believe that, uh, you know, we have seen that various, uh, various innovations
link |
that say like transformer or, or dropout or so they can, uh, you know, pass the
link |
help, but it's also remarkable to me that this algorithm from sixties or so, uh, or,
link |
I mean, you can even say that the gradient descent was invented by Leibniz in, I
link |
guess, 18th century or so that actually is the core of learning in the past.
link |
In the past people are, it's almost like a, out of the, maybe an ego, people are
link |
saying that it cannot be the case that such a simple algorithm is there, you
link |
know, uh, could solve complicated problems.
link |
So they were in search for the other algorithms.
link |
And as I'm saying, like, I believe that actually we are in the game where there
link |
is, there are actually frankly three levers.
link |
There is compute, there are algorithms and there is data.
link |
And, uh, if we want to build intelligent systems, we have to pull, uh, all three
link |
levers and they are actually multiplicative.
link |
Um, it's also interesting.
link |
So you ask, is it only compute?
link |
Uh, people internally, they did the studies to determine how much gains they
link |
were coming from different levers.
link |
And so far we have seen that more gains came from compute than algorithms, but
link |
also we are in the world that in case of compute, there is a kind of, you know,
link |
exponential increase in funding and at some point it's impossible to, uh, invest
link |
more, it's impossible to, you know, invest $10 trillion as we are speaking about
link |
the, let's say all taxes in us.
link |
Uh, but you're talking about money that could be innovation in the compute.
link |
That's that's true as well.
link |
Uh, so I mean, they're like a few pieces.
link |
So one piece is human brain is an incredible supercomputer and they're like
link |
a, it, it, it has a hundred trillion parameters or like a, if you try to count
link |
the various quantities in the brain, they're like a neuron synapses that small
link |
number of neurons, there is a lot of synapses it's unclear even how to map, uh,
link |
synapses to, uh, to parameters of neural networks, but it's clear that there are
link |
Yeah. Um, so it might be the case that our networks are still somewhat small.
link |
Uh, it also might be the case that they are more efficient than brain or less
link |
efficient by some, by some huge factor.
link |
Um, I also believe that there will be like a, you know, at the moment we are at
link |
the stage that the, these neural networks, they require thousand X or, or like a
link |
huge factor of more data than humans do.
link |
And it will be a matter of, uh, um, there will be algorithms that vastly decrease
link |
sample complexity, I believe so, but that place where we are heading today is
link |
there are domains which contains million X more data.
link |
And even though computers might be 1000 times slower than humans in learning,
link |
that's not a problem.
link |
Like, uh, for instance, uh, I believe that, uh, it should be possible to create
link |
super human therapist, uh, by, uh, and, and the, the, like, uh, even simple
link |
steps of, of, of doing what, of, of doing it.
link |
And, you know, the, the core reason is there is just machine will be able to
link |
read way more transcripts of therapies, and then it should be able to speak
link |
simultaneously with many more people and it should be possible to optimize it,
link |
uh, all in parallel.
link |
And, uh, well, there's now you're touching on something I deeply care about
link |
and think is way harder than we imagine.
link |
Um, what's the goal of a therapist?
link |
What's the goal of therapies?
link |
So, okay, so one goal now this is terrifying to me, but there's a lot of
link |
people that, uh, contemplate suicide, suffer from depression, uh, and they
link |
could significantly be helped with therapy and the idea that an AI algorithm
link |
might be in charge of that, it's like a life and death task.
link |
It's, uh, the stakes are high.
link |
So one goal for a therapist, whether human or AI is to prevent suicide
link |
ideation to prevent suicide.
link |
How do you achieve that?
link |
So to be clear, I don't think that the current models are good enough for such
link |
a task because it requires insane amount of understanding, empathy, and the
link |
models are far from this place, but it's.
link |
But do you think that understanding empathy, that signal is in the data?
link |
Um, I think there is some signal in the data.
link |
I mean, there are plenty of transcripts of conversations and it is possible to,
link |
it is possible from it to understand personalities.
link |
It is possible from it to understand, uh, if conversation is, uh,
link |
friendly, uh, amicable, uh, uh, antagonistic, it is, I believe that the,
link |
you know, given the fact that the models that we train now, they can, uh, they
link |
can have, they are chameleons that they can have any personality, they might
link |
turn out to be better in understanding, uh, personality of other people than
link |
anyone else and they empathetic to be empathetic.
link |
Yeah, interesting. Uh, but I wonder if there's some level of, uh, multiple
link |
modalities required to be able to, um, be empathetic of the human experience,
link |
whether language is not enough to understand death, to understand fear,
link |
to understand, uh, childhood trauma, to understand, uh, wit and humor required
link |
when you're dancing with a person who might be depressed or suffering both
link |
humor and hope and love and all those kinds of things.
link |
So there's another underlying question, which is self supervised versus
link |
So can you get that from the data by just reading a huge number of transcripts?
link |
I actually, so I think that reading huge number of transcripts is a step one.
link |
It's like at the same way as you cannot learn to dance if just from YouTube by
link |
watching it, you have to actually try it out yourself.
link |
And so I think that here that's a similar situation.
link |
I also wouldn't deploy the system in the high stakes situations right away, but
link |
kind of see gradually where it goes.
link |
And, uh, obviously initially, uh, it would have to go hand in hand with humans.
link |
But, uh, at the moment we are in the situation that actually there is many
link |
more people who actually would like to have a therapy or, or speak with, with
link |
someone than there are therapies out there.
link |
I can, you know, I was so, so fundamentally I was thinking, what are
link |
the things that, uh, can vastly increase people's well being therapy is one of
link |
them being meditation is other one, I guess maybe human connection is a third
link |
one, and I guess pharmacologically it's also possible, maybe direct brain
link |
stimulation or something like that.
link |
But these are pretty much options out there.
link |
Then let's say the way I'm thinking about the AGI endeavor is by default,
link |
that's an endeavor to, uh, increase amount of wealth.
link |
And I believe that we can invest the increase amount of wealth for everyone
link |
and simultaneously.
link |
So, I mean, there are like a two endeavors that make sense to me.
link |
One is like essentially increase amount of wealth.
link |
And second one is, uh, increase overall human wellbeing.
link |
And those are coupled together and they, they can, like, uh, I would
link |
say these are different topics.
link |
One can help another and, uh, you know, therapist is a, is a funny word
link |
because I see friendship and love as therapy.
link |
I mean, so therapist broadly defined as just friendship as a friend.
link |
So like therapist is, has a very kind of clinical sense to it, but what
link |
is human connection you're like, uh, not to get all Camus and Dostoevsky on you,
link |
but you know, life is suffering and we draw, we seek connection with the
link |
humans as we, uh, desperately try to make sense of this world in a deep
link |
overwhelming loneliness that we feel inside.
link |
So I think connection has to do with understanding.
link |
And I think that almost like a lack of understanding causes suffering.
link |
If you speak with someone and do you, do you feel ignored that actually causes pain?
link |
If you are feeling deeply understood that actually they, they, they might
link |
not even tell you what to do in life, but like a pure understanding
link |
or just being heard, understanding is a kind of, that's a lot, you know,
link |
just being heard, feel like you're being heard, like somehow that's a
link |
alleviation temporarily of the loneliness that if somebody knows
link |
you're here with their body language, with the way they are, with the way
link |
they look at you, with the way they talk, do you feel less alone for a brief moment?
link |
Yeah, very much agree.
link |
So I thought in the past about, um, somewhat similar question to yours,
link |
which is what is love, uh, rather what is connection.
link |
Yes. And, um, and obviously I think about these things from AI perspective.
link |
What would it mean?
link |
Um, so I said that, um, you know, intelligence has to do with some compression,
link |
which is more or less like I can say, almost understanding of what is going around.
link |
It seems to me that, uh, other aspect is there seem to be reward functions and you
link |
can have, uh, uh, you know, reward for, uh, food, for maybe human connection, for,
link |
uh, let's say warmth, uh, sex and so on.
link |
And, um, and it turns out that the various people might be optimizing slightly
link |
different, uh, reward functions.
link |
They essentially might care about different things.
link |
And, uh, uh, in case of, uh, love at least the love between two people, you can say
link |
that the, um, you know, boundary between people dissolves to such extent that, uh,
link |
they end up optimizing each other reward functions and yeah, oh, that's interesting.
link |
Um, celebrate the success of each other.
link |
In some sense, I would say love means, uh, helping others to optimize their, uh,
link |
reward functions, not your reward functions, not the things that you think are
link |
important, but the things that the person cares about, you try to help them to,
link |
So love is, uh, if you think of two reward functions, you just, it's a condition.
link |
You combine them together, pretty much maybe like with a weight and it depends
link |
like the dynamic of the relationship.
link |
I mean, you could imagine that if you're fully, uh, optimizing someone's reward
link |
function without yours, then, then maybe are creating codependency or something
link |
like that, but I'm not sure what's the appropriate weight, but the interesting
link |
thing is I even, I even think that the, uh, individual reward function is
link |
saying that the individual person, uh, uh, we ourselves, we are actually less
link |
of a unified insight.
link |
So for instance, if you look at, at the donut on the one level, you might think,
link |
oh, this is like, it looks tasty.
link |
I would like to eat it on other level.
link |
You might tell yourself, I shouldn't be doing it because I want to gain muscles.
link |
So, and you know, you might do it regardless kind of against yourself.
link |
So it seems that even within ourselves, they're almost like a kind of intertwined
link |
personas and, um, I believe that the self love means that, uh, the love between all
link |
these personas, which also means being able to love, love yourself when we are
link |
angry or stressed or so combining all those reward functions of the different
link |
selves you have and accepting that they are there, like, uh, you know, often
link |
people, they have a negative self talk or they say, I don't like when I'm angry.
link |
And like, I try to imagine, try to imagine if there would be like a small
link |
baby Lex, like a five years old, angry, and then they are like, you shouldn't
link |
Like stop being angry.
link |
But like an instant, actually you want the Lex to come over, give him a hug and
link |
just like, I say, it's fine.
link |
It's going to be angry as long as you want.
link |
And then he would stop or, or maybe not, or maybe not, but you cannot expect it
link |
But still, that doesn't explain the why of love.
link |
Like why is love part of the human condition?
link |
Why is it useful to combine the reward functions?
link |
It seems like that doesn't, I mean, I don't think reinforcement learning
link |
frameworks can give us answers to why even, even the Hutter framework has
link |
an objective function that's static.
link |
So we came to existence as a consequence of evolutionary process.
link |
And in some sense, the purpose of evolution is survival.
link |
And then the, this complicated optimization objective baked into us, let's
link |
say compression, which might help us operate in the real world and it baked
link |
into us various reward functions.
link |
Then to be clear at the moment we are operating in the regime, which is somewhat
link |
out of distribution, where they even evolution optimized us.
link |
It's almost like love is a consequence of a cooperation that we've discovered is
link |
In some way it's even the case.
link |
If you, I just love the idea that love is like the out of distribution.
link |
Or it's not out of distribution.
link |
It's like, as you said, it evolved for cooperation.
link |
And I believe that the cop, like in some sense, cooperation ends up helping each
link |
of us individually, so it makes sense evolutionary and there is a, in some
link |
sense, and, you know, love means there is this dissolution of boundaries that you
link |
have a shared reward function and we evolve to actually identify ourselves with
link |
larger groups, so we can identify ourselves, you know, with a family, we can
link |
identify ourselves with a country to such extent that people are willing to give
link |
away their life for country.
link |
So there is, we are wired actually even for love.
link |
And at the moment, I guess, the, maybe it would be somewhat more beneficial if you
link |
will, if we would identify ourselves with all the humanity as a whole.
link |
So you can clearly see when people travel around the world, when they run into
link |
person from the same country, they say, oh, which CPR and all this, like all the
link |
sudden they find all these similarities.
link |
They find some, they befriended those folks earlier than others.
link |
So there is like a sense, some sense of the belonging. And I would say, I think
link |
it would be overall good thing to the world for people to move towards, I think
link |
it's even called open individualism, move toward the mindset of a larger and
link |
So the challenge there, that's a beautiful vision and I share it to expand
link |
that circle of empathy, that circle of love towards the entirety of humanity.
link |
But then you start to ask, well, where do you draw the line?
link |
Because why not expand it to other conscious beings?
link |
And then finally, for our discussion, something I think about is why not
link |
expand it to AI systems?
link |
Like we, we start respecting each other when the, the person, the entity on the
link |
other side has the capacity to suffer.
link |
Cause then we develop a capacity to sort of empathize.
link |
And so I could see AI systems that are interacting with humans more and more
link |
having conscious, like displays.
link |
So like they display consciousness through language and through other means.
link |
And so then the question is like, well, is that consciousness?
link |
Because they're acting conscious.
