Lex Fridman PodcastMatt Botvinick: Neuroscience, Psychology, and AI at DeepMind | Lex Fridman Podcast #106
EVERY SPOKEN WORD
150 min read · 30,030 words- 0:00 – 3:29
Introduction
- LFLex Fridman
The following is a conversation with Matt Botvinick, director of neuroscience research at DeepMind. He's a brilliant cross-disciplinary mind, navigating effortlessly between cognitive psychology, computational neuroscience, and artificial intelligence. Quick summary of the ads. Two sponsors, The Jordan Harbinger Show and Magic Spoon Cereal. Please consider supporting the podcast by going to jordanharbinger.com/lex and also going to magicspoon.com/lex and using code LEX at checkout after you buy all of their cereal. Click the links, buy the stuff. It's the best way to support this podcast and the journey I'm on. If you enjoy this podcast, subscribe on YouTube, review it with five stars on Apple Podcast, follow on Spotify, support on Patreon, or connect with me on Twitter @lexfridman, spelled surprisingly without the E, just F-R-I-D-M-A-N. As usual, I'll do a few minutes of ads now and never any ads in the middle that can break the flow of the conversation. This episode is supported by The Jordan Harbinger Show. Go to jordanharbinger.com/lex. It's how he knows I sent you. On that page, subscribe to his podcast on Apple Podcasts, Spotify, and you know where to look. I've been binging on his podcast. Jordan is a great interviewer and even a better human being. I recently listened to his conversation with Jack Barsky, former sleeper agent for the KGB in the '80s and author of Deep Undercover, which is a memoir that paints yet another interesting perspective on the Cold War era. I've been reading a lot about the Stalin and then Gorbachev and Putin eras of Russia, but this conversation made me realize that I need to do a deep dive into the Cold War era to get a complete picture of Russia's recent history. Again, go to jordanharbinger.com/lex, subscribe to his podcast. It's how he knows I sent you. It's awesome. You won't regret it. This episode is also supported by Magic Spoon, low-carb, keto-friendly, super amazingly delicious cereal. I've been on a keto or very low-carb diet for a long time now. It helps with my mental performance, it helps with my physical performance, even doing this crazy pushup, uh, pull-up challenge I'm doing, including the running. It just feels great. Uh, I used to love cereal. Obviously, I can't have it, uh, now because most cereals have a crazy amount of sugar, which is terrible for you. So I quit it years ago, but Magic Spoon amazingly somehow is a totally different thing. Zero sugar, 11 grams of protein, and only three net grams of carbs. It tastes delicious. It has a lot of flavors, two new ones, including peanut butter. But if you know what's good for you, you'll go with cocoa, my favorite flavor and the flavor of champions. Click the magicspoon.com/lex link in the description and use code LEX at checkout for free shipping and to let them know I sent you. They've agreed to sponsor this podcast for a long time. They're an amazing sponsor and an even better cereal. I highly recommend it. It's delicious. It's good for you. You won't regret it. And now here's my conversation with Matt Botvinick.
- 3:29 – 14:26
How much of the brain do we understand?
- LFLex Fridman
How much of the human brain do you think we understand?
- MBMatt Botvinick
I think we're at a, a weird moment in the history of neuroscience in the sense that, um, there's a, there... I feel like we understand a lot about the brain at a very high level, but a very, very coarse level.
- LFLex Fridman
When you say high level, what are you thinking? Are you thinking functional?
- MBMatt Botvinick
Yeah. Yeah.
- LFLex Fridman
Are you thinking structurally?
- MBMatt Botvinick
So in other words, what is, what is the brain for? You know, what, what, what kinds of computation does the brain do? Um, you know, what kinds of, uh, behaviors would we have to, uh, would we have to explain if we were going to look down at the mechanistic level? Um, and at that level, I feel like we understand much, much more about the brain than we did when I was in, in high school, but what, but it's, it's at a very, it's almost like we're seeing it through a fog. It's only at a very coarse level. We don't really understand what the, the neuronal mechanisms are that underlie these computations. We've gotten better at saying, you know, what are the functions that the brain is computing that we would have to understand, you know, if we were going to get down to the neuronal level? And at the other end of the spectrum, we, w- w- you know, in the last few years, incredible progress has been made in terms of, um, technologies that allow us to see, you know, actually literally see in some cases what's going on at the, at the, um, single unit level, even the dendritic level, and then there's this yawning gap in between.
- LFLex Fridman
Oh, that's interesting. So at the high level, so that's almost like cognitive science level?
- MBMatt Botvinick
Yeah, yeah.
- LFLex Fridman
And then at the neuronal level, that's neurobiology and neuroscience-
- MBMatt Botvinick
Yeah.
- LFLex Fridman
... just studying single neurons, the, the, the, the, the synaptic connections and all the dopamine-
- MBMatt Botvinick
Yeah.
- LFLex Fridman
... all the kind of neurotransmitters.
- MBMatt Botvinick
One blanket statement I should probably make is that, uh, as I've gotten older, I have become more and more reluctant to make a distinction between psychology and neuroscience. To me, the point of neuroscience is to study what the brain is for. If you, if you, if you're, if you're a nephrologist and you want to learn about the kidney, w- you start by a- by saying, "What is this thing for?" Well, um, it seems to be for taking, uh, blood on one side that has, uh, metabolites in it that are, that shouldn't be there, uh, sucking them out of the blood-... while leaving the good stuff behind and then excreting that in the form of urine. That's what the kidney is for. (laughs) It's, like, obvious. Um, so the rest of the work is deciding how it does that. And this, it seems to me, is the right approach to take to the brain. You say, "Well, what is the brain for?" The brain, as far as I can tell, is for producing behavior. It's from going... it's, it's for going from perceptual inputs to behavioral outputs. Um, and the behavioral outputs should be adaptive. Uh, so that's what psychology is about. It's about understanding the structure of that function and then the rest of neuroscience is about figuring out how those operations are actually carried out at a, at a mech- mechanistic level.
- LFLex Fridman
That's really interesting, but... So unlike the kidney, the, the brain, the, the gap between the electrical signal and behavior... So you truly see neuroscience as the science of, that, that touches behavior, how the brain generates behavior or how the brain converts raw visual information into understanding? Like, under- like, you, you basically see cognitive science, psychology, and neuroscience as all one science.
- MBMatt Botvinick
Yeah. That's a, that's a-
- LFLex Fridman
Is that a-
- MBMatt Botvinick
... it's a personal statement. I don't mean to-
- LFLex Fridman
Is that a hopeful... Is that, is that a hopeful or a realistic statement? So certainly you will be correct in your feeling in some number of years, but that number of years could be 200, 300 years from now.
- MBMatt Botvinick
Oh, well, well, there's a...
- LFLex Fridman
Is that aspirational or is that-
- MBMatt Botvinick
There's-
- LFLex Fridman
... a pragmatic engineering, uh, feeling that you have?
- MBMatt Botvinick
It's, it's both in the sense that this is what I hope and expect will, uh, bear fruit, um, over the coming decades. But it's also pragmatic in the sense that I'm not sure what we're doing in either, in either psychology or neuroscience if that's not the framing. I don't, I don't un- I don't know what it means to understand the brain if there's no... if, if part of the enterprise is not about understanding the behavior that's being produced.
- LFLex Fridman
I mean, yeah, but I would, I would've compare it to maybe astronomers looking at the movement of the s- the, the planets and the stars and without any interest of the underlying physics, right? And I would argue that there, at least in the early days, there is some value to just tracing the movement of the planets and the stars without thinking about the physics too much because it's such a big leap to start thinking about the physics before you even understand even the basic structural elements of-
- MBMatt Botvinick
Oh, I agree with that. I agree, so-
- LFLex Fridman
But you're saying in the end, the goal should be-
- MBMatt Botvinick
Yeah.
- LFLex Fridman
... to deeply understand.
- MBMatt Botvinick
Well, right. And I, I think... So I thought about this a lot when I was in grad school, because a lot of what I studied in grad school was psychology and s- I found myself a little bit confused about what it meant to... It seems like what, what we were talking about a lot of the time were virtual causal mechanisms. Like, uh, "Oh, well, you know, attentional selection then selects some object in the environment and that is then passed on to the motor... you know, information about that is passed on to the motor system."
- 14:26 – 22:53
Psychology
- MBMatt Botvinick
random.
- LFLex Fridman
Now, y- you've kind of talked about this, uh, bridging of the gap between psychology and neuroscience, but do you think it's possible... Like, my love is, um, like I fell in love with psychology and psychiatry in general with Freud when I was really young.
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
And I ho- hoped to understand the mind. And for me, understanding the mind, at least at that young age before I discovered AI and even neuroscience, was to, um... is psychology. And do you think it's possible to understand the mind without getting into all of the messy details of, uh, neuroscience? Like you kind of mentioned to you it's appealing to try to understand the mechanisms at the lowest level, but do you think that's needed, that's required to understand how the mind works?
- MBMatt Botvinick
That's an important part of the whole picture. But, uh, I would be the last person on Earth to suggest that that reality renders psychology in its own right, um, unproductive. I trained as a psychologist. I've, I am fond of saying that I have learned much more from psychology than I have from neuroscience. To me, psychology is a hugely important discipline. And, and one thing that warms my heart is that ways of, ways of investigating behavior that have been native to cognitive psychology since its, you know, dawn in the '60s are starting to become... they're, they're starting to become interesting to AI researchers for a variety of reasons. And that's been exciting for me to see.
- LFLex Fridman
Can you maybe talk a little bit about what's, what you see as, um, beautiful aspects of psychology? Maybe limiting aspects of psychology? I mean, maybe just start it off as, as a science, as a field. To me, it was when I understood what psychology is, analytical psychology, like the way it's actually carried out, it was really disappointing to see, uh, two aspects. One is how few, how small the end is. How many s- how small the number of subject is in the studies. And two, it was disappointing to see how controlled the entire... how, how much it was in the lab. How... it wasn't studying humans in the wild. There was no mechanism for studying humans in the wild. So that's where I became a little bit disillusioned-
- MBMatt Botvinick
Mm-hmm. Mm-hmm.
- LFLex Fridman
... to, uh, psychology. And then the modern world of the internet is so exciting to me. The Twitter data or YouTube da- like, da- data of human behavior on the internet becomes exciting, because then the end grows and then in the wild grows. But that's just my narrow sense. Like, do you have a s- optimistic or pessimistic cynical view of psychology? How do you see the field broadly?
- MBMatt Botvinick
When I was in graduate school, it was early enough that there was still a thrill in seeing that there were ways of doing, there were ways of doing experimental science, um, that provided insight to the structure of the mind. One thing that impressed me most when I was at that stage in my education was neuro-psychology. Looking at, looking at the... analyzing the behavior of populations who had, um, brain damage of different kinds and trying to understand what, what the, what the specific deficits were that arose from a lesion in a particular part of the brain. And the, the kind of experimentation that was done and that's still being done to get answers in that context w- was so creative and it was so, um...... deliberate, you know. Uh, the, it, it was good science. Y- y- uh, an experiment answered one question but raised another, and somebody would do an experiment that answered that question, and you really felt like you were narrowing in on some kind of approximate understanding of what this part of the brain was for.
- LFLex Fridman
Do you have an example from memory of, uh, what kind of aspects of the mind could be studied in this kind of way?
- MBMatt Botvinick
Oh, sure. I mean, the very detailed neuropsychological studies of language, um, language function, looking at production and reception and the relationship between, uh, you know, visual function, uh, you know, ha- reading and auditory and semantic. And I mean, there were these bea- and still are, these beautiful models that came out of that kind of research that really made you feel like you understood something that you hadn't underst- stood before about how, you know, language processing is organized in the brain-
- LFLex Fridman
Yeah.
- MBMatt Botvinick
... but having said all that, um, uh, you know, I, I think, y- I think you are... I mean, I agree with you that the cost of doing highly controlled experiments is that you, by construction, miss out on the richness and complexity of the real world. One thing that... So, I, I, I was drawn into science by what in those days was called connectionism, which is of course the, you know, what we now call deep learning, um, and at that point in history, neural networks were m- primarily being used in order to model human cognition. Um, they weren't yet really useful for industrial applications.
- LFLex Fridman
So, you always found neural networks in biological form beautiful?
- MBMatt Botvinick
Oh, neural networks were very concretely the thing that drew me into science. I was handed... Are you familiar with the, the PDP books from, from the '80s?
- LFLex Fridman
Yeah. Mm-hmm.
- MBMatt Botvinick
Some- when I was in... I went to medical school before I went into science, and, uh-
- LFLex Fridman
Oh, really?
- MBMatt Botvinick
Yeah.
- LFLex Fridman
Interesting. Wow.
- MBMatt Botvinick
I also (laughs) I also did a graduate degree in art history, so I'm-
- LFLex Fridman
(laughs)
- MBMatt Botvinick
... I kind of explored, um-
- LFLex Fridman
Well, art history I understand.
- MBMatt Botvinick
(laughs)
- LFLex Fridman
That's, that's, that's just a curious, uh, creative mind, but medical school with a dream of what, if we take that slight tangent? Uh-
- MBMatt Botvinick
Yeah.
- LFLex Fridman
... what, did y- did you want to be a surgeon?
- MBMatt Botvinick
I actually was quite interested in surgery. I wa- I was interested in surgery and psychiatry, and I thought that must be, I must be the only person, uh, on the planet who had... who was torn between those two fields. And I, I, I, I said exactly that to my advisor in medical school who, who turned out, uh, I found out later to be a famous, uh, psychoanalyst. And, and he said to me, "No, no, it's actually not so uncommon to be interested in surgery and psychiatry." And he, he, he conjectured that the reason that people develop these, these two interests is that both fields are about going beneath the surface and kind of getting into the kind of secret-
- LFLex Fridman
Yeah.
- 22:53 – 32:23
The paradox of the human brain
- LFLex Fridman
I apologize for the romanticized question-
- MBMatt Botvinick
(laughs)
- LFLex Fridman
... but what, uh, what idea in the space of neuroscience, in the space of the human brain is, to you, the most beautiful, mysterious, surprising?
- MBMatt Botvinick
What, what had always fascinated me, um, even when I was a pretty young kid, I think, uh, was the, um, the, the paradox that lies in the fact that, uh, the brain is so mysterious and so... and seems so distant. Uh, um, but at the same time, it's responsible for the, the, the, the full transparency of everyday life (laughs) . It's-
- LFLex Fridman
Yeah.
- MBMatt Botvinick
The brain is literally what makes everything obvious and, uh, familiar (laughs) . And, and, um, and, and there's always one in the room with you (laughs) .
- LFLex Fridman
Yeah.