link |
And so, you know, the reason we don't like torturing animals is because
link |
they look like they're suffering when they're tortured and if AI looks like
link |
it's suffering when it's tortured, how is that not requiring of the same kind
link |
of empathy from us and respect and rights that animals do and other humans do?
link |
I think it requires empathy as well.
link |
I mean, I would like, I guess us or humanity or so make a progress in
link |
understanding what consciousness is, because I don't want just to be speaking
link |
about that, the philosophy, but rather actually make a scientific, uh, to have
link |
a, like, you know, there was a time that people thought that there is a force of
link |
life and, uh, the things that have this force, they are alive.
link |
And, um, I think that there is actually a path to understand exactly what
link |
consciousness is and how it works.
link |
Understand exactly what consciousness is.
link |
And, uh, um, in some sense, it might require essentially putting
link |
probes inside of a human brain, uh, what Neuralink, uh, does.
link |
So the goal there, I mean, there's several things with consciousness
link |
that make it a real discipline, which is one is rigorous
link |
measurement of consciousness.
link |
And then the other is the engineering of consciousness,
link |
which may or may not be related.
link |
I mean, you could also run into trouble.
link |
Like, for example, in the United States for the department, DOT,
link |
department of transportation, and a lot of different places
link |
put a value on human life.
link |
I think DOT is, uh, values $9 million per person.
link |
Sort of in that same way, you can get into trouble.
link |
If you put a number on how conscious of being is, because then you can start
link |
making policy, if a cow is a 0.1 or like, um, 10% as conscious as a human,
link |
then you can start making calculations and it might get you into trouble.
link |
But then again, that might be a very good way to do it.
link |
I would like, uh, to move to that place that actually we have scientific
link |
understanding what consciousness is.
link |
And then we'll be able to actually assign value.
link |
And I believe that there is even the path for the experimentation in it.
link |
So, uh, you know, w we said that, you know, you could put the
link |
probes inside of the brain.
link |
There is actually a few other things that you could do with
link |
devices like Neuralink.
link |
So you could imagine that the way even to measure if AI system is conscious
link |
is by literally just plugging into the brain.
link |
Um, I mean, that, that seems like it's kind of easy, but the plugging
link |
into the brain and asking person if they feel that their consciousness
link |
expanded, um, this direction of course has some issues.
link |
You can say, you know, if someone takes a psychedelic drug, they might
link |
feel that their consciousness expanded, even though that drug
link |
itself is not conscious.
link |
So like, you can't fully trust the self report of a person saying their,
link |
their consciousness is expanded or not.
link |
Let me ask you a little bit about psychedelics is, uh, there've been
link |
a lot of excellent research on, uh, different psychedelics, psilocybin,
link |
MDMA, even DMT drugs in general, marijuana too.
link |
Uh, what do you think psychedelics do to the human mind?
link |
It seems they take the human mind to some interesting places.
link |
Is that just a little, uh, hack, a visual hack, or is there some
link |
profound expansion of the mind?
link |
So let's see, I don't believe in magic.
link |
I believe in, uh, I believe in, uh, in science in, in causality, um, still,
link |
let's say, and then as I said, like, I think that the brain, that the, our
link |
subjective experience of reality is, uh, we live in the simulation run by our
link |
brain and the simulation that our brain runs, they can be very pleasant or very
link |
hellish drugs, they are changing some hyper parameters of the simulation.
link |
It is possible thanks to change of these hyper parameters to actually look back
link |
on your experience and even see that the given things that we took for
link |
granted, they are changeable.
link |
So they allow to have a amazing perspective.
link |
There is also, for instance, the fact that after DMT people can see the
link |
full movie inside of their head, gives me further belief that the brain can generate
link |
that full movie, that the brain is actually learning the model of reality
link |
to such extent that it tries to predict what's going to happen next.
link |
Very high resolution.
link |
So it can replay reality.
link |
Extremely high resolution.
link |
It's also kind of interesting to me that somehow there seems to be some similarity
link |
between these, uh, drugs and meditation itself.
link |
And I actually started even these days to think about meditation as a psychedelic.
link |
Do you practice meditation?
link |
I practice meditation.
link |
I mean, I went a few times on the retreats and it feels after like after
link |
second or third day of meditation, uh, there is a, there is almost like a
link |
sense of, you know, tripping what, what does the meditation retreat entail?
link |
So you w you wake up early in the morning and you meditate for extended
link |
period of time, uh, and yeah, so it's optimized, even though there are other
link |
people, it's optimized for isolation.
link |
So you don't speak with anyone.
link |
You don't actually look into other people's eyes and, uh, you know, you sit
link |
on the chair and say Vipassana meditation tells you, uh, to focus on the breath.
link |
So you try to put, uh, all the, all attention into breathing and, uh,
link |
breathing in and breathing out.
link |
And the crazy thing is that as you focus attention like that, uh, after some
link |
time, their stems starts coming back, like some memories that you completely
link |
forgotten, it almost feels like, uh, that you'll have a mailbox and then you know,
link |
you are just like a archiving email one by one.
link |
And at some point, at some point there is this like a amazing feeling of getting
link |
to mailbox zero, zero emails.
link |
And, uh, it's very pleasant.
link |
It's, it's kind of, it's, it's, it's crazy to me that, um, that once you
link |
resolve these, uh, inner store stories or like inner traumas, then once there is
link |
nothing, uh, left that default, uh, state of human mind is extremely peaceful and
link |
happy, extreme, like, uh, some sense it, it feels that the, it feels at least to
link |
me that way, how, when I was a child that I can look at any object and it's very
link |
beautiful, I have a lot of curiosity about the simple things and that's where
link |
the usual meditation takes me.
link |
Are you, what are you experiencing?
link |
Are you just taking in simple sensory information and they're just enjoying
link |
the rawness of that sensory information?
link |
So there's no, there's no memories or all that kind of stuff.
link |
You're just enjoying being.
link |
Yeah, pretty much.
link |
I mean, still there is, uh, that it's, it's thoughts are slowing down.
link |
Sometimes they pop up, but it's also somehow the extended meditation takes you
link |
to the space that they are way more friendly, way more positive.
link |
Um, there is also this, uh, this thing that, uh, we've, it almost feels that the.
link |
It almost feels that the, we are constantly getting a little bit of a reward
link |
function and we are just spreading this reward function on various activities.
link |
But if you'll stay still for extended period of time, it kind of accumulates,
link |
accumulates, accumulates, and, uh, there is a, there is a sense, there is a sense
link |
that some point it passes some threshold and it feels as drop is falling into kind
link |
of ocean of love and this, and that's like, uh, this is like a very pleasant.
link |
And that's, I'm saying like, uh, that corresponds to the subjective experience.
link |
Some people, uh, I guess in spiritual community, they describe it that that's
link |
the reality, and I would say, I believe that they're like, uh, all sorts of
link |
subjective experience that one can have.
link |
And, uh, I believe that for instance, meditation might take you to the
link |
subjective experiences with the subject.
link |
Vision might take you to the subjective experiences, which are
link |
very pleasant, collaborative.
link |
And I would like a word to move toward a more collaborative, uh, uh, place.
link |
I would say that's very pleasant and I enjoy doing stuff like that.
link |
I, um, I wonder how that maps to your, uh, mathematical model of love with, uh,
link |
the reward function, combining a bunch of things, it seems like our life then is
link |
just, we have this reward function and we're accumulating a bunch of stuff
link |
in it with weights, it's like, um, like multi objective and what meditation
link |
is, is you just remove them, remove them until the weight on one, uh, or
link |
just a few is very high and that's where the pleasure comes from.
link |
So something similar, how I'm thinking about this.
link |
So I told you that there is this like, uh, that there is a story of who you are.
link |
And I think almost about it as a, you know, text prepended to GPT.
link |
And, uh, some people refer to it as ego.
link |
There's like a story who, who, who you are.
link |
So ego is the prompt for GPT three or GPT.
link |
And that's description of you.
link |
And then with meditation, you can get to the point that actually you experience
link |
things without the prompt and you experience things like as they are, you
link |
are not biased over the description, how they supposed to be, uh, that's very
link |
And then we've respected the reward function.
link |
Uh, it's possible to get to the point that the, there is the solution of self.
link |
And therefore you can say that the, or you're having a, your, or like a, your
link |
brain attempts to simulate the reward function of everyone else or like
link |
everything that's that there is this like a love, which feels like a oneness with
link |
And that's also, you know, very beautiful, very pleasant.
link |
At some point you might have a lot of altruistic thoughts during that moment.
link |
And then the self, uh, always comes back.
link |
How would you recommend if somebody is interested in meditation, like a big
link |
thing to take on as a project, would you recommend a meditation retreat?
link |
How many days, what kind of thing would you recommend?
link |
I think that actually retreat is the way to go.
link |
Um, it almost feels that, uh, um, as I said, like a meditation is a psychedelic,
link |
but, uh, when you take it in the small dose, you might barely feel it.
link |
Once you get the high dose, actually you're going to feel it.
link |
Um, so even cold turkey, if you haven't really seriously meditated for a long
link |
period of time, just go to a retreat.
link |
How many days, how many days?
link |
Start weekend one weekend.
link |
So like two, three days.
link |
And it's like, uh, it's interesting that first or second day, it's hard.
link |
And at some point it becomes easy.
link |
There's a lot of seconds in a day.
link |
How hard is the meditation retreat just sitting there in a chair?
link |
So the thing is actually, it literally just depends on your, uh, on the,
link |
your own framing, like if you are in the mindset that you are waiting for it to
link |
be over, or you are waiting for a Nirvana to happen, you are waiting
link |
it will be very unpleasant.
link |
And in some sense, even the difficulty, it's not even in the lack of being
link |
able to speak with others, like, uh, you're sitting there, your legs
link |
will hurt from sitting in terms of like the practical things.
link |
Do you experience kind of discomfort, like physical discomfort of just
link |
sitting, like your, your butt being numb, your legs being sore, all that kind of
link |
You experience it.
link |
And then the, the, they teach you to observe it rather.
link |
And it's like, uh, the crazy thing is you at first might have a feeling
link |
toward trying to escape it and that becomes very apparent that that's
link |
extremely unpleasant.
link |
And then you just, just observe it.
link |
And then at some point it just becomes, uh, it just is, it's like, uh, I remember
link |
that we've, Ilya told me some time ago that, uh, you know, he takes a cold
link |
shower and he's the mindset of taking a cold shower was to embrace suffering.
link |
This is your style?
link |
Yeah, it's my style.
link |
So my style is actually, I also sometimes take cold showers.
link |
It is purely observing how the water goes through my body, like a purely being
link |
present, not trying to escape from there.
link |
And I would say then it actually becomes pleasant.
link |
It's not like, ah, well, that that's interesting.
link |
Um, I I'm also that mean that's, that's the way to deal with anything really
link |
difficult, especially in the physical space is to observe it to say it's pleasant.
link |
It's a D I would use a different word.
link |
You're, um, you're accepting of the full beauty of reality.
link |
I would say, cause say pleasant.
link |
But yeah, I mean, in some sense it is pleasant.
link |
That's the only way to deal with a cold shower is to, to, uh, become an
link |
observer and to find joy in it.
link |
Um, same with like really difficult, physical, um, exercise or like running
link |
for a really long time, endurance events, just anytime you're, any kind of pain.
link |
I think the only way to survive it is not to resist it is to observe it.
link |
You mentioned, you mentioned, um, you mentioned, um, you mentioned
link |
Ilya, Ilya says, it's very, he's our chief scientist, but also
link |
he's very close friend of mine.
link |
He cofounded open air with you.
link |
I've spoken with him a few times.
link |
I really enjoy talking to him.
link |
His mind, just like yours works in fascinating ways.
link |
Now, both of you are not able to define deep learning simply.
link |
Uh, what's it like having him as somebody you have technical discussions with on
link |
in the space of machine learning, deep learning, AI, but also life.
link |
What's it like when these two, um, agents get into a self play situation in a room?
link |
What's it like collaborating with him?
link |
So I believe that we have, uh, extreme, uh, respect to each other.
link |
So, uh, in, I love Ilya's insight, both like, uh, I guess about
link |
consciousness, uh, life AI, but, uh, in terms of the, it's interesting to
link |
me, cause you're a brilliant, uh, Thinker in the space of machine
link |
learning, like intuition, like digging deep in what works, what doesn't,
link |
why it works, why it doesn't, and so is Ilya.
link |
I'm wondering if there's interesting deep discussions you've had with him in the
link |
past or disagreements that were very productive.
link |
So I can say, I also understood over the time, where are my strengths?
link |
So obviously we have plenty of AI discussions and, um, um, and do you
link |
know, I myself have plenty of ideas, but like I consider Ilya, uh, what
link |
of the most prolific AI scientists in the entire world.
link |
And, uh, I think that, um, I realized that maybe my super skill, um, is, uh,
link |
being able to bring people to collaborate together, that I have some level of
link |
empathy that is unique in AI world.