- MBMatt Botvinick
Uh, I, I, I used to teach... Wh- when I taught at Princeton, I used to teach a, a cognitive neuroscience course, and, uh, the very last thing I would say to the students was, you know, "People often... When people think of, uh, scientific inspiration, the, the metaphor is often, 'Well look to the stars,' you know? 'The stars will inspire you to wonder at the universe and, and, uh, you know, think about your place in it and how things work.'" And, and I'm all for looking at the stars, but I- I've always been much more inspired and, and kind of my sense of wonder comes from the...... not from the distant, mysterious stars, but from the extremely intimately close brain.
- LFLex Fridman
Yeah.
- MBMatt Botvinick
There's something just endlessly fascinating to me about that.
- LFLex Fridman
The, like, just like you said, the, the, the one, it's close and yet distant in, in, in terms of our understanding of it. Do you, are you also captivated by the, the fact that this very conversation is happening because two brains are communicating each other?
- MBMatt Botvinick
Yes (laughs) . Exactly.
- LFLex Fridman
So the, the, the, I guess what I mean is the subjective nature of the experience, if we can take a small tangent into the, the mystical of it-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... the consciousness-
- MBMatt Botvinick
Mm.
- LFLex Fridman
... or, or when you're saying you're captivated by the idea of the brain, you're, are you talking about specifically the mechanism of cognition or are you also just, like, at least for me, it, the, it's almost like paralyzing the beauty and the mystery of the fact that it creates the entirety of the experience, not just the reasoning capability, but the experience.
- MBMatt Botvinick
Yes. Well, I, I definitely resonate with that, that latter thought. And I, I often find discussions of artificial intelligence to be, um, disappointingly narrow.
- LFLex Fridman
Mm-hmm.
- MBMatt Botvinick
Uh, you know, speaking as someone who has always had an interest in, in, in art-
- LFLex Fridman
Right.
- MBMatt Botvinick
... you know, the-
- LFLex Fridman
I was just gonna go there 'cause it sounds like somebody who has an interest in art (laughs) .
- MBMatt Botvinick
Yeah, I mean, uh, I, there, there, there, there are many layers to, you know, to full bore human experience. And, and, um, in some ways it's not enough to say, "Oh, well, don't worry, you know, we, w- w- we're talking about cognition, but we'll add emotion," you know?
- LFLex Fridman
Yeah.
- MBMatt Botvinick
There, there's, there's, there's an incredible scope to, um, what humans go through in, in every moment. And, um, uh, and yes, so i- i- it's, that's part of what fascinates me is that, um, is that our brains are producing that, uh, but at the same time it's so mysterious to us how-
- LFLex Fridman
Yeah.
- MBMatt Botvinick
... you know? (laughs) Like, we literally, like our brains are literally in our heads producing this experience-
- LFLex Fridman
Producing the experience.
- MBMatt Botvinick
... and yet they're, th- and yet there's, ther- it's so mysterious to us. And so, and, and the scientific challenge of getting at the, the, the actual explanation for that is so overwhelming. There's so, that's just, I don't know. That e- uh, uh, certain people have fixations on particular questions, and that's always, that's just always been mine.
- 32:23 – 39:34
Cognition is a function of the environment
- MBMatt Botvinick
you know, if, if you take an introductory computer science course and they are introducing you to the notion of, uh, Turing machines, w- w- one way of, uh, articulating w- what the significance of a Turing machine is, is that it, it's a machine emulator. Uh, it, it can emulate any other machine. Um, and that, that to me, you know, that, that, and of... (sighs) That way of looking at a, at a Turing machine, you know, really sticks with me. I think of humans as maybe sharing in some of that, um, character. We're capacity limited, we're not Turing machines obviously, but we have the ability to adapt behaviors that are, uh, very much unlike anything we've done before, but there's some basic mechanism that's implemented in our brain that allows us to run, run software.
- LFLex Fridman
But just on that point, you mentioned Turing machine, but nevertheless it's fundamentally our brains are just computational devices in your view? Is that what you're getting at? Like, is... (laughs) I, I, it was a little bit unclear to this line you drew.
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
Is, is, is there any magic in there or is it just basic computation?
- MBMatt Botvinick
I'm happy to think of it as just basic computation.
- LFLex Fridman
(laughs)
- MBMatt Botvinick
But mind you, I won't be satisfied until somebody explains to me how, what the basic computations are that are leading to the full richness of human cognition, uh-
- LFLex Fridman
Yes.
- MBMatt Botvinick
I mean, it's not gonna be enough for, for me to, you know, understand what the computations are that allow people to, you know, do arithmetic or play chess. I want, I want the whole, the whole, you know, the whole thing.
- LFLex Fridman
And a, and a small tangent because you kind of mentioned coronavirus, the, there's group behavior.
- MBMatt Botvinick
Oh, sure.
- LFLex Fridman
Uh, I, is that, is there something interesting to your search of understanding the human mind, uh, where l- behavior of large groups or just behavior of groups is interesting? You, you know, seeing that as a collective mind, as a collective intelligence? Perhaps seeing the groups of people as a single intelligent organism, especially looking at the reinforcement learning work-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... you've done recently?
- MBMatt Botvinick
Well, yeah. I can't, I can't... I mean, I, I, I have the, I have the, um, the honor of working with a lot of incredibly smart people s- and I wouldn't want to take any credit for-
- LFLex Fridman
(laughs)
- MBMatt Botvinick
... for leading the way on the, the multi-agent work that's come out of, out of my group or DeepMind lately, but, um, I do find it fascinating. And I mean, I think they are... you know, I think, uh, it's, it, it can't be debated, you know? (laughs) The human behavior, uh, arises within communities. That just seems to me self-evident. Um-
- LFLex Fridman
But, but to me, so it, it is self-evident, but that seems to be a profound aspects of something that created... That was like, if you look at, like, 2001: Space Odyssey when the, when the monkeys to- touched the-
- MBMatt Botvinick
Yeah.
- LFLex Fridman
Uh, like, that's the magical moment I think Yuval Harari argues that the b- the ability of our... a large numbers of humans to hold an idea, to converge towards idea together, like you said-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... shaking hands versus bumping elbows-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... somehow converge, like, without even, like, like, without, you know, without being in a room all together, just kind of this, like, distributed convergence towards an idea-
- MBMatt Botvinick
Yeah.
- LFLex Fridman
... over a particular period of time seems to be fundamental to, to, uh, just every aspect of our cognition, of our intelligence because humans, we'll talk about reward-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... but it seems like we don't really have a clear objective function under which we operate, but we all kind of converge (laughs) towards one somehow. And that, that to me has always been a mystery, uh, that, uh, I think is somehow productive for also understanding AI systems. But I, I guess, I guess that's the next step. The first step is try to understand the mind.
- MBMatt Botvinick
Well, I, I don't know. I mean, um, I think there's something to the argument that, uh, that kind of bottom, like, strictly bottom-up approach is, uh, wrongheaded. In other words, you know, there are...... there are basic phenomena that, you know, uh, you know, basic aspects of human intelligence that are, you know, can only be understood in, in the context of groups. I, I'm perfectly open to that. I've never been particularly convinced by the notion that we should be s- we should consider intelligence to inhere at the level of communities. I, I, I don't know why, I just, I'm sort of stuck on the notion that the basic unit that we want to understand is individual humans. And if, if we have to understand that in the context of other humans, fine. Um, but for me, intelligence is just, I'm stubbornly, I stubbornly define it as something that is, you know, an aspect of an individual human. That's just my-
- LFLex Fridman
But that's-
- 39:34 – 53:27
Prefrontal cortex
- LFLex Fridman
context. If we can step back into another sort of beautiful world which is the actual mechanics, the dirty mess of it, of the human brain, is, is there something for people who might not know, uh, is there something you can comment on or describe the key parts of the brain that are important for intelligence or just in general what are the different parts of the brain that you're curious about, that you've studied and, uh, that are just good to know about when you're thinking about cognition?