link |
And that might come, you know, from either meditation, psychedelics, or
link |
let's say I read just hundreds of books on this topic.
link |
So, and I also went through a journey of, you know, I developed a
link |
lot of, uh, algorithms, so I think that maybe I can, that's my super human skill.
link |
Uh, Ilya is, uh, one of the best AI scientists, but then I'm pretty
link |
good in assembling teams and I'm also not holding to people.
link |
Like I'm growing people and then people become managers at OpenAI.
link |
I grew many of them, like a research managers.
link |
So you, you find, you find places where you're excellent and he finds like his,
link |
his, his deep scientific insights is where he is and you find ways you can,
link |
the puzzle pieces fit together.
link |
Like, uh, you know, ultimately, for instance, let's say Ilya, he doesn't
link |
manage people, uh, that's not what he likes or so.
link |
Um, I like, I like hanging out with people.
link |
By default, I'm an extrovert and I care about people.
link |
Oh, interesting. Okay. All right. Okay, cool.
link |
So that, that fits perfectly together.
link |
But I mean, uh, I also just like your intuition about various
link |
problems in machine learning.
link |
He's definitely one I really enjoy.
link |
I remember talking to him about something I was struggling with, which
link |
is coming up with a good model for pedestrians, for human beings across
link |
the street in the context of autonomous vehicles, and I was like, okay,
link |
in the context of autonomous vehicles.
link |
And he immediately started to like formulate a framework within which you
link |
can evolve a model for pedestrians, like through self play, all that kind of
link |
mechanisms, the depth of thought on a particular problem, especially problems
link |
he doesn't know anything about is, is fascinating to watch.
link |
It makes you realize like, um, yeah, the, the, the limits of the, that the human
link |
intellect may be limitless, or it's just impressive to see a descendant of
link |
ape come up with clever ideas.
link |
I mean, so even in the space of deep learning, when you look at various
link |
people, there are people now who invented some breakthroughs once, but
link |
there are very few people who did it multiple times.
link |
And you can think if someone invented it once, that might be just a sheer luck.
link |
And if someone invented it multiple times, you know, if a probability of
link |
inventing it once is one over a million, then probability of inventing it twice
link |
or three times would be one over a million square or, or to the power of
link |
three, which, which would be just impossible.
link |
So it literally means that it's, it's given that, uh, it's not the luck.
link |
And Ilya is one of these few people who, uh, uh, who have, uh, a lot of
link |
these inventions in his arsenal.
link |
It also feels that, um, you know, for instance, if you think about folks
link |
like Gauss or Euler, uh, you know, at first they read a lot of books and then
link |
they did thinking and then they figure out math and that's how it feels with
link |
Ilya, you know, at first he read stuff and then like he spent his thinking cycles.
link |
And that's a really good way to put it.
link |
When I talk to him, I, I see thinking.
link |
He's actually thinking, like, he makes me realize that there's like deep
link |
thinking that the human mind can do.
link |
Like most of us are not thinking deeply.
link |
Uh, like you really have to put in a lot of effort to think deeply.
link |
Like I have to really put myself in a place where I think deeply about a
link |
problem, it takes a lot of effort.
link |
It's like, uh, it's like an airplane taking off or something.
link |
You have to achieve deep focus.
link |
He he's just, uh, he's what is it?
link |
He said, what does it, his brain is like a vertical takeoff in
link |
terms of airplane analogy.
link |
So it's interesting, but it, I mean, Cal Newport talks about
link |
this as ideas of deep work.
link |
It's, you know, most of us don't work much at all in terms of like, like deeply
link |
think about particular problems, whether it's a math engineering, all that kind
link |
of stuff, you want to go to that place often and that's real hard work.
link |
And some of us are better than others at that.
link |
So I think that the big piece has to do with actually even engineering
link |
your environment that says that it's conducive to that.
link |
So, um, see both Ilya and I, uh, on the frequent basis, we kind of disconnect
link |
ourselves from the world in order to be able to do extensive amount of thinking.
link |
So Ilya usually, he just, uh, leaves iPad at hand.
link |
He loves his iPad.
link |
And, uh, for me, I'm even sometimes, you know, just going for a few days
link |
to different location to Airbnb, I'm turning off my phone and there is no
link |
access to me and, uh, that's extremely important for me to be able to actually
link |
just formulate new thoughts, to do deep work rather than to be reactive.
link |
And the, the, the older I am, the more of these random tasks are at hand.
link |
Before I go on to that, uh, thread, let me return to our friend, GPT.
link |
And let me ask you another ridiculously big question.
link |
Can you give an overview of what GPT three is, or like you say in
link |
your Twitter bio, GPT N plus one, how it works and why it works.
link |
So, um, GPT three is a humongous neural network.
link |
Um, let's assume that we know what is neural network, the definition, and it
link |
is trained on the entire internet and just to predict next word.
link |
So let's say it sees part of the, uh, article and it, the only task that it
link |
has at hand, it is to say what would be the next word and what would be the next
link |
word and it becomes a really exceptional at the task of figuring out what's the
link |
next word. So you might ask, why would, uh, this be an important, uh, task?
link |
Why would it be important to predict what's the next word?
link |
And it turns out that a lot of problems, uh, can be formulated, uh, as a text
link |
completion problem.
link |
So GPT is purely, uh, learning to complete the text.
link |
And you could imagine, for instance, if you are asking a question, uh, who is
link |
the president of the United States, then GPT can give you an answer to it.
link |
It turns out that many more things can be formulated this way.
link |
You can format text in the way that you have sentence in English.
link |
You make it even look like some content of a website, uh, elsewhere, which would
link |
be teaching people how to translate things between languages.
link |
So it would be EN colon, uh, text in English, FR colon, and then you'll
link |
uh, uh, and then you'll ask people and then you ask model to, to continue.
link |
And it turns out that the, such a model is predicting translation from English
link |
The crazy thing is that this model can be used for way more sophisticated tasks.
link |
So you can format text such that it looks like a conversation between two people.
link |
And that might be a conversation between you and Elon Musk.
link |
And because the model read all the texts about Elon Musk, it will be able to
link |
predict Elon Musk words as it would be Elon Musk.
link |
It will speak about colonization of Mars, about sustainable future and so on.
link |
And it's also possible to, to even give arbitrary personality to the model.
link |
You can say, here is a conversation that we've a friendly AI bot.
link |
And the model, uh, will complete the text as a friendly AI bot.
link |
So, I mean, how do I express how amazing this is?
link |
So just to clarify, uh, a conversation, generating a conversation between me and
link |
Elon Musk, it wouldn't just generate good examples of what Elon would say.
link |
It would get the same results as the conversation between Elon Musk and me.
link |
Say it would get the syntax all correct.
link |
So like interview style, it would say like Elon call and Lex call, like it,
link |
it's not just like, uh, inklings of, um, semantic correctness.
link |
It's like the whole thing, grammatical, syntactic, semantic, it's just really,
link |
really impressive, uh, generalization.
link |
I mean, I also want to, you know, provide some caveats so it can generate
link |
few paragraphs of coherent text, but as you go to, uh, longer pieces,
link |
it, uh, it actually goes off the rails.
link |
If you try to write a book, it won't work out this way.
link |
What way does it go off the rails, by the way?
link |
Is there interesting ways in which it goes off the rails?
link |
Like what falls apart first?
link |
So the model is trained on the, all the existing data, uh, that is out there,
link |
which means that it is not trained on its own mistakes.
link |
So for instance, if it would make a mistake, then, uh, I kept,
link |
so to give you, give you an example.
link |
So let's say I have a conversation with a model pretending that is Elon Musk.
link |
And then I start putting some, uh, I'm start actually making up
link |
things which are not factual.
link |
Um, I would say like Twitter, but I got you.
link |
Um, like, uh, I don't know.
link |
I would say that Elon is my wife and the model will just keep on carrying it on.
link |
And as if it's true.
link |
And in some sense, if you would have a normal conversation with Elon,
link |
he would be what the fuck.
link |
There'll be some feedback between, so the model is trained on things
link |
that humans have written, but through the generation process, there's
link |
no human in the loop feedback.
link |
That's fascinating.
link |
So it's magnified.
link |
It's like the errors get magnified and magnified and it's also interesting.
link |
I mean, first of all, humans have the same problem.
link |
It's just that we, uh, we'll make fewer errors and magnify the errors slower.
link |
I think that actually what happens with humans is if you have a wrong
link |
belief about the world as a kid, then very quickly we'll learn that it's
link |
not correct because they are grounded in reality and they are learning
link |
from your new experience.
link |
But do you think the model can correct itself too?
link |
Won't it through the power of the representation.
link |
And so the absence of Elon Musk being your wife information on the
link |
internet, won't it correct itself?
link |
There won't be examples like that.
link |
So the errors will be subtle at first.
link |
And in some sense, you can also say that the data that is not out there is
link |
the data, which would represent how the human learns and maybe model would
link |
be learned, trained on such a data.
link |
Then it would be better off.
link |
How intelligent is GPT3 do you think?
link |
Like when you think about the nature of intelligence, it
link |
seems exceptionally impressive.
link |
But then if you think about the big AGI problem, is this
link |
footsteps along the way to AGI?
link |
So let's see, it seems that intelligence itself is, there are multiple axis of it.
link |
And I would expect that the systems that we are building, they might end up being
link |
superhuman on some axis and subhuman on some other axis.
link |
It would be surprising to me on all axis simultaneously, they would become superhuman.
link |
Of course, people ask this question, is GPT a spaceship that would take us to
link |
the moon or are we putting a, building a ladder to heaven that we are just
link |
building bigger and bigger ladder.
link |
And we don't know in some sense, which one of these two.
link |
Which one is better?
link |
I'm trying to, I like stairway to heaven.
link |
So I'm not exactly sure which one is better, but you're saying like the
link |
spaceship to the moon is actually effective.
link |
So people who criticize GPT, they say, you guys just building a
link |
taller, a ladder, and it will never reach the moon.
link |
And at the moment, I would say the way I'm thinking is, is like a scientific question.
link |
And I'm also in heart, I'm a builder creator and like, I'm thinking, let's try out, let's
link |
see how far it goes.
link |
And so far we see constantly that there is a progress.
link |
So do you think GPT four, GPT five, GPT N plus one will, um, there'll be a phase
link |
shift, like a transition to a, to a place where we'll be truly surprised.
link |
Then again, like GPT three is already very like truly surprising.
link |
The people that criticize GPT three as a stair, as a, what is it?
link |
I think too quickly get accustomed to how impressive it is that they're
link |
impressive, it is that the prediction of the next word can achieve such depth of
link |
semantics, accuracy of syntax, grammar, and semantics.
link |
Um, do you, do you think GPT four and five and six will continue to surprise us?
link |
I mean, definitely there will be more impressive models that there is a
link |
question of course, if there will be a phase shift and, uh, the, also even the
link |
way I'm thinking about the, about these models is that when we build these
link |
models, you know, we see some level of the capabilities, but we don't even fully
link |
understand everything that the model can do.
link |
And actually one of the best things to do is to allow other people to probe the
link |
model to even see what is possible.
link |
Hence the, the using GPT as an API and opening it up to the world.
link |
I mean, so when I'm thinking from perspective of like, uh, obviously
link |
various people are, that have concerns about AGI, including myself.
link |
Um, and then when I'm thinking from perspective, what's the strategy even to
link |
deploy these things to the world, then the one strategy that I have seen many
link |
times working is that iterative deployment that you deploy, um, slightly
link |
better versions and you allow other people to criticize you.
link |
So you actually, or try it out, you see where are their fundamental issues.
link |
And it's almost, you don't want to be in that situation that you are holding
link |
into powerful system and there's like a huge overhang, then you deploy it and it
link |
might have a random chaotic impact on the world.
link |
So you actually want to be in the situation that they are
link |
gradually deploying systems.
link |
I asked this question of Illya, let me ask you, uh, you this question.
link |
I've been reading a lot about Stalin and power.
link |
If you're in possession of a system that's like AGI, that's exceptionally
link |
powerful, do you think your character and integrity might become corrupted?
link |
Like famously power corrupts and absolute power corrupts.
link |
So I believe that the, you want at some point to work toward distributing the power.
link |
I think that the, you want to be in the situation that actually AGI is not
link |
controlled by a small number of people, uh, but, uh, essentially, uh, by a larger
link |
So the thing is that requires a George Washington style move in the ascent to
link |
power, there's always a moment when somebody gets a lot of power and they
link |
have to have the integrity and, uh, the moral compass to give away that power.
link |
That humans have been good and bad throughout history at this particular
link |
And I wonder, I wonder we like blind ourselves in a, for example, between
link |
nations, a race, uh, towards, um, they, yeah, AI race between nations, we might
link |
blind ourselves and justify to ourselves the development of AI without distributing
link |
the power because we want to defend ourselves against China, against Russia,
link |
that kind of, that kind of logic.