- MBMatt Botvinick
Well, m- my, my area of expertise, if I have one, is, um, prefrontal cortex. So, uh, you know-
- LFLex Fridman
What's that?
- MBMatt Botvinick
(laughs)
- LFLex Fridman
Or do we-
- MBMatt Botvinick
It depends on who you ask. The, the, the, the, uh, the, the technical definition is, has ha- is anatomical, it, there are, there are, um, uh, parts of your brain that are responsible for motor behavior and they're very easy to identify. Um, and, um, the region of your cerebral cortex, the out, you know, the sort of outer crust of your brain that lies in front of those is defined as the prefrontal cortex.
- LFLex Fridman
And when you say an- anatomical, sorry to interrupt, so that's referring to, uh, sort of the geographic region?
- MBMatt Botvinick
Yeah.
- LFLex Fridman
As opposed to some kind of functional definition.
- MBMatt Botvinick
Exactly. So that, it, this is kind of the coward's way out.
- LFLex Fridman
(laughs)
- MBMatt Botvinick
You know, I'm telling you what the prefrontal cortex is, just in terms of, like, what part of the real estate it occupies.
- LFLex Fridman
It's the thing in the front of the brain.
- MBMatt Botvinick
Yeah, exactly. And, and in fact, um, the early history of, uh, you know, the, of neuroscientific, um, investigation of what this, like, front part of the brain does is sort of funny to read because, uh, you know, it was really, it was really World War I that started people down this road of trying to figure out what different parts of the brain, the human brain do, in the sense that, uh, there were a lot of people with brain damage, who came back from the war with brain damage, and it, that provided, as tragic as that was, it provided an opportunity for scientists to try to identify the functions of different brain regions, and it wa- that was an, actually incredibly productive. But one of the frustrations that neuropsychologists faced was they couldn't really identify exactly what the deficit was, um, that arose from damage to this, these most, you know, kind of frontal parts of the brain. It was just a very difficult thing to, you know, to, um, you know, to pin down. There were a couple of neuropsychologists who identified, uh, through, through a large amount of clinical experience and close observation, they started to, um, put their finger on a syndrome that was associated with frontal damage. Actually, one of them was a, a Russian, uh, neuropsychologist named Luria, who, you know, students of cognitive psychology still read. Um, and, and what he started to-... figure out was that the frontal cortex was somehow involved in flexibility. The, i- i- in, in, in guiding behaviors that required someone to override a habit, uh, or to do something unusual, or to change what they were doing in every flexible way from one moment to another.
- LFLex Fridman
So it focused on, l- like, new experiences? So the, um, so the way your brain processes and acts w- in new experiences?
- MBMatt Botvinick
Yeah. What later helped bring this function into better focus was a distinction between controlled and automatic behavior. Or, or to, in, in, in other literatures, this is referred to as habitual behavior versus goal-directed behavior. So it's very, very clear that the human brain has pathways that are dedicated to habits, to things that you do all the time and they need to be automatized so that they don't require you to concentrate too much so th- that leaves your cognitive capacity free to do other things. Um, just think about wh- the difference between, uh, wh- wh- driving when you're learning to drive versus driving after you're fairly expert. Uh, there are brain pathways that slowly absorb those frequently, uh, performed behaviors so that they can be habits, so that they can be automatic.
- LFLex Fridman
So that, that's kind of, like, the purest form of learning, I guess, is happening there, which is why, uh, I mean, this is kind of jumping ahead, which is why that perhaps is the most useful for us to focusing on and trying to see how artificial intelligence systems can learn. Is that the way you-
- MBMatt Botvinick
It's interesting. I, I, I do think about this distinction between controlled and automatic or goal-directed and, and habitual behavior a lot in thinking about where we are in AI research. Um, but, um, but just to finish, fi- to finish the, the kind of dissertation here, the, the, the role of the front, of the prefrontal cortex is generally understood these days sort of in, in contradistinction to that habitual domain. In other words, the prefrontal cortex is what helps you override those habits. It alle-
- LFLex Fridman
(laughs) .
- MBMatt Botvinick
It's what allows you to say, "Whoa, whoa. What I usually do in this situation is X but given the context, I probably should do Y." I mean, th- the elbow bump is a great example, right? If, uh, you know, reaching out and shaking hands is a, probably a habitual behavior and it's the prefrontal cortex that allows us to bear in mind that there's something unusual going on right now and in this situation, I need to not do the usual thing, uh, th- the kind of behaviors that Luria reported and, you know, he built tests for, you know, detecting these kinds of things, were exactly like this. So in other words, when I stuck out my hand, I want you instead to present your elbow. A, a patient with frontal damage would have a great deal of trouble with that. You know, if somebody proffering their hand would elicit, you know, a handshake. Um, the prefrontal cortex is what allows us to say, "Hold on, hold on. That's the usual thing but I'm, I have the ability to r- bear in mind even very unusual contexts and to reason about what behavior is appropriate there."
- LFLex Fridman
Y- just to get a sense, is, uh, are us humans special in the presence of the prefrontal cortex? Uh, do mice have a prefrontal cortex? Do other mammals that we can study? I- uh, if y- if know, then how do they integrate new experiences?
- MBMatt Botvinick
Yeah. That's a, that's a really tricky question and a very timely question because we have, um, revolutionary, uh, new technologies for, um, m- monitoring, measuring, and also causally influencing neural behavior in mice and fruit flies. And these techniques are not, um, fully available even for studying, um, brain function in, in monkeys, uh, let alone humans. Um, and so it's a, it's a very, sort of, for me at least, a very urgent question whether the kinds of things that we want to understand about human intelligence can be pursued in these other organisms. And, um, you know, to put it briefly, there's disagreement.
- LFLex Fridman
(laughs) .
- MBMatt Botvinick
Um, th- you know, y- people who study fruit flies will often tell you, "Hey, fruit flies are smarter than you think." Uh, and they'll point to experiments where fruit flies were able to learn new behaviors, were able to generalize, f- um, from one stimulus to another in a way that suggests that they have abstractions that guide their generalization. Um, uh, th- I've had many conversations in which, uh, I will start by observing, m- you know, m- recounting some, some o- observation a- about mouse behavior where it seemed like mice were taking an awfully long time to learn a task that for a human would be profoundly trivial and, um, I will conclude from that that mice really don't have the cognitive flexibility that we want to explain. And then a mouse researcher will say to me, "Well, you know, hold on. That experiment may not have worked because you asked a mouse to deal with"... stimuli and behaviors that were very unnatural for the mouse. If instead you kept the logic of the experiment the same but put, you know, kind of put it in a, you know, uh, uh, presented it, the information in a way that aligns with what mice are used to dealing with in their natural habitats, you might find that a mouse actually has more intelligence than you think. Um, and then they'll go on to show you videos of mice doing things in their natural habitat which seem strikingly intelligent, you know, dealing with, you know, physical problems, you know, "I have to drag this piece of food back to my, uh, you know, back to my lair, but there's something in my way, and how do I get rid of that thing?" So I think, I think these are open questions, to put it, you know, to sum that up.