link |
And, um, I wonder how we, um, how we design governance mechanisms that, um,
link |
prevent us from becoming power hungry and in the process, destroying ourselves.
link |
So let's see, I have been thinking about this topic quite a bit, but I also want
link |
to admit that, uh, once again, I actually want to rely way more on Sam Altman on it.
link |
He wrote an excellent blog on how even to distribute wealth.
link |
Um, and he's proper, he proposed in his blog, uh, to tax, uh, equity of the companies
link |
rather than profit and to distribute it.
link |
And this is, this is an example of, uh, Washington move.
link |
I guess I personally have insane trust in some here already spent plenty of money
link |
running, uh, universal basic income, uh, project.
link |
That like, uh, gives me, I guess, maybe some level of trust to him, but I also,
link |
I guess love him as a friend.
link |
I wonder because we're sort of summoning a new set of technologies.
link |
I wonder if we'll be, um, cognizant, like you're describing the process of open AI,
link |
but it could also be at other places like in the U S government, right?
link |
Uh, both China and the U S are now full steam ahead on autonomous
link |
weapons systems development.
link |
And that's really worrying to me because in the framework of something being a
link |
national security danger or military danger, you can do a lot of pretty dark
link |
things that blind our moral compass.
link |
And I think AI will be one of those things, um, in some sense, the, the mission
link |
and the work you're doing in open AI is like the counterbalance to that.
link |
So you want to have more open AI and less autonomous weapons systems.
link |
I, I, I, I like these statements, like to be clear, like this interesting and I'm
link |
thinking about it myself, but, uh, this is a place that I, I, I put my trust
link |
actually in Sam's hands, because it's extremely hard for me to reason about it.
link |
I mean, one important statement to make is, um, it's good to think about this.
link |
No question about it.
link |
No question, even like low level quote unquote engineer, like there's such a,
link |
um, I remember I, I programmed a car, uh, our RC car, um, and it was, it was
link |
programmed a car, uh, our RC car, they went really fast, like 30, 40 miles an hour.
link |
And I remember I was like sleep deprived.
link |
So I programmed it pretty crappily and it like, uh, the, the, the code froze.
link |
So it's doing some basic computer vision and it's going around on track,
link |
but it's going full speed.
link |
And, uh, there was a bug in the code that, uh, the car just went, it didn't turn.
link |
Went straight full speed and smash into the wall.
link |
And I remember thinking the seriousness with which you need to approach the
link |
design of artificial intelligence systems and the programming of artificial
link |
intelligence systems is high because the consequences are high, like that
link |
little car smashing into the wall.
link |
For some reason, I immediately thought of like an algorithm that controls
link |
nuclear weapons, having the same kind of bug.
link |
And so like the lowest level engineer and the CEO of a company all need to
link |
have the seriousness, uh, in approaching this problem and thinking
link |
about the worst case consequences.
link |
So I think that is true.
link |
I mean, the, what I also recognize in myself and others even asking this
link |
question is that it evokes a lot of fear and fear itself ends up being
link |
actually quite debilitating.
link |
The place where I arrived at the moment might sound cheesy or so, but it's
link |
almost to build things out of love rather than fear, like a focus on how, uh, I can,
link |
you know, maximize the value, how the systems that I'm building might be, uh,
link |
I'm not saying that the fear doesn't exist out there and like it totally
link |
makes sense to minimize it, but I don't want to be working because, uh, I'm
link |
scared, I want to be working out of passion, out of curiosity, out of the,
link |
you know, uh, looking forward for the positive future.
link |
With, uh, the definition of love arising from a rigorous practice of empathy.
link |
So not just like your own conception of what is good for the world, but
link |
always listening to others.
link |
Like the love where I'm considering reward functions of others.
link |
Others to limit to infinity is like a sum of like one to N where N is, uh,
link |
7 billion or whatever it is.
link |
Not, not projecting my reward functions on others.
link |
Can we just take a step back to something else?
link |
Super cool, which is, uh, OpenAI Codex.
link |
Can you give an overview of what OpenAI Codex and GitHub Copilot is, how it works
link |
and why the hell it works so well?
link |
So with GPT tree, we noticed that the system, uh, you know, that system train
link |
on all the language out there started having some rudimentary coding capabilities.
link |
So we're able to ask it, you know, to implement addition function between
link |
two numbers and indeed it can write item or JavaScript code for that.
link |
And then we thought, uh, we might as well just go full steam ahead and try to
link |
create a system that is actually good at what we are doing every day ourselves,
link |
which is programming.
link |
We optimize models for proficiency in coding.
link |
We actually even created models that both have a comprehension of language and code.
link |
And Codex is API for these models.
link |
So it's first pre trained on language and then codex.
link |
Then I don't know if you can say fine tuned because there's a lot of code,
link |
but it's language and code.
link |
It's language and code.
link |
It's also optimized for various things.
link |
I can, let's say low latency and so on.
link |
Codex is the API, the similar to GPT tree.
link |
We expect that there will be proliferation of the potential products that can use
link |
coding capabilities and I can, I can speak about it in a second.
link |
Copilot is a first product and developed by GitHub.
link |
So as we're building, uh, models, we wanted to make sure that these
link |
models are useful and we work together with GitHub on building the first product.
link |
Copilot is actually, as you code, it suggests you code completions.
link |
And we have seen in the past, there are like a various tools that can suggest
link |
how to like a few characters of the code or a line of code.
link |
Then the thing about Copilot is it can generate 10 lines of code.
link |
You, it's often the way how it works is you often write in the comment
link |
what you want to happen because people in comments, they describe what happens next.
link |
So, um, these days when I code, instead of going to Google to search, uh, for
link |
the appropriate code to solve my problem, I say, Oh, for this area, could you
link |
smooth it and then, you know, it imports some appropriate libraries and say it
link |
uses NumPy convolution or so I, that I was not even aware that exists and
link |
it does the appropriate thing.
link |
Um, so you, uh, you write a comment, maybe the header of a function
link |
and it completes the function.
link |
Of course, you don't know what is the space of all the possible small
link |
programs that can generate.
link |
What are the failure cases?
link |
How many edge cases, how many subtle errors there are, how many big errors
link |
there are, it's hard to know, but the fact that it works at all in a large
link |
number of cases is incredible.
link |
It's like, uh, it's a kind of search engine into code that's
link |
been written on the internet.
link |
So for instance, when you search things online, then usually you get to the,
link |
some particular case, like if you go to stack overflow and people describe
link |
that one particular situation, uh, and then they seek for a solution.
link |
But in case of a copilot, it's aware of your entire context and in
link |
context is, Oh, these are the libraries that they are using.
link |
That's the set of the variables that is initialized.
link |
And on the spot, it can actually tell you what to do.
link |
So the interesting thing is, and we think that the copilot is one
link |
possible product using codecs, but there is a place for many more.
link |
So internally we tried out, you know, to create other fun products.
link |
So it turns out that a lot of tools out there, let's say Google
link |
calendar or Microsoft word or so, they all have a internal API
link |
to build plugins around them.
link |
So there is a way in the sophisticated way to control calendar or Microsoft word.
link |
Today, if you want, if you want more complicated behaviors from these
link |
programs, you have to add the new button for every behavior.
link |
But it is possible to use codecs and tell for instance, to calendar, uh,
link |
could you schedule an appointment with Lex next week after 2 PM and it
link |
writes corresponding piece of code.
link |
And that's the thing that actually you want.
link |
So you figure out is there's a lot of programs with which
link |
you can interact through code.
link |
And so there you can generate that code from natural language.
link |
That's fascinating.
link |
And that's somewhat like also closest to what was the promise of Siri or Alexa.
link |
So previously all these behaviors, they were hard coded and it seems
link |
that codecs on the fly can pick up the API of let's say, given software.
link |
And then it can turn language into use of this API.
link |
So without hard coding, you can find, it can translate to machine language.
link |
To, uh, so for example, this would be really exciting for me, like for, um,
link |
Adobe products, like Photoshop, uh, which I think action scripted, I think
link |
there's a scripting language that communicates with them, same with Premier.
link |
And do you could imagine that that allows even to do coding by voice on your phone?
link |
So for instance, in the past, okay.
link |
As of today, I'm not editing Word documents on my phone because it's
link |
just the keyboard is too small.
link |
But if I would be able to tell, uh, to my phone, you know, uh, make the
link |
header large, then move the paragraphs around and that's actually what I want.
link |
So I can tell you one more cool thing, or even how I'm thinking about codecs.
link |
So if you look actually at the evolution of, uh, of computers, we started with
link |
a very primitive interfaces, which is a punch card and punch card.
link |
So Charlie, you make a holes in the, in the plastic card to indicate zeros and ones.
link |
And, uh, during that time, there was a small number of specialists
link |
who were able to use computers.
link |
And by the way, people even suspected that there is no need for many
link |
more people to use computers.
link |
Um, but then we moved from punch cards to at first assembly and see, and
link |
at these programming languages, they were slightly higher level.
link |
They allowed many more people to code and they also, uh, led to more
link |
of a proliferation of technology.
link |
And, uh, you know, further on, there was a jump to say from C++ to Java and Python.
link |
And every time it has happened, more people are able to code
link |
and we build more technology.
link |
And it's even, you know, hard to imagine now, if someone will tell you that you
link |
should write code in assembly instead of let's say, Python or Java or JavaScript.
link |
And codecs is yet another step toward kind of bringing computers closer to
link |
humans such that you communicate with a computer with your own language rather
link |
than with a specialized language, and, uh, I think that it will lead to an
link |
increase of number of people who can code.
link |
And then, and the kind of technologies that those people will create is it's
link |
innumerable, it could, you know, it could be a huge number of technologies.
link |
We're not predicting at all because that's less and less requirement
link |
of having a technical mind, a programming mind, you're not opening it to the world
link |
of, um, other kinds of minds, creative minds, artistic minds, all that kind of stuff.
link |
I would like, for instance, biologists who work on DNA to be able to program
link |
and not to need to spend a lot of time learning it.
link |
And I, I believe that's a good thing to the world.
link |
And I would actually add, I would add, so at the moment I'm a managing codecs
link |
team and also language team, and I believe that there is like a plenty
link |
of brilliant people out there and they should have a lot of experience.
link |
There and they should apply.
link |
So what's the language and the codecs is, so those are kind of,
link |
they're overlapping teams.
link |
It's like GPT, the raw language, and then the codecs is like applied to programming.
link |
And they are quite intertwined.
link |
There are many more things involved making this, uh, models,
link |
uh, extremely efficient and deployable.
link |
For instance, there are people who are working to, you know, make our data
link |
centers, uh, amazing, or there are people who work on putting these
link |
models into production or, uh, or even pushing it at the very limit of the scale.
link |
So all aspects from, from the infrastructure to the actual machine.
link |
So I'm just saying there are multiple teams while the, and the team working
link |
on codecs and language, uh, I guess I'm, I'm directly managing them.
link |
I would like, I would love to hire more interested in machine learning.
link |
This is probably one of the most exciting problems and like systems
link |
to be working on is it's actually, it's, it's, it's pretty cool.
link |
Like what, what, uh, the program synthesis, like generating a
link |
programs is very interesting, very interesting problem that has echoes
link |
of reasoning and intelligence in it.
link |
It's and I think there's a lot of fundamental questions that you might
link |
be able to sneak, uh, sneak up to by generating programs.
link |
Yeah, that one more exciting thing about the programs is that, so I said
link |
that the, um, you know, the, in case of language, that one of the travels
link |
is even evaluating language.
link |
So when the things are made up, you, you need somehow either a human to,
link |
to say that this doesn't make sense or so in case of program, there is one extra
link |
lever that we can actually execute programs and see what they evaluate to.
link |
So that process might be somewhat, uh, more automated in, in order to improve
link |
the, uh, qualities of generations.
link |
Oh, that's fascinating.
link |
So like the, wow, that's really interesting.
link |
So, so for the language, the, you know, the simulation to actually
link |
execute it as a human mind.
link |
For programs, there is a, there is a computer on which you can evaluate it.
link |
That's a brilliant little insight.
link |
Insight that the thing compiles and runs that's first and second, you can evaluate
link |
on a, like do automated unit testing and in some sense, it seems to me that we'll
link |
be able to make a tremendous progress.
link |
You know, we are in the paradigm that there is way more data.
link |
There is like a transcription of millions of, uh, of, uh, software engineers.
link |
So, uh, I mean, you just mean, cause I was going to ask you about reliability.
link |
The thing about programs is you don't know if they're going to, like a program
link |
that's controlling a nuclear power plant has to be very reliable.
link |
So I wouldn't start with controlling nuclear power plant maybe one day,
link |
but that's not actually, that's not on the current roadmap.
link |
That's not the step one.
link |
And you know, it's the Russian thing.
link |
You just want to go to the most powerful, destructive, most powerful
link |
the most powerful, destructive thing right away run by JavaScript.
link |
So this is a lower impact, but nevertheless, when you make me
link |
realize it is possible to achieve some levels of reliability by doing testing.
link |
And you could, you could imagine that, you know, maybe there are ways for
link |
model to write event code for testing itself and so on, and there exists
link |
a ways to create the feedback loops that the model could keep on improving.
link |
Yeah. By writing programs that generate tests for the instance, for instance.
link |
And that's how we get consciousness, because it's metacompression.
link |
That's what you're going to write.
link |
That's the comment.
link |
That's the prompt that generates consciousness.
link |
Compressor of compressors.
link |
You just write that.
link |
Do you think the code that generates consciousness will be simple?
link |
I mean, ultimately, the core idea behind will be simple,
link |
but there will be also decent amount of engineering involved.
link |
Like in some sense, it seems that, you know, spreading these models
link |
on many machines, it's not that trivial.
link |
And we find all sorts of innovations that make our models more efficient.
link |
I believe that first models that I guess are conscious or like a truly intelligent,
link |
they will have all sorts of tricks, but then again, there's a Richard Sutton
link |
argument that maybe the tricks are temporary things that they might be
link |
temporary things and in some sense, it's also even important to, to know
link |
that even the cost of a trick.
link |
So sometimes people are eager to put the trick while forgetting that
link |
there is a cost of maintenance or like a long term cost, long term cost
link |
or maintenance, or maybe even flexibility of code to actually implement new ideas.
link |
So even if you have something that gives you 2x, but it requires, you know,
link |
1000 lines of code, I'm not sure if it's actually worth it.
link |
So in some sense, you know, if it's five lines of code and 2x, I would take it.
link |
And we see many of this, but also, you know, that requires some level of,
link |
I guess, lack of attachment to code that we are willing to remove it.
link |
So you led the OpenAI robotics team.
link |
Can you give an overview of the cool things you were able to
link |
accomplish, what are you most proud of?
link |
So when we started robotics, we knew that actually reinforcement learning works
link |
and it is possible to solve fairly complicated problems.
link |
Like for instance, AlphaGo is an evidence that it is possible to build superhuman
link |
Go players, DOTA2 is an evidence that it's possible to build superhuman agents
link |
playing DOTA, so I asked myself a question, you know, what about robots out there?
link |
Could we train machines to solve arbitrary tasks in the physical world?
link |
Our approach was, I guess, let's pick a complicated problem that if we would
link |
solve it, that means that we made some significant progress in the domain.
link |
And if can progress the domain, and then we went after the problem.
link |
So we noticed that actually the robots out there, they are kind of at the moment
link |
optimized per task, so you can have a robot that it's like, if you have a robot
link |
opening a bottle, it's very likely that the end factor is that bottle opener.
link |
And the, and in some sense, that's a hack to be able to solve a task,
link |
which makes any task easier and ask myself, so what would be a robot that
link |
can actually solve many tasks?
link |
And we conclude that human hands have such a quality that indeed they are, you
link |
know, you have five kind of tiny arms attached individually.
link |
They can manipulate pretty broad spectrum of objects.
link |
So we went after a single hand, like trying to solve Rubik's cube single handed.
link |
We picked this task because we thought that there is no way to hard code it.
link |
And it's also, we picked the robot on which it would be hard to hard code it.
link |
And we went after the solution such that it could generalize to other problems.
link |
And just to clarify, it's one robotic hand solving the Rubik's cube.
link |
The hard part isn't the solution to the Rubik's cube is the manipulation of the,
link |
of like having it not fall out of the hand, having it use the, uh, five baby
link |
arms to, uh, what is it like rotate different parts of the Rubik's cube to
link |
achieve the solution.
link |
So what, uh, what was the hardest part about that?
link |
What was the approach taken there?
link |
What are you most proud of?
link |
Obviously we have like a strong belief in reinforcement learning.
link |
And, uh, you know, one path it is to do reinforcement learning, the real
link |
world other path is to, uh, uh, that simulation in some sense, the tricky
link |
part about the real world is at the moment, our models, they require a lot
link |
of data and there is essentially no data.
link |
And, uh, I did, we decided to go through the path of the simulation.
link |
And in simulation, you can have infinite amount of data.
link |
The tricky part is the fidelity of the simulation.
link |
And also can you in simulation represent everything that you represent
link |
otherwise in the real world.
link |
And, you know, it turned out that, uh, that, you know, because there is
link |
lack of fidelity, it is possible to what we, what we arrived at is training
link |
a model that doesn't solve one simulation, but it actually solves the
link |
entire range of simulations, which, uh, uh, in terms of like, uh, what's
link |
the, exactly the friction of the cube or the weight or so, and the single AI
link |
that can solve all of them ends up working well with the reality.
link |
How do you generate the different simulations?
link |
So, uh, you know, there's plenty of parameters out there.
link |
We just pick them randomly.
link |
And, uh, and in simulation model just goes for thousands of years and keeps
link |
on solving Rubik's cube in each of them.
link |
And the thing is that neural network that we used, it has a memory.
link |
And as it presses, for instance, the side of the, of the cube, it can sense,
link |
oh, that's actually, this side was, uh, difficult to press.
link |
I should press it stronger and throughout this process kind of, uh, learn it's even
link |
how to, uh, how to solve this particular instance of the Rubik's cube, like even
link |
mass, it's kind of like, uh, you know, sometimes when you go to a gym and after,
link |
um, after bench press, you try to leave the class and you kind of forgot, uh, and,
link |
and your head goes like up right away because kind of you got used to maybe
link |
different weight and it takes a second to adjust and this kind of, of a memory,
link |
the model gained through the process of interacting with the cube in the
link |
simulation, I appreciate you speaking to the audience with the bench press,
link |
all the bros in the audience, probably working out right now.
link |
There's probably somebody listening to this actually doing bench press.
link |
Um, so maybe, uh, put the bar down and pick up the water bottle and you'll
link |
know exactly what, uh, what Jack is talking about.
link |
So what, uh, what was the hardest part of getting the whole thing to work?
link |
So the hardest part is at the moment when it comes to, uh, physical work, when it
link |
comes to robots, uh, they require maintenance, it's hard to replicate a
link |
million times it's, uh, it's also, it's hard to replay things exactly.
link |
I remember this situation that one guy at our company, he had like a model that
link |
performs way better than other models in solving Rubik's cube.
link |
And, uh, you know, we kind of didn't know what's going on, why it's that.
link |
And, uh, it turned out, uh, that, you know, he was running it from his laptop
link |
that had better CPU or better, maybe local GPU as well.
link |
And, uh, because of that, there was less of a latency and the model was the same.
link |
And that actually made solving Rubik's cube more reliable.
link |
So in some sense, there might be some subtle bugs like that when it comes
link |
to running things in the real world.
link |
Even hinting on that, you could imagine that the initial models you would like
link |
to have models, which are insanely huge neural networks, and you would like to
link |
give them even more time for thinking.
link |
And when you have these real time systems, uh, then you might be constrained
link |
actually by the amount of latency.
link |
And, uh, ultimately I would like to build a system that it is worth for you to wait
link |
five minutes because it gives you the answer that you're willing to wait for
link |
So latency is a very unpleasant constraint under which to operate.
link |
And also there is actually one more thing, which is tricky about robots.
link |
Uh, there is actually, uh, no, uh, not much data.
link |
So the data that I'm speaking about would be a data of, uh, first person
link |
experience from the robot and like a gigabytes of data like that, if we would
link |
have gigabytes of data like that, of robots solving various problems, it would
link |
be very easy to make a progress on robotics.
link |
And you can see that in case of text or code, there is a lot of data, like a
link |
first person perspective, they don't writing code.
link |
Yeah. So you had this, you mentioned this really interesting idea that if you were
link |
to build like a successful robotics company, so open as mission is much
link |
bigger than robotics, this is one of the, one of the things you've worked on, but
link |
if it was a robotics company, they, you wouldn't so quickly dismiss supervised
link |
learning, uh, correct that you would build a robot that, uh, was perhaps what
link |
like, um, an empty shell, like dumb, and they would operate under teleoperation.
link |
So you would invest, that's just one way to do it, invest in human supervision,
link |
like direct human control of the robots as it's learning and over time, add
link |
more and more automation.
link |
So let's say that's how I would build a robotics company today.
link |
If I would be building a robotics company, which is, you know, spend 10
link |
million dollars or so recording human trajectories, controlling a robot.
link |
After you find a thing that the robot should be doing, that there's a market
link |
fit for, like you can make a lot of money with that product.
link |
Uh, so I would record data and then I would essentially train supervised
link |
learning model on it.
link |
That might be the path today.
link |
I think that actually what is needed is to have a robot that can
link |
train powerful models over video.
link |
So, um, you have seen maybe a models that can generate images like Dali and people
link |
are looking into models, generating videos, they're like, uh, bodies,
link |
algorithmic questions, even how to do it.
link |
And it's unclear if there is enough compute for this purpose, but, uh, I, I
link |
suspect that the models that which would have a level of understanding of video,
link |
same as GPT has a level of understanding of text, could be used, uh, to train
link |
robots to solve tasks.
link |
They would have a lot of common sense.
link |
If one day, I'm pretty sure one day there will be a robotics company by robotics
link |
company, I mean, the primary source of income is, is from robots that is worth
link |
What do you think that company will do?
link |
I think self driving cars.
link |
No, it's interesting.
link |
Cause my mind went to personal robotics, robots in the home.
link |
It seems like there's much more market opportunity there.
link |
I think it's very difficult to achieve.
link |
I mean, this, this, this might speak to something important, which is I understand
link |
self driving much better than understand robotics in the home.
link |
So I understand how difficult it is to actually solve self driving to a, to a
link |
level, not just the actual computer vision and the control problem and just the
link |
basic problem of self driving, but creating a product that would undeniably
link |
be, um, that will cost less money.
link |
Like it will save you a lot of money, like orders of magnitude, less money
link |
that could replace Uber drivers, for example.
link |
So car sharing that's autonomous, that creates a similar or better
link |
experience in terms of how quickly you get from A to B or just whatever, the
link |
pleasantness of the experience, the efficiency of the experience, the value
link |
of the experience, and at the same time, the car itself costs cheaper.
link |
I think that's very difficult to achieve.
link |
I think there's a lot more, um, low hanging fruit in the home.
link |
That, that, that could be, I also want to give you a perspective on like how
link |
challenging it would be at home or like it maybe kind of depends on that exact
link |
problem that you'd be solving.
link |
Like if we're speaking about these robotic arms and hands, these things,
link |
they cost tens of thousands of dollars or maybe a hundred K and, um, you know,
link |
maybe, obviously, maybe there would be economy of scale.
link |
These things would be cheaper, but actually for any household to buy it,
link |
the price would have to go down to maybe a thousand bucks.
link |
I personally think that, uh, so self driving car, it provides a clear service.
link |
I don't think robots in the home, there'll be a trillion dollar company
link |
will just be all about service, meaning it will not necessarily be about like
link |
a robotic arm that's helps you.
link |
I don't know, open a bottle or wash the dishes or, uh, any of that kind of stuff.
link |
It has to be able to take care of that whole, the therapist thing.
link |
You mentioned, I think that's, um, of course there's a line between what
link |
is a robot and what is not like, does it really need a body?
link |
But you know, some, um, uh, AI system with some embodiment, I think.
link |
So the tricky part when you think actually what's the difficult part is,
link |
um, when the robot has like, when there is a diversity of the environment
link |
with which the robot has to interact, that becomes hard.
link |
So, you know, on the one spectrum, you have, uh, industrial robots as they
link |
are doing over and over the same thing, it is possible to some extent to
link |
prescribe the movements and we've very small amount of intelligence, the, the
link |
movement can be repeated millions of times.
link |
Um, the, it, there are also, you know, various pieces of industrial robots
link |
where it becomes harder and harder.
link |
I can, for instance, in case of Tesla, it might be a matter of putting a, a
link |
rack inside of a car and, you know, because the rack kind of moves around,
link |
it's, uh, it's not that easy.
link |
It's not exactly the same every time.
link |
That's not being the case that you need actually humans to do it.
link |
Uh, while, you know, welding cars together, it's a very repetitive process.
link |
Um, then in case of self driving itself, uh, that difficulty has to do with the
link |
diversity of the environment, but still the car itself, um, the problem
link |
that they are solving is you try to avoid even interacting with things.
link |
You are not touching anything around because touching itself is hard.
link |
And then if you would have in the home, uh, robot that, you know, has to
link |
touch things and like if these things, they change the shape, if there is a huge
link |
variety of things to be touched, then that's difficult.
link |
If you are speaking about the robot, which there is, you know, head that
link |
is smiling in some way with cameras that either doesn't, you know, touch things.
link |
That's relatively simple.
link |
Okay. So to both agree and to push back.
link |
So you're referring to touch, like soft robotics, like the actual touch, but.