- LFLex Fridman
A- and, and then taking a small step back, so related to that, is you kinda mentioned, or taken a little shortcut by saying it's a geographic, geographic part of the, uh, the prefrontal cortex is a region of the brain, but if we, uh... What's your sense, in a bigger philosophical view, prefrontal cortex and the brain in general, do you have a sense that it's a set of subsystems in the way we've kind of implied-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... that are, that are pretty distinct? Or, uh, to what degree is it that, or to what degree is it a giant interconnected mess where everything kinda does everything and it's impossible to disentangle them?
- MBMatt Botvinick
Mm-hmm. I think there's overwhelming evidence that there's functional differentiation, that it's clearly not the case that all parts of the brain are doing the same thing. This follows immediately from the kinds of studies of brain damage that we were chatting about before. It, it's obvious from what you see if you stick an electrode in the brain and measure what's going on at the level of u- neural activity. Um, having said that, uh, there are two other things to add, which kind of, I don't know, maybe tug in the other direction. One is that, um, it's when you look carefully at functional differentiation in the brain, what you usually end up concluding, uh, at least this is my observation of the literature, is that th- the differences between regions are, are graded rather than being discrete. So it doesn't seem like it's easy to divide the brain up into, uh, uh, true modules, um, where, you know, that are, are, you know, that have clear boundaries and that have, you know, um, like, uh, you know, uh, d- like, cha- clear channels of communication between them. Instead-
- LFLex Fridman
And this applies to the prefrontal cortex?
- MBMatt Botvinick
Yeah. Oh, yeah.
- 53:27 – 1:00:11
Information processing in the brain
- LFLex Fridman
that, on that topic, you kinda mentioned information. Just to get a sense, I imagine something that there's still mystery and disagreement on is how does the brain carry information and signal? Like, what in your sense is, uh, the basic mechanism of, uh, communication in the brain?
- MBMatt Botvinick
Well, I, I, I guess I'm old-fashioned in that I consider, um, the networks that we use in deep learning research to be a reasonable approximation to, uh, you know, w- the, the mechanisms that carry information in the brain. So the, the, the usual way of articulating that is to say what really matters is a rate code. It, it, what matters is how, how, uh, how quickly is a, an individual neuron spiking? How f- you know, what's the frequency at which it's spiking? Is it fir-
- LFLex Fridman
So the timing of the spiking.
- MBMatt Botvinick
Yeah. Is it, is it firing fast or slow? Let's, you know, let's put a number on that, and that number is enough to capture what, what neurons are doing. There's, you know, there's still uncertainty about whether that's an, uh, an adequate, um, description of how information is, uh-... uh, is transmitted within the brain. There, you know, there, there are studies that suggest that the precise timing of spikes matters. There are, um, studies that suggest that there are computations that go on within the dendritic tree, within a neuron, that, um, are quite rich and structured and that really don't equate to anything that we're doing in our artificial neural networks. Having said that, I feel like we can get... I feel like, I feel like we're getting somewhere by, um, sticking to this high level of abstraction.
- LFLex Fridman
Just the rate and, by the way, we're talking about the electrical signal, is that... I remember reading some vague paper somewhere recently where the mechanical signal, like the vibrations or something of the, of, of the neurons also communicates information.
- MBMatt Botvinick
I haven't seen that, but, um...
- LFLex Fridman
The, the... So there, somebody was arguing that the, uh, the electrical signal, this is in a Nature paper or something like that, where the electrical signal is actually a side effect of the mechanical signal. But I don't think that changes the story-
- MBMatt Botvinick
(laughs)
- LFLex Fridman
... but, but it's almost an interesting idea that there could be a deeper... It's like a... it's always like, uh, in physics with quantum mechanics, there's always a deeper story that could be underlying the whole thing. But you think it's basically the rate of spiking that gets us... that's like the lowest hanging fruit that can get us really far.
- MBMatt Botvinick
This is a, this is a classical view. I mean, this is, this is, this is not... The only way in which this stance would be controversial is i- i- you know, in the sense that there are, there are members of the neuroscience community who are interested in alternatives, but this is really a very mainstream view. The way that neurons communicate is that, uh, neurotransmitters arrive, uh, you know, at, at a, at, uh, you know, they, they wash up on a neuron. Uh, the neuron has receptors for those transmitters. The, the, the, the, the meeting of the transmitter with these receptors changes the voltage of the neuron, and if enough voltage change occurs, then a spike occurs, right? One-
- LFLex Fridman
Right.
- MBMatt Botvinick
... of these, like discrete events. And it's that spike that gets conducted down the axon and leads to neurotransmitter release. This is just, this is just like Neuroscience 101.
- LFLex Fridman
Yeah.
- MBMatt Botvinick
This is like the way the brain is supposed to work. Now, what we do when we build artificial neural networks of the kind that are now popular in the AI community, um, is that we don't worry about those individual spikes. We just worry about the frequency at which those spikes are being, um, generated. And the, you know, we can sit... You know, people talk about that as the activity-
- LFLex Fridman
Mm-hmm.
- MBMatt Botvinick
... of a neuron. And, you know, so the, the activity of units in a deep learning system is, you know, broadly analogous to the spike rate of a neuron. There, there are people who, who believe that there are other forms of communication in the brain. In fact, I've been involved in some research recently that suggests that, um, the, the voltage, the voltage fluctuations that occur, uh, in populations of neurons that aren't, um, you know, that, that are sort of below the level of, um, of spike production may be important for, for communication. But I'm still pretty old school in the sense that I think that the, the things that we're building in AI research constitute reasonable, um, models of how a brain would work.
- LFLex Fridman
Let me ask just for fun a crazy question, uh, 'cause I can.
- MBMatt Botvinick
Yeah.
- LFLex Fridman
Do, do you think it's possible we're completely wrong about the way, uh, this basic mechanism of neur- neuronal communication, that the information is stored as some very different kind of way in the brain?
- MBMatt Botvinick
Oh, heck yes. Uh, I mean, I wouldn't... Look, I wouldn't be a scientist if I didn't think there was any chance we were wrong. But, but I mean, if you look, if you look at the history of, um, deep learning research as, as it's been applied to neuroscience... Of course, the vast majority of deep learning research these days isn't about neuroscience, but, um, you know, if you go back to the 1980s, there's a, you know, sort of an unbroken chain of, of research in, in which a particular strategy is taken, which is, hey, let's train, uh, a deep, uh, a deep learning system. Let's train a, a, a, a multi-layered neural network on this task that we, uh, trained our, you know, rat on or our monkey on or this human being on. And then let's look at what the units deep in the system are doing, and let's ask whether what they're doing resembles what we know about what neurons deep in the brain are doing. And over and over and over and over, that strategy works in the sense that, um, the learning algorithms that we have access to, which typically center on backpropagation, they give rise to, you know, patterns of activity, patterns of response, um, patterns of like neuronal behavior in these, in these artificial models that look hauntingly s- hauntingly-
- LFLex Fridman
Yeah.
- MBMatt Botvinick
... similar to what you see in the brain. And, uh, you know, is that a c- I mean-
- LFLex Fridman
Yeah, at a cert-
- MBMatt Botvinick
... is that a coincidence? Like... (laughs)
- LFLex Fridman
At a certain point, it starts looking like such coincidence is unlikely to not be deeply meaningful. Yeah.
- MBMatt Botvinick
Yeah. That's... Yeah. The circumstantial evidence is overwhelming.
- LFLex Fridman
(laughs)
- MBMatt Botvinick
But it could be.
- LFLex Fridman
But you're always open to total, uh, flipping of the table.
- MBMatt Botvinick
Hey, of course.
- 1:00:11 – 1:15:18
Meta-reinforcement learning
- MBMatt Botvinick
- LFLex Fridman
Yeah. So y- you have, uh, co-authored several recent papers that sort of weave beautifully between the world of, uh, neuroscience and artificial intelligence. And, uh, this... Maybe if we could, can we just try to dance around and talk about some of them, maybe try to pick out interesting ideas that jump to your mind from memory? So maybe looking at... We were talking about the prefrontal cortex. The 2018, I believe, paper called, uh, The Prefrontal Cortex as a Meta-Reinforcement Learning System.
- MBMatt Botvinick
Yeah.
- LFLex Fridman
What... Is there a key idea that, um, you can speak to from that paper?
- MBMatt Botvinick
Yeah, the, I mean, the key idea is about meta learning. So...
- LFLex Fridman
What is meta learning?
- MBMatt Botvinick
Meta learning is, by definition, uh, a situation in which you have a learning algorithm and the learning algorithm operates in such a way that it gives rise to another learning algorithm. In the, in the earliest applications of this idea, you had one learning algorithm sort of adjusting the parameters on another learning algorithm. But the case that we're interested in this paper is one where you start with just one learning algorithm and then another learning algorithm kind of emerges out of, like out of thin air.
- LFLex Fridman
Mm-hmm.
- MBMatt Botvinick
I can say more about what I mean by that, I don't mean to be, um, you know-
- LFLex Fridman
(laughs) .
- MBMatt Botvinick
... obscurantist.
- LFLex Fridman
Yeah.
- MBMatt Botvinick
But, um, uh, that's the idea of meta learning. You, it, uh, it, it relates to the old idea in psychology of learning to learn. Um, uh, situations where you, you, you have experiences that make you better at learning something new. Like a gr- a familiar example would be learning a foreign language. The first time you learn a foreign language it may be, you know, quite laborious and disorienting and, uh, and novel, but if you, let's say you've learned two, two foreign languages, the third foreign language obviously is gonna be much easier to pick up. And why? Because you've learned how to learn, you know how this goes, you know, "Okay, I'm gonna have to learn how to conjugate, I'm gonna have to..." Um, that's a f- that's a simple form of meta learning, right? In the sense that there's some slow learning mechanism that's giving, that's helping you kind of update your fast learning mechanism. Does that, does that-
- LFLex Fridman
Does that make sense?
- MBMatt Botvinick
... bring it into focus?
- LFLex Fridman
Yeah. So, how, from, from our understand... From the psychology world, from, from n- in neuroscience under our understanding how meta learning works, might work in the human brain, what, uh, what lessons can we draw from that that we can bring into the artificial intelligence world?
- MBMatt Botvinick
Well, yeah. So we... The origin of that paper was in AI work that, that we were doing in my group. We were, we were looking at what happens when you train a recurrent neural network using standard reinforcement learning algorithms, but, but you train that network not just in one task, but you train it in a bunch of interrelated tasks. And then you ask what happens when you give it yet another task in that sort of line of re- of interrelated tasks. And, and what we started to realize is that, um, a form of meta learning spontaneously happens in recurrent neural networks. And, and the simplest way to explain it is to say, a recurrent, a recurrent neural network has a kind of memory in its activation patterns. It, it's recurrent, by definition, in the sense that you have units that connect to other units that connect to other units. So you have sort of loops of connectivity which allows activity to stick around and be updated over time. In psychology we call... In neuroscience we call this working memory, it's like actively holding something in mind. And, um, and, uh, and, and so that memory gives the recurrent neural network a, a, uh, a dynamics, right? The way that the activity pattern evolves over time is inherent to the connectivity of the recurrent neural network. Okay? So that's, that's idea number one. Now, the dynamics of that network are shaped by the connectivity, by the synaptic weights, and those synaptic weights are being shaped by this reinforcement learning algorithm that you're, you know, training the network with. So the punchline is, if you train a recurrent neural network with a reinforcement learning algorithm that's adjusting its weights, and you do that for long enough, the activation dynamics will become very interesting, right? So imagine, imagine I give you a task where you have to press one button or another, left button or right button. And, uh, sometime... And, and there's some probability that I'm gonna give you an M&M if you press, uh, the left button, and there's some probability I'll give you an M&M if you press the other button. And you have to figure out what those probabilities are just by trying things out. But as I said before, instead of just giving you one of these tasks, I give you a whole sequence. You know, I give you two buttons and you figure out which one's best and I go, "Good job. Here's, here's a new box, two new buttons, you have to figure out which one's best. Good job, here's a new box." And every box has its own probabilities and you have to figure. So if you train a neural n- a recurrent neural network on that kind of sequence of tasks, the, what, what happens... It seemed almost magical to, to us when we first started kind of realizing what was going on. The slow learning algorithm that's adjusting the, the pr- synaptic weights, tho- those slow synaptic changes give rise to a network dynamics that themsel- that, you know, the dynamics themselves turn into a learning algorithm. So in other words you can, you can tell this is happening by just freezing the synaptic weights, saying, "Okay, no more learning, you're done. Here's a new box. Figure out which button is best." And the recurrent neural network will do this just fine. There's no... Like, it, it, it figures out which, which button is best, it tran- it kind of transitions from exploring the two buttons to just pressing the one that it likes best in a very rational way. How is that happening? It's happening because the activity of the na- the activity dynamics of the network have been shaped by this slow learning process that's occurred over many, many boxes. And so what's happened is that this slow learning algorithm that's, uh, slowly adjusting the weights is changing the dynamics of the network, the activity dynamics into its own learning algorithm. And as we were, as we were kind of realizing that this is a thing, we-It just so happened that the, the group that was working on this included a bunch of neuroscientists, and it started kind of ringing a bell for us.
- LFLex Fridman
Mm-hmm.
- MBMatt Botvinick
Which is to say that we thought, "This sounds a lot like the distinction between synaptic learning and activity, uh, synaptic memory and activity-based memory in the brain." And it also reminded us of recurrent connectivity that's very characteristic of prefrontal function.
- LFLex Fridman
Mm-hmm.
- MBMatt Botvinick
Um, so there, this is, this is kind of why it's good to have people working on AI that know a little bit about neuroscience and vice versa. Um, because we started thinking about whether we could apply this principle to, to neuroscience, and that's where the paper came from.
- LFLex Fridman
So the, the kind of principle fo- of, uh, the, the recurrence they can see in the prefrontal cortex, then you start to realize that it's possible to, for something like an idea of a learning to learn, uh, emerging from this learning process as long as you keep varying the environment sufficiently.