link |
I would argue that you could formulate just basic interaction between, um, like
link |
non contact interaction is also a kind of touch and that might be very difficult
link |
to solve that's the basic, this not disagreement, but that's the basic open
link |
question to me with self driving cars and this agreement with Elon, which
link |
is how much interaction is required to solve self driving cars.
link |
How much touch is required?
link |
You said that in your intuition, touch is not required.
link |
And my intuition to create a product that's compelling to use, you're going
link |
to have to, uh, interact with pedestrians, not just avoid pedestrians,
link |
but interact with them when we drive around.
link |
In major cities, we're constantly threatening everybody's life with
link |
our movements, um, and that's how they respect us.
link |
There's a game to ready going out with pedestrians and I'm afraid you can't
link |
just formulate autonomous driving as a collision avoidance problem.
link |
So I think it goes beyond like a collision avoidance is the
link |
first order approximation.
link |
Uh, but then at least in case of Tesla, you can't just
link |
at least in case of Tesla, they are gathering data from people driving their
link |
cars and I believe that's an example of supervised data that they can train
link |
their models, uh, on, and they are doing it, uh, which, you know, can give
link |
a model dislike, uh, another level of, uh, of, uh, behavior that is needed
link |
to actually interact with the real world.
link |
It's interesting how much data is required to achieve that.
link |
Um, w what do you think of the whole Tesla autopilot approach, the computer
link |
vision based approach with multiple cameras and there's a data engine.
link |
It's a multitask, multiheaded neural network, and it's this fascinating
link |
process of, uh, similar to what you're talking about with the robotics
link |
approach, uh, which is, you know, you deploy in your own network and
link |
then there's humans that use it and then it runs into trouble in a bunch
link |
of places and that stuff is sent back.
link |
So like the deployment discovers a bunch of edge cases and those edge
link |
cases are sent back for supervised annotation, thereby improving the
link |
neural network and that's deployed again.
link |
It goes over and over until the network becomes really good at the task of
link |
driving becomes safer and safer.
link |
What do you think of that kind of approach to robotics?
link |
I believe that's the way to go.
link |
So in some sense, even when I was speaking about, you know, collecting
link |
trajectories from humans, that's like a first step and then you deploy
link |
the system and then you have humans revising the, all the issues.
link |
And in some sense, like at this approach converges to system that doesn't make
link |
mistakes because for the cases where there are mistakes, you got their
link |
data, how to fix them and the system will keep on improving.
link |
So there's a very, to me, difficult question of how hard that, you know,
link |
how long that converging takes, how hard it is.
link |
The other aspect of autonomous vehicles, this probably applies to certain
link |
robotics applications is society, right?
link |
They put as, as the quality of the system converges.
link |
So one, there's a human factors perspective of psychology of humans being
link |
able to supervise those even with teleoperation, those robots.
link |
And the other is society willing to accept robots.
link |
Currently society is much harsher on self driving cars than it is on human
link |
driven cars in terms of the expectation of safety.
link |
So the bar is set much higher than for humans.
link |
And so if there's a death in an autonomous vehicle, that's seen as a much more,
link |
much more dramatic than a death in the human driven vehicle.
link |
Part of the success of deployment of robots is figuring out how to make robots
link |
part of society, both on the, just the human side, on the media side, on the
link |
media journalist side, and also on the policy government side.
link |
And that seems to be, maybe you can put that into the objective function to
link |
optimize, but that is, that is definitely a tricky one.
link |
And I wonder if that is actually the trickiest part for self driving cars or
link |
any system that's safety critical.
link |
It's not the algorithm, it's the society accepting it.
link |
Yeah, I would say, I believe that the part of the process of deployment is actually
link |
showing people that the given things can be trusted and, you know, trust is also
link |
like a glass that is actually really easy to crack it and damage it.
link |
And I think that's actually very common with, with innovation, that there's
link |
some resistance toward it and it's just the natural progression.
link |
So in some sense, people will have to keep on proving that indeed these
link |
systems are worth being used.
link |
And I would say, I also found out that often the best way to convince people
link |
is by letting them experience it.
link |
That's the case with Tesla autopilot, for example, that's the case with, yeah,
link |
with basically robots in general.
link |
It's kind of funny to hear people talk about robots.
link |
Like there's a lot of fear, even with like legged robots, but when they
link |
actually interact with them, there's joy.
link |
I love interacting with them.
link |
And the same with the car, with a robot, if it starts being useful, I think
link |
people immediately understand.
link |
And if the product is designed well, they fall in love.
link |
It's actually even similar when I'm thinking about the car.
link |
It's actually even similar when I'm thinking about Copilot, the GitHub Copilot.
link |
There was a spectrum of responses that people had.
link |
And ultimately the important piece was to let people try it out.
link |
And then many people just loved it.
link |
Especially like programmers.
link |
Yeah, programmers, but like some of them, you know, they came with a fear.
link |
But then you try it out and you think, actually, that's cool.
link |
And, you know, you can try to resist the same way as, you know, you could
link |
resist moving from punch cards to, let's say, C++ or so.
link |
And it's a little bit futile.
link |
So we talked about generation of program, generation of language, even
link |
self supervised learning in the visual space for robotics and then
link |
reinforcement learning.
link |
What do you, in like this whole beautiful spectrum of AI, do you think is a
link |
good benchmark, a good test to strive for to achieve intelligence?
link |
That's a strong test of intelligence.
link |
You know, it started with Alan Turing and the Turing test.
link |
Maybe you think natural language conversation is a good test.
link |
So, you know, it would be nice if, for instance, machine would be able to
link |
solve Riemann hypothesis in math.
link |
That would be, I think that would be very impressive.
link |
So theorem proving, is that to you, proving theorems is a good, oh, oh,
link |
like one thing that the machine did, you would say, damn.
link |
That would be quite, quite impressive.
link |
I mean, the tricky part about the benchmarks is, you know, as we are
link |
getting closer with them, we have to invent new benchmarks.
link |
There is actually no ultimate benchmark out there.
link |
See, my thought with the Riemann hypothesis would be the moment the
link |
machine proves it, we would say, okay, well then the problem was easy.
link |
That's what happens.
link |
And I mean, in some sense, that's actually what happens over the years
link |
in AI that like, we get used to things very quickly.
link |
You know something, I talked to Rodney Brooks.
link |
I don't know if you know who that is.
link |
He called AlphaZero homework problem.
link |
Cause he was saying like, there's nothing special about it.
link |
It's not a big leap.
link |
And I didn't, well, he's coming from one of the aspects that we referred
link |
to is he was part of the founding of iRobot, which deployed now tens
link |
of millions of robot in the home.
link |
So if you see robots that are actually in the homes of people as the
link |
legitimate instantiation of artificial intelligence, then yes, maybe an AI
link |
that plays a silly game like go and chess is not a real accomplishment,
link |
but to me it's a fundamental leap.
link |
But I think we as humans then say, okay, well then that that game of
link |
chess or go wasn't that difficult compared to the thing that's currently
link |
So my intuition is that from perspective of the evolution of these AI
link |
systems will at first seen the tremendous progress in digital space.
link |
And the, you know, the main thing about digital space is also that you
link |
can, everything is that there is a lot of recorded data.
link |
Plus you can very rapidly deploy things to billions of people.
link |
While in case of a physical space, the deployment part takes multiple
link |
You have to manufacture things and, you know, delivering it to actual
link |
people, it's very hard.
link |
So I'm expecting that the first and that prices in digital space of
link |
goods, they would go, you know, down to the, let's say marginal costs
link |
And also the question is how much of our life will be in digital because
link |
it seems like we're heading towards more and more of our lives being in
link |
the digital space.
link |
So like innovation in the physical space might become less and less
link |
Like why do you need to drive anywhere if most of your life is spent in
link |
I still would like, you know, to at least at the moment, my impression
link |
is that I would like to have a physical contact with other people.
link |
And that's very important to me.
link |
We don't have a way to replicate it in the computer.
link |
It might be the case that over the time it will change.
link |
Like in 10 years from now, why not have like an arbitrary infinite number
link |
of people you can interact with?
link |
Some of them are real, some are not with arbitrary characteristics that
link |
you can define based on your own preferences.
link |
I think that's maybe where we are heading and maybe I'm resisting the
link |
Yeah, I'm telling you, if I got to choose, if I could live in Elder
link |
Scrolls Skyrim versus the real world, I'm not so sure I would stay with
link |
Yeah, I mean, the question is, so will VR be sufficient to get us there
link |
or do you need to, you know, plug electrodes in the brain?
link |
And it would be nice if these electrodes wouldn't be invasive.
link |
Or at least like provably non destructive.
link |
But in the digital space, do you think we'll be able to solve the
link |
Turing test, the spirit of the Turing test, which is, do you think we'll
link |
be able to achieve compelling natural language conversation between
link |
people, like have friends that are AI systems on the internet?
link |
I totally think it's doable.
link |
Do you think the current approach of GPT will take us there?
link |
So there is, you know, the part of at first learning all the content
link |
out there and I think that Steel System should keep on learning as
link |
it speaks with you.
link |
Yeah, and I think that should work.
link |
The question is how exactly to do it.
link |
And, you know, obviously we have people at OpenAI asking these
link |
questions and kind of at first pre training on all existing content
link |
is like a backbone and is a decent backbone.
link |
Do you think AI needs a body connecting to our robotics question to
link |
truly connect with humans or can most of the connection be in the
link |
So let's see, we know that there are people who met each other online
link |
and they fell in love.
link |
So it seems that it's conceivable to establish connection, which is
link |
purely through internet.
link |
Of course, it might be more compelling the more modalities you add.
link |
So it would be like you're proposing like a Tinder, but for AI, you
link |
like swipe right and left and half the systems are AI and the other is
link |
humans and you don't know which is which.
link |
That would be our formulation of Turing test.
link |
The moment AI is able to achieve more swipe right or left, whatever,
link |
the moment it's able to be more attractive than other humans, it
link |
passes the Turing test.
link |
Then you would pass the Turing test in attractiveness.
link |
Well, no, like attractiveness just to clarify.
link |
There will be conversation.
link |
It's also attractiveness with wit and humor and whatever makes
link |
conversation is pleasant for humans.
link |
So you're saying it's possible to achieve in the digital space.
link |
In some sense, I would almost ask that question.
link |
Why wouldn't that be possible?
link |
Well, I have this argument with my dad all the time.
link |
He thinks that touch and smell are really important.
link |
So they can be very important.
link |
And I'm saying the initial systems, they won't have it.
link |
Still, there are people being born without these senses and I believe
link |
that they can still fall in love and have meaningful life.
link |
I wonder if it's possible to go close to all the way by just training
link |
on transcripts of conversations.
link |
I wonder how far that takes us.
link |
So I think that actually still you want images like I would like.
link |
So I don't have kids, but like I could imagine having AI Tutor.
link |
It has to see, you know, kids drawing some pictures on the paper.
link |
And also facial expressions, all that kind of stuff.
link |
We use dogs and humans use their eyes to communicate with each other.
link |
I think that's a really powerful mechanism of communication.
link |
Body language too, that words are much lower bandwidth.
link |
And for body language, we still, you know, we kind of have a system
link |
that displays an image of its or facial expression on the computer.
link |
Doesn't have to move, you know, mechanical pieces or so.
link |
So I think that, you know, that there is like kind of a progression.
link |
You can imagine that text might be the simplest to tackle.
link |
But this is not a complete human experience at all.
link |
You expand it to, let's say images, both for input and output.
link |
And what you describe is actually the final, I guess, frontier.
link |
What makes us human, the fact that we can touch each other or smell or so.
link |
And it's the hardest from perspective of data and deployment.
link |
And I believe that these things might happen gradually.
link |
Are you excited by that possibility?
link |
This particular application of human to AI friendship and interaction?
link |
Like would you, do you look forward to a world?
link |
You said you're living with a few folks and you're very close friends with them.
link |
Do you look forward to a day where one or two of those friends are AI systems?
link |
So if the system would be truly wishing me well, rather than being in the situation
link |
that it optimizes for my time to interact with the system.
link |
The line between those is, it's a gray area.
link |
I think that's the distinction between love and possession.
link |
And these things, they might be often correlated for humans, but you might find that there are
link |
some friends with whom you haven't spoke for months.
link |
And then you pick up the phone, it's as the time hasn't passed.
link |
They are not holding to you.
link |
And I will, I wouldn't like to have AI system that, you know, it's trying to convince me
link |
to spend time with it.
link |
I would like the system to optimize for what I care about and help me in achieving my own goals.
link |
But there's some, I mean, I don't know, there's some manipulation, there's some possessiveness,
link |
there's some insecurities, this fragility, all those things are necessary to form a close
link |
friendship over time, to go through some dark shit together, some bliss and happiness together.
link |
I feel like there's a lot of greedy self centered behavior within that process.