- MBMatt Botvinick
Exactly. So, so the, the, the kind of metaphorical transition we made to neuroscience was to think, "Okay, well, we know that the prefrontal cortex is highly recurrent. We know that it's an important locus for working memory, for active, activation-based memory, so maybe the prefrontal cortex supports reinforcement learning." In other words, you, y- what is reinforcement learning? You take an action, you see how much reward you got, you update your policy of behavior. Maybe the prefrontal cortex is doing that sort of thing strictly in its activation patterns. It's keeping around a memory in its activity patterns of what you did, how much reward you got, uh, and it's using that, that activity-based me- uh, memory as a basis for updating behavior. But then the question is, well, how did the prefrontal cortex get, get so smart? In other words, how did it... where did these activity dynamics come from? How did that program that's implemented in the recurrent dynamics of the prefrontal cortex arise? And one answer that became evident in this work was, well, maybe, maybe the mechanisms that operate on the synaptic level, uh, which we believe are mediated by dopamine, are, are responsible for shaping those dynamics.
- LFLex Fridman
So, um, this may be a silly question, but because this kind of several temp- temporal sort of classes of learning are happening and so the learning to learn is em- emerges, can it just- can you keep building stacks of learning to learn to learn-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... learning to learn, to learn, to learn, to learn? Because it keeps, I mean, basically abstractions of more powerful abilities to generalize, of learning complex rules.
- MBMatt Botvinick
Yeah.
- LFLex Fridman
Or is, is this that's overstretching the, uh, this kind of mechanism?
- MBMatt Botvinick
Well, one of, one of the, um, one of the people in AI who started thinking about meta learning, um, uh, from ver- very early on, Juergen Schmidhuber, uh, sort of, um, cheekily suggested, I think in his, it may have been in his di- his, uh, PhD thesis, that, uh, we should think about meta, meta, meta, meta, meta, meta learning.
- LFLex Fridman
Yeah. (laughs)
- 1:15:18 – 1:19:01
Dopamine
- LFLex Fridman
If we could talk a little bit about dopamine, you have really ... You were part of, uh, co-authoring really exciting recent paper, very recent, or in terms of release, on dopamine and temporal difference learning. Can you describe, uh, the key ideas of that paper?
- MBMatt Botvinick
Sure, yeah. I mean, one thing I want to pause to do is acknowledge my co-authors on actually both of the papers we're talking about. So the, this dopamine paper-
- LFLex Fridman
I'll, I'll, I'll just, uh, I'll, I'll certainly post all their names-
- MBMatt Botvinick
Okay, wonderful.
- LFLex Fridman
... and, and all, all that kind of stuff.
- MBMatt Botvinick
Yeah, 'cause I, you know, I'm, I'm sort of abashed to be the spokesperson for these papers when, um, I had such amazing collaborators on both. Um, so it's, uh, it's a comfort to me to know that y'all-
- LFLex Fridman
Yeah.
- MBMatt Botvinick
... y'all acknowledge them. Um-
- LFLex Fridman
Yeah, this is an incredible team there, but yeah, so-
- MBMatt Botvinick
Oh, yeah.
- LFLex Fridman
...
- NANarrator
it was-
- MBMatt Botvinick
I think it's such a, it's so much fun. Um, and, and in the case of the, the dopamine paper, uh, we also, um, collaborated with Nao Ichida at Harvard, who, you know, o- obviously a paper simply wouldn't have happened without him. But, um, so, so you were asking for like a, a thumbnail sketch of...
- LFLex Fridman
Yeah, thumbnail sketch or key ideas or, you know-
- MBMatt Botvinick
Yeah.
- LFLex Fridman
... things, the insights that, you know, continue on our kind of discussion here between neuroscience and AI.
- MBMatt Botvinick
Yeah, I mean, this was another... A lot of the work that we've done so far is, um, taking ideas that have bubbled up in AI and, m- m- you know, asking the question of whether the brain might be doing something related, which I think on the surface sounds like something that's really mainly of use to neuroscience. Um, we see it also as a way of validating what we're doing on the AI side. If we can gain some evidence that the brain is using some technique that we've been trying out in our AI work, um, that gives us confidence that, you know, it may be a good idea, that it'll, you know, scale to rich complex tasks, that it'll interface well with other mechanisms. So-
- LFLex Fridman
So you see it as a two-way road?
- MBMatt Botvinick
Yeah, for sure.
- LFLex Fridman
Just because a particular paper is a little bit focused on from one to the oth- uh, from A- AI, from neural networks to neuroscience, ultimately the discussion, the thinking, the productive long-term aspect of it is the, the two-way road nature of the whole interaction.
- MBMatt Botvinick
Yeah, I mean, we, we've talked about the notion of a virtuous circle between AI and neuroscience. And, you know, the way I see it, that's always been there, um, since the two fields, uh, you know, jointly existed. Um, there have been some phases in that history when AI was sort of ahead. There are some phases when neuroscience was sort of ahead. I feel like given the burst of innovation that's happened recently on the AI side, AI is kind of ahead in the sense that there are all of these ideas that w- we, you know, w- w- we, you know, for which it's exciting to consider that there might be neural analogs. Um, and neuroscience, uh, you know, in a sense has been focusing on approaches to studying behavior that come from, you know, that are kind of derived from this earlier era of cognitive psychology. Um, and, you know, so i- in some ways fail to connect with some of the issues that we're, you know, grappling with in AI. Like how do we deal with, you know, large, you know, complex environments. Um, but, uh, I, you know, I think it's inevitable that this circle will keep turning and there will be a moment-... in the not too different- distant future when neuroscience is, uh, pelting AI researchers with, uh, insights that, um, may change the direction of our work.
- LFLex Fridman
Just a- uh, just a quick human question.
- MBMatt Botvinick
Mm-hmm.
- 1:19:01 – 1:23:37
Neuroscience and AI research
- MBMatt Botvinick
- LFLex Fridman
Is it, uh, you have a l- s- parts of your brain... this is very meta-
- MBMatt Botvinick
(laughing)
- LFLex Fridman
... but they're able to both think about neuroscience and AI. You know, I don't often meet people like that. And so do you- do you think, uh... let me ask a meta-plasticity question.
- MBMatt Botvinick
Uh-
- LFLex Fridman
Do you- do you think a human being can be both good at AI and neuroscience? Because, like, what... on the team-
- MBMatt Botvinick
Mm-hmm.
- LFLex Fridman
... at- at DeepMind, what kind of human can occupy these two realms? And is that something you see everybody should be doing, can be doing? Or is that a v- a very special few can kind of jump... just like we talk about art history. I would think it's a special person that can major in art history and also consider being a surgeon.
- MBMatt Botvinick
Otherwise known as a dilettante? (laughing)
- LFLex Fridman
A dil- a dilettante, yeah. Easily distracted. (laughing) No.
- MBMatt Botvinick
Um, I- I, um, I think it does take a special kind of person to be truly world-class at both AI and neuroscience, and I am not on that list. Um, I happen to be someone who, whose interest in neuroscience and psychology, uh, involved y- using the kinds of modeling techniques that are now, um, very central in AI. And that sort of, I guess, bought me a ticket to be involved in all of the amazing things that are going on in AI research right now. Um, I do know a few people who I would consider pretty expert on both fronts, uh, and I won't embarrass them by naming them. But th- you know, there are- there are, like, exceptional people out there, um, who are like this. The- the one- the one thing that I find, um, is a- is a barrier to being truly world-class on both fronts is th- is, um, the- just the- the complexity of the technology that's, uh, involved in both disciplines now. So, um, the- the engineering expertise that it takes to- to do, you know, truly frontline, hands-on AI research is really, really considerable.
- LFLex Fridman
The learning curve of the tools, just like the specifics of just... whether it's programming or the kind of tools necessary to collect the data, to manage the data, to distribute, to compute, all that kind of stuff.