link |
My intuition, but I might be wrong, is that human computer interaction doesn't have to
link |
go through a computer being greedy, possessive, and so on.
link |
It is possible to train systems, maybe, that they actually
link |
they are, I guess, prompted or fine tuned or so to truly optimize for what you care about.
link |
And you could imagine that, you know, the way how the process would look like is at
link |
some point, we as humans, we look at the transcript of the conversation or like an entire
link |
interaction and we say, actually here, there was more loving way to go about it.
link |
And we supervise system toward being more loving, or maybe we train the system such
link |
that it has a reward function toward being more loving.
link |
Or maybe the possibility of the system being an asshole and manipulative and possessive
link |
every once in a while is a feature, not a bug.
link |
Because some of the happiness that we experience when two souls meet each other, when two humans
link |
meet each other, is a kind of break from the assholes in the world.
link |
And so you need assholes in AI as well, because, like, it'll be like a breath of fresh air
link |
to discover an AI that the three previous AIs you had are too friendly or no, or cruel
link |
It's like some kind of mix.
link |
And then this one is just right, but you need to experience the full spectrum.
link |
Like, I think you need to be able to engineer assholes.
link |
Because there's some level to us being appreciated to appreciate the human experience.
link |
We need the dark and the light.
link |
So that kind of reminds me.
link |
I met a while ago at the meditation retreat, one woman, and she told me, you know,
link |
beautiful, beautiful woman, and she had a she had a crutch.
link |
She had the trouble walking on one leg.
link |
I asked her what has happened.
link |
And she said that five years ago she was in Maui, Hawaii, and she was eating a salad and
link |
some snail fell into the salad.
link |
And apparently there are neurotoxic snails over there.
link |
And she got into coma for a year.
link |
And apparently there is, you know, high chance of even just dying.
link |
But she was in the coma.
link |
At some point, she regained partially consciousness.
link |
She was able to hear people in the room.
link |
People behave as she wouldn't be there.
link |
You know, at some point she started being able to speak, but she was mumbling like a
link |
barely able to express herself.
link |
Then at some point she got into wheelchair.
link |
Then at some point she actually noticed that she can move her toe and then she knew that
link |
she will be able to walk.
link |
And then, you know, that's where she was five years after.
link |
And she said that since then she appreciates the fact that she can move her toe.
link |
And I was thinking, hmm, do I need to go through such experience to appreciate that I have
link |
I can move my toe?
link |
Wow, that's a really good story and really deep example.
link |
And in some sense, it might be the case that we don't see light if we haven't went through
link |
But I wouldn't say that we should.
link |
We shouldn't assume that that's the case, which it may be able to engineer shortcuts.
link |
Ilya had this, you know, belief that maybe one has to go for a week or six months to
link |
do some challenging camp to just experience, you know, a lot of difficulties and then comes
link |
back and actually everything is bright, everything is beautiful.
link |
I'm with Ilya on this.
link |
It must be a Russian thing.
link |
Where are you from originally?
link |
I'm tempted to say that explains a lot.
link |
But yeah, there's something about the Russian, the necessity of suffering.
link |
I believe suffering or rather struggle is necessary.
link |
I believe that struggle is necessary.
link |
I mean, in some sense, you even look at the story of any superhero in the movie.
link |
It's not that it was like everything goes easy, easy, easy, easy.
link |
I like how that's your ground truth is the story of superheroes.
link |
You mentioned that you used to do research at night and go to bed at like 6 a.m.
link |
I still do that often.
link |
What sleep schedules have you tried to make for a productive and happy life?
link |
Like, is there is there some interesting wild sleeping patterns that you engaged that you
link |
found that works really well for you?
link |
I tried at some point decreasing number of hours of sleep like a gradually like a half
link |
an hour every few days to this.
link |
You know, I was hoping to just save time.
link |
That clearly didn't work for me.
link |
Like at some point, there's like a phase shift and I felt tired all the time.
link |
You know, there was a time that I used to work during the nights.
link |
The nice thing about the nights is that no one disturbs you.
link |
And even I remember when I was meeting for the first time with Greg Brockman, his
link |
CTO and chairman of OpenAI, our meeting was scheduled to 5 p.m.
link |
And I overstepped for the meeting.
link |
Over slept for the meeting at 5 p.m.
link |
Yeah, now you sound like me.
link |
And at the moment, in some sense, my sleeping schedule also has to do with the fact that
link |
I'm interacting with people.
link |
I sleep without an alarm.
link |
So, yeah, the the team thing you mentioned, the extrovert thing, because most humans operate
link |
during a certain set of hours, you're forced to then operate at the same set of hours.
link |
But I'm not quite there yet.
link |
I found a lot of joy, just like you said, working through the night because it's quiet
link |
because the world doesn't disturb you.
link |
And there's some aspect counter to everything you're saying.
link |
There's some joyful aspect to sleeping through the mess of the day because people are having
link |
people are having meetings and sending emails and there's drama meetings.
link |
I can sleep through all the meetings.
link |
You know, I have meetings every day and they prevent me from having sufficient amount of
link |
time for focused work.
link |
And then I modified my calendar and I said that I'm out of office Wednesday, Thursday
link |
and Friday every day and I'm having meetings only Monday and Tuesday.
link |
And that busty positively influenced my mood that I have literally like at three days for
link |
fully focused work.
link |
So there's better solutions to this problem than staying awake all night.
link |
OK, you've been part of development of some of the greatest ideas in artificial intelligence.
link |
What would you say is your process for developing good novel ideas?
link |
You have to be aware that clearly there are many other brilliant people around.
link |
So you have to ask yourself a question, why the given idea, let's say, wasn't tried by
link |
someone else and in some sense, it has to do with, you know, kind of simple.
link |
It might sound simple, but like a thinking outside of the box.
link |
And what do I mean here?
link |
So, for instance, for a while, people in academia, they assumed that you have a feeling that
link |
you have a fixed data set and then you optimize the algorithms in order to get the best performance.
link |
And that was so in great assumption that no one thought about training models on
link |
anti internet or like that.
link |
Maybe some people thought about it, but it felt to many as unfair.
link |
And in some sense, that's almost like a it's not my idea or so, but that's an example of
link |
breaking at the typical assumption.
link |
So you want to be in the paradigm that you're breaking at the typical assumption.
link |
In the context of the community, getting to pick your data set is cheating.
link |
And in some sense, so that was that was assumption that many people had out there.
link |
And then if you free yourself from assumptions, then they are likely to achieve something
link |
that others cannot do.
link |
And in some sense, if you are trying to do exactly the same things as others, it's very
link |
likely that you're going to have the same results.
link |
Yeah, I but there's also that kind of tension, which is asking yourself the question, why
link |
haven't others done this?
link |
Because, I mean, I get a lot of good ideas, but I think probably most of them suck when
link |
they meet reality.
link |
So so actually, I think the other big piece is getting into habit of generating ideas,
link |
training your brain towards generating ideas and not even suspending judgment of the ideas.
link |
So in some sense, I noticed myself that even if I'm in the process of generating ideas,
link |
if I tell myself, oh, that was a bad idea, then that actually interrupts the process
link |
and I cannot generate more ideas because I'm actually focused on the negative part, why
link |
But I created also environment in the way that it's very easy for me to store new ideas.
link |
So, for instance, next to my bed, I have a voice recorder and it happens to me often
link |
like I wake up during the night and I have some idea.
link |
In the past, I was writing them down on my phone, but that means, you know, turning on
link |
the screen and that wakes me up or like pulling a paper, which requires, you know, turning
link |
These days, I just start recording it.
link |
What do you think, I don't know if you know who Jim Keller is.
link |
I know Jim Keller.
link |
He's a big proponent of thinking harder on a problem right before sleep so that he can
link |
sleep through it and solve it in his sleep or like come up with radical stuff in his
link |
sleep that's trying to get me to do this.
link |
So it happened from my experience perspective, it happened to me many times during the high
link |
school days when I was doing mathematics that I had a solution to my problem as I woke up.
link |
At the moment, regarding thinking hard about the given problem is I'm trying to actually
link |
devote substantial amount of time to think about important problems, not just before
link |
I'm organizing amount of the huge chunks of time such that I'm not constantly working
link |
on the urgent problems, but I actually have time to think about the important one.
link |
So you do it naturally.
link |
But his idea is that you kind of prime your brain to make sure that that's the focus.
link |
Oftentimes people have other worries in their life that's not fundamentally deep problems
link |
like I don't know, just stupid drama in your life and even at work, all that kind of stuff.
link |
He wants to kind of pick the most important problem that you're thinking about and go
link |
I think that's wise.
link |
I mean, the other thing that comes to my mind is also I feel the most fresh in the morning.
link |
So during the morning, I try to work on the most important things rather than just being
link |
pulled by urgent things or checking email or so.
link |
What do you do with the...
link |
Because I've been doing the voice recorder thing too, but I end up recording so many
link |
messages it's hard to organize.
link |
I have the same problem.
link |
Now I have heard that Google Pixel is really good in transcribing text and I might get
link |
a Google Pixel just for the sake of transcribing text.
link |
Yeah, people listening to this, if you have a good voice recorder suggestion that transcribe,
link |
please let me know.
link |
Some of it has to do with OpenAI codecs too.
link |
Like some of it is simply like the friction.
link |
I need apps that remove that friction between voice and the organization of the resulting
link |
transcripts and all that kind of stuff.
link |
But yes, you're right.
link |
Absolutely, like during, for me it's walking, sleep too, but walking and running, especially
link |
running, get a lot of thoughts during running and there's no good mechanism for recording
link |
So one more thing that I do, I have a separate phone which has no apps.
link |
Maybe it has like audible or let's say Kindle.
link |
No one has this phone number, this kind of my meditation phone.
link |
And I try to expand the amount of time that that's the phone that I'm having.
link |
It has also Google Maps if I need to go somewhere and I also use this phone to write down ideas.
link |
Ah, that's a really good idea.
link |
That's a really good idea.
link |
Often actually what I end up doing is even sending a message from that phone to the other
link |
So that's actually my way of recording messages or I just put them into notes.
link |
What advice would you give to a young person, high school, college, about how to be successful?
link |
You've done a lot of incredible things in the past decade, so maybe, maybe have some.
link |
There's something, there might be something.
link |
There might be something.
link |
I mean, might sound like a simplistic or so, but I would say literally just follow your
link |
passion, double down on it.
link |
And if you don't know what's your passion, just figure out what could be a, what could
link |
So that might be an exploration.
link |
When I was in elementary school was math and chemistry.
link |
And I remember for some time I gave up on math because my school teacher, she told me
link |
And I guess maybe an advice would be just ignore people if they tell you that you're
link |
You're dumb. You mentioned something offline about chemistry and explosives.
link |
What was that about?
link |
So a story goes like that.
link |
I got into chemistry.
link |
Maybe I was like a second grade of my elementary school, third grade.
link |
I started going to chemistry classes.
link |
I really love building stuff.
link |
And I did all the experiments that they describe in the book, like, you know, how to create
link |
oxygen with vinegar and baking soda or so.
link |
So I did all the experiments and at some point I was, you know, so what's next?
link |
And explosives, they also, it's like a, you have a clear reward signal, you know, if the
link |
thing worked or not.
link |
So I remember at first I got interested in producing hydrogen.
link |
That was kind of funny experiment from school.
link |
You can just burn it.
link |
And then I moved to nitroglycerin.
link |
So that's also relatively easy to synthesize.
link |
I started producing essentially dynamite and detonating it with a friend.
link |
I remember there was a, you know, there was at first like maybe two attempts that I went
link |
with a friend to detonate what we built and it didn't work out.
link |
And like a third time he was like, ah, it won't work.
link |
Like, let's don't waste time.
link |
And, you know, we were, I was carrying this, this, you know, that tube with dynamite, I
link |
don't know, pound or so, dynamite in my backpack, we're like riding on the bike to the edges
link |
Yeah, and attempt number three, this was be attempt number three.
link |
Attempt number three.
link |
And now we dig a hole to put it inside.
link |
It actually had the, you know, electrical detonator.
link |
We draw a cable behind the tree.
link |
I even, I never had, I haven't ever seen like a explosion before.
link |
So I thought that there would be a lot of, you know, a lot of, you know, a lot of, you
link |
know, there will be a lot of sound.
link |
But, you know, we're like laying down and I'm holding the cable and the battery.
link |
At some point, you know, we kind of like a three to one and I just connected it and it
link |
felt like the ground shake.
link |
It was like more like a sound.
link |
And then the soil started kind of lifting up and started falling on us.
link |
And then, you know, the friend said, let's make sure the next time we have helmets.
link |
But also, you know, I'm happy that nothing happened to me.
link |
It could have been the case that I lost the limbo or so.
link |
Yeah, but that's childhood of an engineering mind with a strong reward signal of an
link |
My there's some aspect of chemists that the chemists I know, like my dad with plasma
link |
chemistry, plasma physics, he was very much into explosives, too.
link |
It's a worrying quality of people that work in chemistry that they love.
link |
I think it is that exactly is the strong signal that the thing worked.
link |
There is no doubt.
link |
There's some magic.
link |
It's almost like a reminder that physics works, that chemistry works.
link |
It's almost like a little glimpse at nature that you yourself engineer.
link |
I that's why I really like artificial intelligence, especially robotics, is you create a little
link |
piece of nature and in some sense, even for me with explosives, the motivation was creation
link |
rather than destruction.
link |
In terms of advice, I forgot to ask about just machine learning and deep learning for
link |
people who are specifically interested in machine learning, how would you recommend
link |
they get into the field?
link |
So I would say re implement everything and also there is plenty of courses.
link |
So like from scratch?
link |
So on different levels of abstraction in some sense, but I would say re implement something
link |
from scratch, re implement something from a paper, re implement something, you know,
link |
from podcasts that you have heard about.
link |
I would say that's a powerful way to understand things.
link |
So it's often the case that you read the description and you think you understand, but you truly
link |
understand once you build it, then you actually know what really matter in the description.
link |
Is there a particular topics that you find people just fall in love with?
link |
I tend to really enjoy reinforcement learning because it's much more, it's much easier
link |
to get to a point where you feel like you created something special, like fun games
link |
kind of things that are rewarding.
link |
As opposed to like re implementing from scratch, more like supervised learning kind of things.
link |
So, you know, if someone would optimize for things to be rewarding, then it feels that
link |
the things that are somewhat generative, they have such a property.
link |
So you have, for instance, adversarial networks, or do you have just even generative language
link |
And you can even see, internally, we have seen this thing with our releases.
link |
So we have, we released recently two models.
link |
There is one model called Dali that generates images, and there is other model called Clip
link |
that actually you provide various possibilities, what could be the answer to what is on the
link |
picture, and it can tell you which one is the most likely.
link |
And in some sense, in case of the first one, Dali, it is very easy for you to understand
link |
that actually there is magic going on.
link |
And in the case of the second one, even though it is insanely powerful, and you know, people
link |
from a vision community, they, as they started probing it inside, they actually understood
link |
How far it goes, it's difficult for a person at first to see how well it works.
link |
And that's the same, as you said, that in case of supervised learning models, you might
link |
not kind of see, or it's not that easy for you to understand the strength.
link |
Even though you don't believe in magic, to see the magic.
link |
To see the magic, yeah.
link |
It's a generative.
link |
That's really brilliant.
link |
So anything that's generative, because then you are at the core of the creation.
link |
You get to experience creation without much effort.
link |
Unless you have to do it from scratch, but.
link |
And it feels that, you know, humans are wired.
link |
There is some level of reward for creating stuff.
link |
Of course, different people have a different weight on this reward.
link |
In the big objective function.
link |
In the big objective function of a person.
link |
You wrote that beautiful is what you intensely pay attention to.
link |
Even a cockroach is beautiful.
link |
If you look very closely, can you expand on this?
link |
So what I'm, I wrote here actually corresponds to my subjective experience that I had through
link |
extended periods of meditation.
link |
It's, it's pretty crazy that at some point the meditation gets you to the place that
link |
you have really increased focus, increased attention.
link |
Increased attention.
link |
And then you look at the very simple objects that were all the time around you can look
link |
at the table or on the pen or at the nature.
link |
And you notice more and more details and it becomes very pleasant to look at it.
link |
And it, once again, it kind of reminds me of my childhood.
link |
Like I just pure joy of being.
link |
It's also, I have seen even the reverse effect that by default, regardless of what we possess,
link |
we very quickly get used to it.
link |
And you know, you can have a very beautiful house and if you don't put sufficient effort,
link |
you're just going to get used to it and it doesn't bring any more joy,
link |
regardless of what you have.
link |
Well, I actually, I find that material possessions get in the way of that experience of pure
link |
So I've always, I've been very fortunate to just find joy in simple things.
link |
Just, just like you're saying, just like, I don't know, objects in my life, just stupid
link |
objects like this cup, like thing, you know, just objects sounds okay.
link |
I'm not being eloquent, but literally objects in the world, they're just full of joy.
link |
Cause it's like, I can't believe when I can't believe that I'm fortunate enough to be alive
link |
to experience these objects.
link |
And then two, I can't believe humans are clever enough to have built these objects.
link |
The hierarchy of pleasure that that provides is infinite.
link |
I mean, even if you look at the cup of water, so, you know, you see first like a level of
link |
like a reflection of light, but then you think, you know, man, there's like a trillions upon
link |
of trillions of particles bouncing against each other.
link |
There is also the tension on the surface that, you know, if the back, back could like a stand
link |
on it and move around.
link |
And you think it also has this like a magical property that as you decrease temperature,
link |
it actually expands in volume, which allows for the, you know, legs to freeze on the,
link |
on the surface and at the bottom to have actually not freeze, which allows for life like a crazy.
link |
You look in detail at some object and you think actually, you know, this table, that
link |
was just a figment of someone's imagination at some point.
link |
And then there was like a thousands of people involved to actually manufacture it and put
link |
And by default, no one cares.
link |
And then you can start thinking about evolution, how it all started from single cell organisms
link |
that led to this table.
link |
And these thoughts, they give me life appreciation and even lack of thoughts, just the pure raw
link |
signal also gives the life appreciation.
link |
See, the thing is, and then that's coupled for me with the sadness that the whole ride
link |
ends and perhaps is deeply coupled in that the fact that this experience, this moment
link |
ends, gives it, gives it an intensity that I'm not sure I would otherwise have.
link |
So in that same way, I tried to meditate on my own death.
link |
Do you think about your mortality?
link |
Are you afraid of death?
link |
So fear of death is like one of the most fundamental fears that each of us has.
link |
We might be not even aware of it.
link |
It requires to look inside, to even recognize that it's out there and there is still, let's
link |
say, this property of nature that if things would last forever, then they would be also
link |
The fact that the things change in some way gives any meaning to them.
link |
I also, you know, found out that it seems to be very healing to people to have these
link |
short experiences, like, I guess, psychedelic experiences in which they experience death
link |
of self in which they let go of this fear and then maybe can even increase the intensity
link |
can even increase the appreciation of the moment.
link |
It seems that many people, they can easily comprehend the fact that the money is finite
link |
while they don't see that time is finite.
link |
I have this like a discussion with Ilya from time to time.
link |
He's like, you know, man, like the life will pass very fast.
link |
At some point I will be 40, 50, 60, 70 and then it's over.
link |
This is true, which also makes me believe that, you know, that every single moment it
link |
is so unique that should be appreciated.
link |
And this also makes me think that I should be acting on my life because otherwise it
link |
I also like this framework of thinking from Jeff Bezos on regret minimization that like
link |
I would like if I will be at that deathbed to look back on my life and not regret that
link |
I haven't done something.
link |
It's usually you might regret that you haven't tried.
link |
I'm fine with failing.
link |
What's the Nietzsche eternal occurrence?
link |
Try to live a life that if you had to live it infinitely many times, that would be the
link |
you'd be okay with that kind of life.
link |
So try to live it optimally.
link |
I can say that it's almost like I'm.
link |
I'm available to me where I am in my life.
link |
I'm extremely grateful for actually people whom I met.
link |
I would say I think that I'm decently smart and so on.
link |
But I think that actually to a great extent where I am has to do with the people who I
link |
Would you be okay if after this conversation you died?
link |
So if I'm dead, then it kind of I don't have a choice anymore.
link |
So in some sense, there's like plenty of things that I would like to try out in my life.
link |
I feel that I'm gradually going one by one and I'm just doing them.
link |
I think that the list will be always infinite.
link |
Yeah, so might as well go today.
link |
Yeah, I mean, to be clear, I'm not looking forward to die.
link |
I would say if there is no choice, I would accept it.
link |
But like in some sense, I'm if there would be a choice, if there would be a possibility
link |
to leave, I would fight for leaving.
link |
I find it's more honest and real to think about, you know, dying today at the end of
link |
That seems to me, at least to my brain, more honest slap in the face as opposed to I still
link |
have 10 years like today, then I'm much more about appreciating the cup and the table and
link |
so on and less about like silly worldly accomplishments and all those kinds of things.
link |
But we have in the company a person who say at some point found out that they have cancer
link |
and that also gives, you know, huge perspective with respect to what matters now.
link |
And, you know, often people in situations like that, they conclude that actually what
link |
matters is human connection.
link |
And love and that's people conclude also if you have kids, kids as family.
link |
You, I think, tweeted, we don't assign the minus infinity reward to our death.
link |
Such a reward would prevent us from taking any risk.
link |
We wouldn't be able to cross the road in fear of being hit by a car.
link |
So in the objective function, you mentioned fear of death might be fundamental to the
link |
So, as I said, let's assume that they're like a reward functions in our brain.
link |
And the interesting thing is even realization, how different reward functions can play with
link |
As a matter of fact, I wouldn't say that you should assign infinite negative reward to
link |
anything because that messes up the math.
link |
The math doesn't work out.
link |
It doesn't work out.
link |
And as you said, even, you know, government or some insurance companies, you said they
link |
assign $9 million to human life.
link |
And I'm just saying it with respect to, that might be a hard statement to ourselves, but
link |
in some sense that there is a finite value of our own life.
link |
I'm trying to put it from perspective of being less, of being more egoless and realizing
link |
fragility of my own life.
link |
And in some sense, the fear of death might prevent you from acting because anything can
link |
And I'm sure actually, if you were to put death in the objective function, there's probably
link |
so many aspects to death and fear of death and realization of death and mortality.
link |
There's just whole components of finiteness of not just your life, but every experience
link |
and so on that you're going to have to formalize mathematically.
link |
And also, you know, that might lead to you spending a lot of compute cycles on this like
link |
a deliberating this terrible future instead of experiencing now.
link |
And then in some sense, it's also kind of unpleasant simulation to run in your head.
link |
Do you think there's an objective function that describes the entirety of human life?
link |
So, you know, usually the way you ask that is what is the meaning of life?
link |
Is there a universal objective functions that captures the why of life?
link |
So, yeah, I mean, I suspect that they will ask this question, but it's also a question
link |
that I ask myself many, many times.
link |
See, I can tell you a framework that I have these days to think about this question.
link |
So I think that fundamentally, meaning of life has to do with some of our reward actions
link |
that we have in brain and they might have to do with, let's say, for instance, curiosity
link |
or human connection, which might mean understanding others.
link |
It's also possible for a person to slightly modify their reward function.
link |
Usually they mostly stay fixed, but it's possible to modify reward function and you can pretty
link |
So in some sense, the reward functions, optimizing reward functions, they will give you a life
link |
Is there some randomness in the function?
link |
I think when you are born, there is some randomness.
link |
You can see that some people, for instance, they care more about building stuff.
link |
Some people care more about caring for others.
link |
Some people, there are all sorts of default reward functions.
link |
And then in some sense, you can ask yourself, what is the satisfying way for you to go after
link |
this reward function?
link |
And you just go after this reward function.
link |
And, you know, some people also ask, are you satisfied with your life?
link |
And, you know, some people also ask, are these reward functions real?
link |
I almost think about it as, let's say, if you would have to discover mathematics, in
link |
mathematics, you are likely to run into various objects like complex numbers or differentiation,
link |
some other objects.
link |
And these are very natural objects that arise.
link |
And similarly, the reward functions that we are having in our brain, they are somewhat
link |
very natural, that, you know, there is a reward function for understanding, like a comprehension,
link |
curiosity, and so on.
link |
So in some sense, they are in the same way natural as their natural objects in mathematics.
link |
So, you know, there's the old sort of debate, is mathematics invented or discovered?
link |
You're saying reward functions are discovered.
link |
So nature provided some, you can still, let's say, expand it throughout the life.
link |
Some of the reward functions, they might be futile.
link |
Like, for instance, there might be a reward function, maximize amount of wealth.
link |
And this is more like a learned reward function.
link |
But we know also that some reward functions, if you optimize them, you won't be quite satisfied.
link |
Well, I don't know which part of your reward function resulted in you coming today, but
link |
I am deeply appreciative that you did spend your valuable time with me.
link |
Wojtek is really fun talking to you.
link |
You're a good human being.
link |
And it's an honor to meet you and an honor to talk to you.
link |
Thanks for talking today, brother.
link |
Thank you, Lex a lot.
link |
I appreciated your questions, curiosity.
link |
I had a lot of time being here.
link |
Thanks for listening to this conversation with Wojtek Zaremba.
link |
To support this podcast, please check out our sponsors in the description.
link |
And now, let me leave you with some words from Arthur C. Clarke, who is the author of
link |
2001 A Space Odyssey.
link |
It may be that our role on this planet is not to worship God, but to create him.
link |
Thank you for listening, and I hope to see you next time.