- MBMatt Botvinick
Yeah.
- LFLex Fridman
And on the neuroscience, I guess, side, there would be all different sets of tools to work with.
- MBMatt Botvinick
Exactly, especially with the recent explosion in, you know, in neuroscience methods. So but- but h- you know, so having said all that, I- I think- I think the r- I think the best scenario for both neuroscience and AI is to have people who... interacting, who live at every point on the spectrum, from exclusively focused on neuroscience to exclu- exclusively focused on the engineering side of AI. But to- but to have those people, uh, you know, inhabiting a community where they're talking to people who live elsewhere on the- on the spectrum. And I be- I may be someone who's very close to the center, in- in the sense that I have one foot in the neuroscience world and one foot in the AI world.
- LFLex Fridman
Yeah.
- MBMatt Botvinick
And- and that central position, I will admit, prevents me, at least someone with my limited cognitive capacity, fr- from being a truly, you know, true- having true technical expertise in any- in either domain. But at the same time, I- I at least hope that it's worthwhile having people around who can kind of, t- y- you know, see the connections between these two-
- LFLex Fridman
Yeah, the community, the, uh, yeah, the- the emergent intelligence of the community-
- MBMatt Botvinick
Yeah, yeah, yeah.
- LFLex Fridman
... when it's nicely distributed i- is useful. Okay, so-
- MBMatt Botvinick
Exactly, yeah. So hopefully, that... I mean, I've seen that work- I've seen that work out well at DeepMind. There- there are- there are people who... I mean, even if you just focus on the AI work that happens at DeepMind, it's been a good thing to have some people around doing that kind of work whose PhDs are in neuroscience or psychology. Every- every academic discipline has its kind of, uh, blind spots and kind of unfortunate obsessions, and its metaphors and its reference points. And having some intellectual diversity, uh, is- is really healthy. People get each other unstuck, I think. Um, I see it all the time at DeepMind. And, uh, you know, I- I like to think that the people who bring some neuroscience, um, background to the table are- are helping
- 1:23:37 – 1:39:56
Human side of AI
- MBMatt Botvinick
with that.
- LFLex Fridman
So one of the- one of my, like, probably the deepest passion for me, what I would say, maybe we kind of sp- spoke off mic a little bit about it, but, um, that- that I think is a blind spot for at least robotics and AI folks is human-robot interaction, human-agent interaction. Um, maybe... do you have thoughts about how we, uh, reduce the size of that blind spot? Do you also share the- the- the feeling that not enough folks are studying this aspect, uh, of interaction?
- MBMatt Botvinick
Well, I- I'm- I'm actually pretty intensively interested in this issue now. And there are people in my group who've actually pivoted pretty hard over the last few years from doing more traditional cognitive psychology and cognitive neuroscience to doing experimental work on human-agent interaction.
- LFLex Fridman
Mm-hmm.
- MBMatt Botvinick
And, um, there are a couple reasons that I'm pretty passionately interested in this. One is, um, i- i- it's kind of the outcome of having thought for a few years now about what we're up to. Like, what- what are we, like- (laughing)
- LFLex Fridman
(laughing)
- MBMatt Botvinick
... what are we doing?
- LFLex Fridman
Yeah.
- MBMatt Botvinick
Like, wha- what is this- what is this age- AI research for? So what does it mean to make the world a better place? I think......I'm pretty sure that means making life better for humans.
- LFLex Fridman
Yeah.
- MBMatt Botvinick
Um, and so how do you make life better for humans? That's, that's a proposition that when you look at it carefully and honestly, is rather horrendously complicated, especially when the AI systems that you're, that you're building are learning systems. They're not, you're not, you know, programming something that you then introduce to the, into the world and it just works as programmed, like, uh, Google Maps or something. Um, we're building systems that, that learn from experience. So you, th- that, that typically leads to AI safety questions. How do we keep these things from getting out of control? How do we keep them from doing things that harm humans? And I mean, I hasten to say, I consider those hugely important-
- LFLex Fridman
Yes.
- MBMatt Botvinick
...issues and there are large sectors of the research community at DeepMind and of course elsewhere, who are dedicated to thinking hard all day every day about that. But, uh, there's a, there's, I guess, I guess I would say a positive side to this too, which is to say, well, what would it mean to make human life better? W- w- a- and how, how can we imagine learning systems doing that? Uh, and, and in talking to my colleagues about that, we reached the initial conclusion that it's not sufficient to philosophize about that. You actually have to take into account how humans actually work, uh, and what humans want and the difficulties of knowing what humans want, um, and, uh, the difficulties that arise when humans want different things. Um, and, and so human agent interaction has become, uh, you know, a quite, a quite intensive focus of my group lately, i- if for no other reason that in order to really address that, that issue in an adequate way, you have to... I mean, psychology becomes part of the picture.
- LFLex Fridman
Yeah. And, and it... so there's a, there's a few elements there. So if you focus on solving into... like the, if you focus on the robotics problem of say, AGI without humans in the picture is a, it's, you're missing fundamentally the, the final step. You... wh- when you do want to help human civilization, you eventually have to interact with humans.
- MBMatt Botvinick
Yeah.
- LFLex Fridman
And when you create a learning system, just as you said, that will eventually have to interact with humans, the interaction itself has to be... become a... has to become part of the learning process.
- MBMatt Botvinick
Right.
- LFLex Fridman
So you can't just watch... well, my sense is, it sounds like your sense is you can't just watch humans to learn about humans.
- MBMatt Botvinick
Yeah.
- LFLex Fridman
You have to also be part of the human world. You have to interact with humans.
- MBMatt Botvinick
Yeah, exactly. And I mean, th- then questions arise that start imperceptibly but inevitably to slip beyond the realm of engineering. So questions like, if you have an agent that can do something that you can't do, under what conditions do you want that agent to do it? So, you know, if, uh, you know, if I, if I have a, if I have a robot that can play, um, Beethoven sonatas better than any human in the sense that the, you know, the, the sensitivity (laughs) , the like, the expression, the expression is just beyond what any human... Do I, do I wanna listen to that? D- do I want to go to a concert and hear a robot play? These are, these are, these aren't engineering questions. These are questions about human preference and human culture.
- LFLex Fridman
Yeah.
- MBMatt Botvinick
And, um-
- LFLex Fridman
Psychology bordering on philosophy. (laughs)
- MBMatt Botvinick
Yeah. And then, and then you start asking, well, well, even if we knew the answer to that, is it our place as AI engineers to build that into these agents? Probably the agents should interact with humans beyond the population of AI engineers and figure out what those humans want.
- LFLex Fridman
Yes.
- MBMatt Botvinick
Um, and then, you know, when you start... I referred this a moment ago, but even that becomes complicated becau-... what if, what if two a- what if two humans want different things ha... and, and you have only one agent that's able to interact with them and try to satisfy their preferences? Then you're into the realm of, um, of, uh, of l- like economics and, and social choice theory and, and even politics. So there's a sense in which if you, if you kind of follow what we're doing to its logical conclusion, then it goes beyond questions of engineering and technology And, you know, starts to shade imperceptibly into questions about what kind of society do you want? A- and actually that... once, once that dawned on me, I actually felt... I don't know what the right word is... quite refreshed in my, i- in my involvement in AI research. It's almost like this... building this kind of stuff is gonna lead us back to asking really-
Episode duration: 2:00:32
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode 3t06ajvBtl0
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome