Skip to content
Melanie Mitchell: Concepts, Analogies, Common Sense & Future of AI | Lex Fridman Podcast #61
This video isn’t embeddableWatch on YouTube →
Lex Fridman PodcastLex Fridman Podcast

Melanie Mitchell: Concepts, Analogies, Common Sense & Future of AI | Lex Fridman Podcast #61

Lex Fridman and Melanie Mitchell on melanie Mitchell Explores Concepts, Analogies, Common Sense, Future AI.

Lex FridmanhostMelanie Mitchellguest
Dec 28, 20191h 52mWatch on YouTube ↗

EVERY SPOKEN WORD

  1. 0:00 – 2:31

    Melanie Mitchell’s background + podcast framing and sponsor message

    1. LF

      The following is a conversation with Melanie Mitchell. She's a professor of computer science at Portland State University, and an external professor at Santa Fe Institute. She has worked on and written about artificial intelligence from fascinating perspectives, including adaptive complex systems, genetic algorithms, and the copycat cognitive architecture, which places the process of analogy-making at the core of human cognition. From her doctoral work, with her advisors Douglas Hofstadter and John Holland, to today, she has contributed a lot of important ideas to the field of AI, including her recent book, simply called Artificial Intelligence: A Guide for Thinking Humans. This is the Artificial Intelligence podcast. If you enjoy it, subscribe on YouTube, give it five stars on Apple Podcasts, support it on Patreon, or simply connect with me on Twitter, @LexFridman, spelled F-R-I-D-M-A-N. I recently started doing ads at the end of the introduction. I'll do one or two minutes after introducing the episode, and never any ads in the middle that can break the flow of the conversation. I hope that works for you and doesn't hurt the listening experience. I provide timestamps for the start of the conversation, but it helps if you listen to the ad and support this podcast by trying out the product or service being advertised. This show is presented by Cash App, the number one finance app in the App Store. I personally use Cash App to send money to friends, but you can also use it to buy, sell, and deposit bitcoin in just seconds. Cash App also has a new investing feature. You can buy fractions of a stock, say $1 worth, no matter what the stock price is. Broker services are provided by Cash App Investing, a subsidiary of Square and member SIBC. I'm excited to be working with Cash App to support one of my favorite organizations called FIRST, best known for their FIRST robotics and LEGO competitions. They educate and inspire hundreds of thousands of students in over 110 countries, and have a perfect rating on Charity Navigator, which means the donated money is used to maximum effectiveness. When you get Cash App from the App Store or Google Play and use code LEXPODCAST, you'll get $10 and Cash App will also donate $10 to FIRST, which again is an organization that I've personally seen inspire girls and boys to dream of engineering a better world. And now, here's my conversation

  2. 2:31 – 5:15

    Why the term “Artificial Intelligence” is confusing (and what to call it instead)

    1. LF

      with Melanie Mitchell. The name of your new book is Artificial Intelligence, subtitle, A Guide for Thinking Humans. The name of this podcast is Artificial Intelligence. So let me take a step back and ask the old Shakespeare question about roses and, uh, what do you think of the term "artificial intelligence" for our big and complicated and interesting field?

    2. MM

      I'm not crazy about the term. (laughs) I think it has a few problems, um, because it, it's, means so many different things to different people. And intelligence is one of those words that isn't very clearly defined either. There's so many different kinds of intelligence, degrees of intelligence, approaches to intelligence. John McCarthy was the one who came up with the term "artificial intelligence." And what, from what I read, he called it that to differentiate it from cybernetics, which was another related movement at the time. And he later regretted calling it artificial intelligence. Uh, Herbert Simon was pushing for calling it complex information processing. (laughs) Which got nixed, but you know, probably is equally vague, I guess.

    3. LF

      Is it the intelligence or the artificial in terms of words that's-

    4. MM

      I think it-

    5. LF

      ... most problematic, would you say?

    6. MM

      Yeah, I think it's a little of both. But you know, it has some good sides because I personally was attracted to the field because I was interested in phenom- phenomenon of intelligence.

    7. LF

      Mm-hmm.

    8. MM

      And if it was called complex information processing, maybe I'd be doing something wholly different now.

    9. LF

      What do you think of, I've heard the term used cognitive systems, for example. So using cognitive...

    10. MM

      Yeah, I mean, cognitive has certain associations with it and people like to separate things like cognition and perception, which I don't actually think are separate, but often people talk about cognition as being different from sort of other aspects of, of intelligence. Uh, it's sort of higher level

    11. NA

      So to-

    12. LF

      So to you, cognition is this broad, beautiful mess of things that's encompasses the whole thing. Memory...

    13. MM

      Yeah.

    14. LF

      ... perception...

    15. MM

      I, I think it's hard to draw lines like that.

    16. LF

      Right.

    17. MM

      When I was coming out of grad school in the nine- in 1990, which is when I graduated, that was during one of the AI winters. And I was advised to not put AI, artificial intelligence, on my CV, but instead call it intelligent systems.

    18. LF

      Mm-hmm.

    19. MM

      So that was kind of an, an, a euphemism, (laughs) I guess.

  3. 5:15 – 10:06

    Weak vs strong AI, moving goalposts, and whether we’ll ever “cross the line”

    1. LF

      What about, to stick briefly on, uh, on terms and words, the idea of artificial general intelligence or, uh, or like Yann LeCun prefers human-level intelligence, sort of starting to talk about ideas that, that achieve higher and higher levels of intelligence and somehow artificial intelligence seems to be a term used more for the narrow, very specific applications of AI and sort of... The, there's, what set of terms appeal to you to describe the thing that perhaps we strive to create?

    2. MM

      People have been struggling with this for the whole history of the field. And defining exactly what it is that we're talking about. You know, John Searle had this distinction between strong AI and weak AI.

    3. LF

      Mm-hmm.

    4. MM

      And weak AI could be general AI but his, his idea was stro- strong AI was the view that a machine is actually thinking.... that, as opposed to simulating thinking or carrying out intelligent processes that we would call intelligent.

    5. LF

      At a high level, if you look at the founding of the field of McCarthy and S- Searle and so on, are we closer to having a better sense of that line between narrow, weak AI and strong AI?

    6. MM

      Yes, I think we're closer to having a better idea of what that line is. Early on, for example, a lot of people thought that playing chess would be... You couldn't play chess if you didn't have sort of general human level intelligence. And of course once computers were able to play chess better than humans, that revised that view, and people said, "Okay, well maybe now we have to revise what we think of intelligence as," or-

    7. LF

      Right.

    8. MM

      Uh, and so that's kind of been a theme throughout the history of the field, is that once a machine can do some task, we then have to look back and say, "Oh, well, that changes my understanding of what intelligence is, because I don't think that machine is intelligent. At least that's not what I want to call intelligence."

    9. LF

      Do you think that line moves forever? Or will we eventually really feel as a civilization like we crossed a line, if it's possible?

    10. MM

      It's hard to predict, but I don't see any reason why we couldn't, in principle, create something that we would consider intelligent. I don't know how we will know for sure. Maybe our own view of what intelligence is will be refined more and more until we finally figure out what we mean when we talk about it, but I- I think eventually we will create machines in a sense that have intelligence. They may not be the kinds of machines we have now. And one of the things that that's going to produce is- is making us sort of understand our own machine-like qualities, that we, in a sense, are mechanical in the sense that like cells. Cells are kind of mechanical. They pro- they have algorithms they process information by, and somehow out of this mass of cells we get this emergent property that we call intelligence. But underlying it is, uh, really just cellular processing and- and lots and lots and lots of it.

    11. LF

      Do you think we'll be able to... Do you think it's possible to create intelligence without understanding our own mind? You said sort of in that process we'll understand more and more, but do you think it's possible to sort of create without really fully understanding from a mechanistic perspective, sort of from a functional perspective, how our mysterious mind works?

    12. MM

      If I had to bet on it, I would say no, we- we- we do have to understand our own minds, at least to some significant extent, but it- it- I think that's a really big open question. I've been very surprised at how far kind of brute force approaches based on, say, big data and huge networks can- can take us. I wouldn't have expected that.

    13. LF

      Mm-hmm.

    14. MM

      And they have nothing to do with the way our minds work, so that's been surprising to me, so it could be wrong.

  4. 10:06 – 18:38

    Why humans want to create artificial minds—and what kinds of intelligence matter most

    1. LF

      To explore the psychological and the philosophical, do you think we're okay as a species with, uh, something that's more intelligent than us? Do you think perhaps that the reason we're pushing that line further and further is be- we're afraid of acknowledging that there's something stronger, better, smarter than us humans?

    2. MM

      Well, I'm not sure we can define intelligence that way because, you know, smarter than is with- with respect to what- what, you know? Computers are already smarter than us in some areas. They can multiply much better than we can. They- they can figure out driving routes to take much faster and better than we can. They have a lot more information to draw on. They know about, you know, traffic conditions and all that stuff. So for any given particular task, sometimes computers are much better than we are, and we're totally happy with that, right?

    3. LF

      Okay.

    4. MM

      I'm totally happy with that. I don't... It doesn't bother me at all. I guess the question is, you know, what- which things about our intelligence would we feel very sad or- or upset that machines had- had been able to recreate?

    5. LF

      Yes. Yeah.

    6. MM

      So in the book, I talk about my former PhD advisor, Douglas Hofstadter, who encountered a music generation program, and that was really the line for him, that if- if a machine could create beautiful music, that would be terrifying for him because that is something he feels is really at the core of what it is to be human, creating beautiful music, art, literature. I, you know, I don't think... He doesn't like the fact that machines can recognize spoken language really well. Like he doesn't, he personally doesn't like using speech recognition, but I don't think it bothers him to his core because it's like, okay, that's not at the core of humanity, but it may be different for every person what- what really...... they feel would, er, usurp their humanity. (laughs) And I think maybe it's a generational thing also. Maybe our children or our children's children will be adapted. They'll, they'll adapt to these new devices that can do all these tasks and, and say, "Yes, this thing is smarter than me in all these areas, but, uh, that's great 'cause it helps me." (laughs)

    7. LF

      Looking at the broad history of our species, why do you think so many humans have dreamed of creating artificial life and artificial intelligence throughout the history of our civilization? So not just this century or the 20th century, but really many... Throughout many centuries that preceded it?

    8. MM

      That's a really good question, and I have wondered about that. 'Cause I'm... I, myself, you know, was driven by curiosity about my own thought processes and thought it would be fantastic to be able to get a computer to mimic some of my thought processes and... I, I'm not sure why we're so driven. I think we want to understand ourselves better, and we also want machines to do things for us. But I don't know, there's something more to it because it's so deep in, in the kind of mythology or the ethos of our, our species. And I don't think other species have this drive. (laughs) So I don't know.

    9. LF

      If, if you were to sort of psychoanalyze yourself and your... And your own interest in AI, are you... What excites you about creating intelligence? You said understanding our own selves?

    10. MM

      Yeah. I think that's what drives me particularly. I'm really interested in human intelligence, but I'm al- I'm also interested in the, sort of the phenomenon of intelligence more generally. And I don't think humans are the only thing with intelligence, you know? I... And, uh, or even animals. But I think intelligence is a concept that encompasses a lot of complex systems. And if you think of things like, uh, insect colonies or, uh, cellular processes or the immune system or all kinds of different biological or even soc- soc- societal processes have, as an emergent property, some aspects of what we would call intelligence. You know, they have memory. They do process information. They have goals. They accomplish their goals, et cetera. And, um, to me that... The question of what is this thing we're, we're talking about here was, uh, really fascinating to me. And, and exploring it using computers seemed to be a good way to approach the question.

    11. LF

      So do you think, kind of, of intelligence... Do you think of the... our universe as a kinda hierarchy of complex systems and then intelligence is just the property of any... You, you can look at any level and every level has some a- aspect of intelligence? So we're just, like, one little speck in that giant hierarchy of complex systems?

    12. MM

      I don't know if I would say any system like that has intelligence. But I guess, I... What I wanna... I don't have a good enough definition of intelligence to say that.

    13. LF

      So let me, let me do-

    14. MM

      (laughs)

    15. LF

      ... sort of, uh, multiple choice, I guess.

    16. MM

      (laughs)

    17. LF

      So, uh, so you said ant colonies. So are ant colonies intelligent? Are the bacteria in our body int- intelligent? And then l- going to the ph- the physics world, molecules and the behavior at the quantum level of, of electrons and so on, is... Are those kinds of systems, do they possess intelligence? Like where's, where's the line that-

    18. MM

      Yeah.

    19. LF

      ... feels compelling to you?

    20. MM

      I don't know. I mean, I think intelligence is a continuum, and I think that the ability to, in some sense, have intention, have a goal, have, have a... Some kind of self-awareness is part of it. So I'm not sure if... You know, it's hard to know where to draw that line. I think that's kind of a mystery. But I wouldn't say that, say, the, you know... This, the planets orbiting the sun are... Is an intelligent system. I mean, I would find th- that maybe not the right term to describe that. And this is... You know, there's all this debate in the field of, like, what's, what's the right way to define intelligence? What's the right way to model intelligence? Should we think about computation? Should we think about dynamics and, um, should we think about, you know, free energy and all of that stuff? And I think that it's, it's a fantastic time to be in the field because there's so many questions and so much we don't understand. There's so much work to do.

    21. LF

      So are we... Are we the most special kind of intelligence in this kind of... You said there's, uh, a bunch of different elements and characteristics of intelligence systems and colonies. Uh, i- our... Is human intelligence, the thing in our brain, is that the most interesting kind of intelligence in this continuum?

    22. MM

      Well, it's interesting to us 'cause, 'cause it, it is us. (laughs)

    23. LF

      So-

    24. MM

      I mean, interesting to me, yes, and... Because I'm part of, you know, human...

    25. LF

      But to understanding the fundamentals of intelligence, what I'm getting at-

    26. MM

      Yeah, I mean, there-

    27. LF

      ... do we... Is studying the human... Is sort of... I- if everything we've talked about, what you talk about in your book, what, uh... Just the AI field, this notion, yes, it's hard to define, but it's usually talking about something that's very akin to human intelligence.

    28. MM

      Yeah. To me, it is the most interesting because it's the most complex, I think. It's the most self-aware.... it is the only system, at least that- that I know of, that reflects on its own intelligence.

  5. 18:38 – 24:57

    Forecasting AI: why predictions fail and Mitchell’s “100+ years / 100 Nobel Prizes” view

    1. LF

      And you talk about the history of AI, and- and us, in terms of, uh, creating artificial intelligence, being terrible at predicting the future with AI or with tech in general. So, why do you think we're so bad at predicting the future? Are we hopelessly bad? So, no matter what, whether it's this decade or the next few decades, every time we make a prediction, there's just no way of doing it well? Or, as the field matures, we'll be better and better at it?

    2. MM

      I believe as the field matures, we will be better. And I think the reason that we've had so much trouble is that we have so little understanding of our own intelligence.

    3. LF

      Hmm.

    4. MM

      So, there's the famous story about Marvin Minsky assigning computer vision as a summer project-

    5. LF

      Yeah.

    6. MM

      ... to his undergrad students. And I believe that's actually a true story.

    7. LF

      Yeah, no, there's a-

    8. MM

      (laughs)

    9. LF

      ... there's- there's a write-up on it w- that everyone should read. It's like a pr- I think it's like a proposal, uh, that describes everything that sh- should-

    10. MM

      Yeah.

    11. LF

      ... be done in that project. And it's hilarious because it, uh, I mean, you could explain it, but from my sort of recollection, it describes basically all the fundamental problems with computer vision, many of which that still haven't been solved.

    12. MM

      Yeah. And- and- and I don't know how far they really expected to get.

    13. LF

      Right.

    14. MM

      But I think that, and- and they're real, you know, Marvin Minsky was a super smart guy, and very sophisticated thinker. Uh, but I think that no one really understands or understood, still doesn't understand, how complicated, how complex the things that we do are because they're so invisible to us, you know? To us, vision, being able to look out at the world and describe what we see, that's just immediate. It feels like it's no work at all. So it didn't seem like it would be that hard. But there's so much going on unconsciously, sort of invisible to us, that I think we overestimate how easy it will be to get computers to do it.

    15. LF

      And sort of, uh, for me to ask an unfair question, you've done, uh, research, you've thought about many different branches of AI through this book, uh, widespread looking at where AI has been and where it is today. If you were to make a prediction, how many years from now would we, as a society, create something that you would say achieved human-level intelligence or superhuman-level intelligence?

    16. MM

      That is an unfair question (laughs) .

    17. LF

      A prediction that will most likely be wrong, so but it's just your notion because-

    18. MM

      Okay, I'll say, I'll say more than a hundred years.

    19. LF

      More than a hundred years.

    20. MM

      And there, I quoted somebody in my book who said that human-level intelligence is a hundred Nobel Prizes away.

    21. LF

      (laughs)

    22. MM

      (laughs) Which I like 'cause it's a-

    23. LF

      Oh, yeah.

    24. MM

      ... it's a nice way to- to sort of, it's a nice unit for prediction. (laughs)

    25. LF

      (laughs)

    26. MM

      And it's like that many fantastic discoveries have to be made. And of course there's no Nobel Prize in AI. (laughs)

    27. LF

      Right.

    28. MM

      Not yet, at least. (laughs)

    29. LF

      If we look at that a hundred years, your sense is really the journey to intelligence has to go through something, uh, something more complicated, as again, to our own cognitive systems, uh, understanding them, being able to create them in, uh, in artificial systems, as opposed to sort of taking the machine learning approaches of today and really scaling them and scaling them and scaling them exponentially with both compute and hardware and- and, uh, data.

    30. MM

      That would be my- that would be my guess. Um, you know, I think that in- in, uh, the- the- the sort of going along in the narrow AI, that these current th- the current approaches will get better. You know, I think there's some fundamental limits to how far they're gonna get. I might be wrong, but that's what I think. Uh, but, and- and there's some fundamental weaknesses that they have that, um, I talk about in the book, that- that just comes from this approach of- of supervised learning, um, req- req- require- requiring s- s- sort of feed forward networks-

  6. 24:57 – 31:24

    Competing AI worldviews: scaling deep learning vs hybrids, causality, and developmental learning

    1. MM

      It's certainly possible. One thing that surprised me when I was writing the book is how far apart different people are in- in the field are on-

    2. LF

      Yeah, that's interesting.

    3. MM

      ... their opinion of how- how far the field has come and what it's accomplished and what's- what's gonna happen next.

    4. LF

      W- what's your sense of the different... Who are the different people, groups, mindsets, thoughts, uh, in the community about where AI is today?

    5. MM

      Yeah, they're all over the place. So- so there's- there's kind of the- the singularity trans-humanism group. I don't know exactly how to characterize that approach, what- which just-

    6. LF

      Like Ray Kurzweil, the- the-

    7. MM

      Yeah, the sort of exponential-

    8. LF

      ... robot.

    9. MM

      ... exponential progress we're- we're- we're on the, sort of, almost at the- the hugely accelerating part of the exponential, and by... in the next 30 years we're going to see super intelligent AI and all that, and we'll be able to upload our brains and... That. So there- there's that kind of extreme view that most, I think, most people who work in AI don't have.

    10. LF

      Hmm.

    11. MM

      They disagree with that. But there are people who- who are... maybe don't- aren't, you know, singularity people, but- but they're... they do think that the current approach of deep learning is going to scale and is going to kind of go all the way, basically-

    12. LF

      Hmm.

    13. MM

      ... and take us to true AI or human-level AI, or whatever you wanna call it. Uh, and there's quite a few of them. And a lot of them... like, a lot of the people I met who work at, um, big tech companies in- in AI groups kind of have this view that we're really not that far, you know?

    14. LF

      Just to linger on that point, sort of, if- if I can take as an example, like, Yann LeCun. I don't know if you know about his work and so- or- or his viewpoints on this.

    15. MM

      Yeah. Of course. I do.

    16. LF

      He believes that there's a bunch of breakthroughs, like fundamental... like Nobel Prizes, there's- needed still.

    17. MM

      Yeah. Right.

    18. LF

      But I think he thinks those breakthroughs will be built on top of deep learning.

    19. MM

      Right.

    20. LF

      And then there's some people who think we need to kinda put deep learning to the side a little bit as just one module that's helpful in the bigger-

    21. MM

      Right.

    22. LF

      ... cognitive framework.

    23. MM

      Right. So- so- so I think, uh... so what I understand, Yann LeCun is rightly saying supervised learning is not sustainable, we have to figure out how to do unsupervised learning. That that's gonna be the key. Um, and, you know, I think that's probably true. Uh, I think unsupervised learning is gonna be harder than people think (laughs) .

    24. LF

      (laughs)

    25. MM

      I mean, the way that we humans do it. Then there's the opposing view, you know, that there's a... the- the Gary Marcus kind of hybrid view where- where deep learning's one part, but we need to bring back kind of these symbolic approaches and combine them. Of course, no one knows how to do that very well.

    26. LF

      Which is the more important part-

    27. MM

      Right.

    28. LF

      ... to- to emphasize, and how do they... yeah, how do they fit together? What's- what's the foundation? What's the thing that's on top?

    29. MM

      Yeah.

    30. LF

      What's the cake? What's the icing?

  7. 31:24 – 36:47

    Copycat: an analogy-making system built from agents and a shared workspace

    1. LF

      That's definitely something I'd l- I'd love to talk about i- in a little bit, to step into the cognitive world then, if you don't mind, 'cause you've done so many interesting things. If w- if we look to CopyCat, taking a couple of decades step back, you, Douglas Hofstadter, and others have created and developed CopyCat more than 30 years ago.

    2. MM

      (laughs) That's painful to hear. (laughs)

    3. LF

      (laughs) So what is it? What is, w- well, what is CopyCat?

    4. MM

      It's a program that makes analogies in an idealized domain, idealized world of letter strings. So as you say, 30 years ago, wow.

    5. LF

      Yeah.

    6. MM

      Uh, so I started working on it when I started grad school in, um, 1984. Wow. (laughs)

    7. LF

      (laughs)

    8. MM

      Dates me. Um, and it's based on Doug Hofstadter's ideas that, about, um, that analogy is really a core aspect of thinking. Uh, I remember he, he has a really nice quote in, in, in the book by, by himself and Emmanuel Sander called Surfaces and Essences. I don't know if you've seen that book, but it's, it's about analogy.

    9. LF

      Mm-hmm.

    10. MM

      Uh, he says, "Without concepts, there can be no thought, and without analogies, there can be no concepts."

    11. LF

      Mm-hmm.

    12. MM

      So the view is that analogy is not just this kind of reasoning technique where we go, you know, uh, shoe is to foot as glove is to what? You know, these kinds of things that we have on IQ tests or whatever. Uh, that, but that it's much deeper, it's much more per- uh, pervasive in everything we do, in every, our language, our, our thinking, our perception. So we, so he had a view that was a very active perception idea. So the idea was that, um, instead of having kind of wha- a passive, uh, network in which you ha- have input that's being processed through these feedforward layers and then there's an output at the end, that perception is really a, a dynamic process, you know, where, like, our eyes are moving around and they're getting information, and that information is feeding back to what we look at next, influences what we look at next and how we look at it. And so CopyCat was trying to do that, kind of simulate that kind of idea where you have these, um, agents. It's kind of an agent-based system, and you have these agents that are picking things to look at and deciding whether they were interesting or not-

    13. LF

      Mm-hmm.

    14. MM

      ... whether they should be looked at more. And, and that would influence other agents.

    15. LF

      Now, how do they interact? The-

    16. MM

      So they interacted through this global kind of what we call the workspace.

    17. LF

      Mm-hmm.

    18. MM

      So it was actually inspired by the old blackboard systems, where you would have agents that post information on a blackboard, a common blackboard.

    19. LF

      Mm-hmm.

    20. MM

      This is, like, old, very old-fashioned AI.

    21. LF

      Is that, (laughs) is that, are we talking about, like, in physical space? Is this a computer program? What are you-

    22. MM

      It's a computer program. Yeah.

    23. LF

      So a- agents posting concepts on a blackboard kind of thing?

    24. MM

      Yeah, we called it a workspace. And it, it, it, it's, it, it's, the workspace is a data structure. The agents are little pieces of code that, you could think of them as detect- little detectors or little filters that-

    25. LF

      Mm-hmm.

    26. MM

      ... say I'm gonna pick this place to look and I'm gonna look for a certain thing, and is this the thing I, I think is important? Is it there? So it's almost like, you know, convolution in a way-

    27. LF

      Yeah.

    28. MM

      ... except a little bit, uh, more general and saying, and then highlighting it on the, on the wo- in the workspace.

    29. LF

      Wha- wha- what's i- once it's in the workspace, how do the things that are highlighted relate to each other? Like, what's, is it-

    30. MM

      So the, there's different kinds of agents that can build connections between different things. So, so just to give you a concrete example, what CopyCat did was it, it made analogies between strings of letters. So here's an example. ABC changes to ABD. What does IJK change to?

  8. 36:47 – 42:42

    Concepts and analogies: why “essential sameness” underlies perception and thought

    1. LF

      Let, let's step back for a second. So I really like that quote, uh, that you said, "Without concepts, there could be no thought, and without analogies, there could be no concepts." In a, in a Santa Fe presentation, you said that it should be one of the mantras of AI.

    2. MM

      Yes.

    3. LF

      ... and that you also yourself said, uh, how to form and fluidly use concept is the most important open problem in AI.

    4. MM

      Yes.

    5. LF

      How to form and fluidly use concepts is the most important open problem in AI. So let's, uh... What is a concept and what is an analogy?

    6. MM

      A concept is, in some sense, a fundamental unit of thought. So say we have a, uh, concept, uh, o- of a, a dog, okay? And a concept is embedded in a whole space of concepts, so that there are certain concepts that are closer to it or farther away from it.

    7. LF

      Are these concepts... Are they really, like, fundamental, like we mentioned innate, almost like axiomatic, like, very basic and then there's other stuff built on top of it?

    8. MM

      Yeah.

    9. LF

      Or does this include everything? Is... Are there complicated... Th- like-

    10. MM

      You can certainly have, form new concepts.

    11. LF

      Right. I guess that's the question I'm asking.

    12. MM

      Yeah.

    13. LF

      Can you form new concepts that are, uh, combina- complex combinations of other concepts?

    14. MM

      Yes, absolutely. And that's kind of what we, we do-

    15. LF

      Yeah.

    16. MM

      ... you know, learning. Uh-

    17. LF

      And then what's the role of analogies in that structure?

    18. MM

      So analogy is when you recognize that one situation is essentially the same as another situation, and "essentially" is kinda the key word there, uh, 'cause it's not the same. So if I say, um... Last week, I did a podcast interview-

    19. LF

      Mm-hmm.

    20. MM

      ... in... actually, like, three days ago (laughs) -

    21. LF

      (laughs)

    22. MM

      ... in Washington, DC. And that situation was very similar to this situation, although it wasn't exactly the same, you know? It was a different person-

    23. LF

      Mm-hmm.

    24. MM

      ... sitting across from me. We had different kinds of microphones. Uh, the questions were different. The building was different. Uh, uh, there was all kinds of different things, but really, it was analogous.

    25. LF

      Mm-hmm.

    26. MM

      Um, or I can say, i- y- so, so, so bo- doing a podcast interview, that's kind of a conce- it's a new concept.

    27. LF

      Yes, concept.

    28. MM

      You know, I, uh, never had that concept before... (laughs)

    29. LF

      (laughs)

    30. MM

      (laughs) Until this year, essentially.

  9. 42:42 – 55:33

    Mental models and generative perception: top-down expectations shaping what we see

    1. LF

      If you could just linger on it a little bit, like, wha- what, what do you think it takes to engineer a process like that for us in our artificial systems?

    2. MM

      We need to understand better, I think, how, how we do it, how humans do it.And it comes down to internal models, I think. You know, people talk a lot about mental models, that concepts are mental models, that I, I can, in my head, I can do a simulation-

    3. LF

      Mm-hmm.

    4. MM

      ... of a situation, like walking a dog.

    5. LF

      Right.

    6. MM

      And that, there, there's some work in psychology that promotes this idea that all of concepts are really mental simulations, that whenever you encounter a concept or situation in the world, or you read about it or whatever, you do some kind of mental simulation-

    7. LF

      Mm-hmm.

    8. MM

      ... that allows you to predict what's gonna happen, to, to develop expectations of what's gonna happen.

    9. LF

      Mm-hmm.

    10. MM

      So that's the kind of structure I think we need, is that kind of mental model that... And the, you know, in our brains, somehow these mental models are very much interconnected.

    11. LF

      Again, so a lot of the stuff we're talking about are essentially open problems, right? So-

    12. MM

      Yeah.

    13. LF

      ... i- if I ask a question, I don't mean to, uh, that you would know the answer-

    14. MM

      Right.

    15. LF

      I'm only just hypothesizing. But how big do you think is the, the, the network graph data structure of concepts that's in our head? Like, if we were trying to build that ourselves, like, it's, we take it, that's one of the things we take for granted, we think. I mean, that's why we take common sense for granted. We think common sense is trivial. But how big of a thing of concepts is on, that underlies what we think of as common sense, for example?

    16. MM

      Yeah, I don't know, and I, I'm not, I don't even know what units to measure it in. (laughs)

    17. LF

      (laughs)

    18. MM

      When you say, how big is it?

    19. LF

      That's beautifully put, right? What, uh...

    20. MM

      But, you know, we have, uh, you know, it's really hard to know. We have, uh, what, 100 billion neurons or something, I don't know, uh, and they're connected via trillions of synapses, and there's all this chemical processing going on. There, there's just a lot of capacity for (laughs) stuff. And their information's encoded in different ways in the brain. It's encoded in, uh, ch- chemical interactions, it's encoded in elec- electric, like, firing and firing rates. And, and nobody really knows how it's encoded, but it just seems like there's a huge amount of capacity. So I think it's, it's huge. It's just enormous, and it's amazing how much stuff we know.

    21. LF

      Yeah. And, and if we're-

    22. MM

      (laughs)

    23. LF

      But we know, and not just know, like, facts, but it's all integrated into this thing that we can make analogies with.

    24. MM

      Yes.

    25. LF

      There's a dream of semantic web. There's, there's a lot of dreams from expert systems of building giant knowledge bases. W- do you see a hope for these kinds of approaches of building, of converting Wikipedia into something that could be used in analogy-making?

    26. MM

      Uh, sure. And I think people have, have made some progress along those lines. I mean, people have been working on this for a long time. But the problem is, uh, and this, I think, was, is, is the problem of common sense. Like, people have been trying to get these common sense networks. Here at MIT, there's this ConceptNet project, right? Uh, but the problem is that, as I said, most of the knowledge that we have is in- invisible to us. It's not in Wikipedia. (laughs) It's very basic things about, you know, intuitive physics, intuitive psychology, intuitive metaphysics, all that stuff.

    27. LF

      If you were to create a website that described intuitive physics, intuitive psychology, would it be bigger or smaller than Wikipedia? What do you think?

    28. MM

      I guess describe to whom? Uh... (laughs)

    29. LF

      (laughs)

    30. MM

      I'm sorry, but it, it-

  10. 55:33 – 1:09:07

    Limits of feedforward deep learning: attention, feedback, transfer, and the “paddle moved” problem

    1. LF

      But sort of these analogies are very human interpretable.

    2. MM

      Mm-hmm.

    3. LF

      So that's that kind of space, and then you look at something like, uh, the current deep learning approaches, they kinda help you to take raw sensory information and- and to sort of automatically build up hierarchies of- of- of, well, you can even call them concepts. They're just not human interpretable concepts. What- what's your- what's the link here? Do- do you hope... a- eh, sort of the hybrid system question, how do you think the two can start to meet each other? What's the value of learning in this systems of forming of analogy making?

    4. MM

      The- the- the goal of I- you know, the original goal of deep learning in- in at least visual perception was that you would get the system to learn to extract features that- at these different levels of complexities, you know, maybe edge detection-

    5. LF

      Mm-hmm.

    6. MM

      ... and that would lead into learning, you know, simple combinations of edges and then more complex shapes and then whole objects or faces. Uh, and this was based on the- the ideas of the neuroscientists Hubel and Wiesel, who had seen-

    7. LF

      Mm-hmm.

    8. MM

      ... laid out this kind of structure in brain. Um... and I think that is- that's right to some extent. Of course, people have come- found that the whole story is a little more complex than that.

    9. LF

      Yeah.

    10. MM

      And the brain, of course, always is, and there's a lot of feedback and... Um, so I see that as- as absolutely a- a- a good brain-inspired approach to some aspects of perception. But one thing that it's lacking, for example, is all of that feedback, which is extremely important.

    11. LF

      The interactive element that you mentioned, uh...

    12. MM

      The expectation, right?

    13. LF

      The- the-

    14. MM

      The conceptual level.

    15. LF

      ... going back and forth with the- the expectation, the perception-

    16. MM

      Yeah.

    17. LF

      ... and just going back and forth.

    18. MM

      So, right. So that is extremely important. And, you know, one thing about deep neural networks is that in a given situation, like, you know, they- they're trained, right? They get these weights and everything, but then now I give them a new, uh, a new image let's say.

    19. LF

      Yes.

    20. MM

      They treat every part of the image in the same way. You know, they apply the same filters at each layer to all parts of the image.

    21. LF

      Mm-hmm.

    22. MM

      There's no feedback to say like, "Oh, this part of the image is irrelevant."

    23. LF

      Right.

    24. MM

      "I shouldn't care about this part of the image," or- or, "This part of the image is the most important part." And that's kind of what we humans are able to do because we have these con- conceptual expectations.

    25. LF

      The- there's a... by the way, a little bit work in that. There's certainly a lot more in atten- what- what's under the- called attention-

    26. MM

      Attention.

    27. LF

      ... in natural language processing knowledge is- it's an int- it's a... a- and that's exceptionally powerful, and- and it's a very, just as you say, it's really powerful idea. But again, in sort of machine learning, it all kind of operates in an automated way that's not human inter-

    28. MM

      Well, it's not- it's not also... Okay, so the... yeah. Right. It's not dynamic. I mean, in the sense that as a perception of a new example is being processed, those attentions weights don't change.

    29. LF

      Right. So, I mean, there's a- there's a kind of notion that there's not a memory, so you're not aggregating... The- the idea of like this mental model idea.

    30. MM

      Yes.

  11. 1:09:07 – 1:20:21

    Autonomous driving as a common-sense test: long-tail edge cases and social interaction

    1. LF

      Wow. So just for fun, let me, uh, ask you on the topic of autonomous vehicles. It's the area that, um, that I work at- at least these days most closely on, and it's also area that I think is a good example that you use as sort of, uh, an example of things we as humans don't always realize how hard it is to do. It's like the, the constant trend in AI, but the different problems that we think are easy when we first try them, and then we realize how hard it is. Okay. So, why ... You've talked about this a- autonomous driving being a difficult problem, more difficult than we realize, humans give it credit for. Why is it so difficult? What are the most difficult parts in your view?

    2. MM

      I think it's difficult because of the world is so open-ended a- as to wha- what kinds of things can happen. So, you have sort of what normally happens, which is just you drive along and nothing, nothing surprising happens, and autonomous vehicles can do, the ones we have now evidently can do really well on most normal situations as long, as long as, you know, the weather is reasonably good and everything. Um, but if some ... We have this notion of edge case or, or, you know, things in the tail of the distribution, we call it the long-tail problem, which says that there's so many possible things that can happen that w- was not in the training data of the machine that it won't be able to handle it, because it doesn't have common sense.

    3. LF

      Right. It's the old the paddle moved problem. (laughs)

    4. MM

      Yeah. It's the paddle moved problem. Right. And so my understanding, and you probably are more of an expert than I am on this, is that s- current self-driving car v- vision systems have problems with obstacles, meaning that they don't know which obstacles, which, quote/unquote, "obstacles" they should stop for and which ones they shouldn't stop for.

    5. LF

      Mm-hmm.

    6. MM

      And so a- a lot of times I read that they tend to slam on the brakes quite a bit, and the most common accidents with self-driving cars are people rear-ending them, 'cause they were surprised, they weren't expecting the machine, the, the car to stop.

    7. LF

      Yeah, so th- there's, there's a lot of interesting questions there. Uh, whether ... 'cause, 'cause you mentioned kind of two things. So, so one is the, the problem of perception, of understanding, of interpreting the objects that are detected ...

    8. MM

      Right.

    9. LF

      ... correctly. Right? And the other one is more like the, the policy, the, the action that you take or how you respond to it. So a lot of the cars breaking is a kind of notion of, uh, to clarify, there's a lot of different kind of things that people are calling autonomous vehicles, but a lot ... The L4 vehicles with a safety driver are the ones like Waymo and Cruise and all those companies, they tend to be very conservative and cautious, so they tend to be very, very afraid of hurting anything or anyone and getting in any kind of accidents, so their policy is very kind of ... That, it, that results in being exceptionally responsive to anything that could possibly be an obstacle, right?

    10. MM

      Right, which, which, which the human drivers around it, i- i- it's unpredict- it, it behaves unpredictably.

    11. LF

      Yeah, that's not a very human thing to do, caution.

    12. MM

      Yeah.

    13. LF

      That's not the thing we're good at, especially in driving.

    14. MM

      (laughs)

    15. LF

      We're in a hurry, often angry and et cetera, especially in Boston. So, a- and then there's sort of another ap- and, and a lot of times, that's machine learning is not a huge part of that. It's becoming more and more unclear to me how much e- you know, sort of speaking to public information, because a lot of the companies say they're doing deep learning and machine learning just to attract good candidates.

    16. MM

      (laughs)

    17. LF

      Uh, the reality is in many cases, it's still not a huge part of the, of the perception. There's, uh, there's Lidar and there's other sensors that are much more reliable for obstacle detection. And then there's Tesla approach, which is vision only ...

    18. MM

      Mm-hmm.

    19. LF

      And there's, I- I think a few companies doing that, but Tesla most, sort of famously pushing that forward.

    20. MM

      And that's because the LIDAR is too expensive, right?

    21. LF

      Well, I, I mean, yes, but I would say if you were to, for free, give to every Tesla vehicle... I mean, Elon Musk fundamentally believes that LIDAR is a crutch, right? He famously-

    22. MM

      Mm-hmm.

    23. LF

      ... said that. That if you want to solve the problem with machine learning, LIDAR is not, should not be the primary sensor, is the belief.

    24. MM

      Okay.

    25. LF

      Uh, the camera contains a lot more information.

    26. MM

      Mm-hmm.

    27. LF

      So if- so if you wanna learn, you want that information. But if you wanted not to hit obstacles-

    28. MM

      (laughs)

    29. LF

      ... you want LIDAR.

    30. MM

      (laughs)

  12. 1:20:21 – 1:36:11

    Embodiment, emotion, and AI risk: why superintelligence may be a confused concept

    1. LF

      So, you've mentioned embodied intelligence. Wh- what do you think it takes to build a system of human level intelligence? Does it need to have a body?

    2. MM

      I'm not sure, but I f- I'm coming around to that more and more. (laughs)

    3. LF

      And what does it mean to be, I don't mean to keep bringing, uh, up Yann LeCun, but...

    4. MM

      (laughs) He looms very large. (laughs)

    5. LF

      (laughs) Well, he, he certainly has a large personality, yes. Uh, he thinks that the system needs to be grounded, meaning it needs to sort of be able to interact with reality, but doesn't think it necessarily needs to have a body. So, when you think of-

    6. MM

      So, what's the difference?

    7. LF

      I guess I, I wanna ask, when you mean body, do you mean you have to be able to play with the world? Or do you also mean, like, there's a body that you, that you have to preserve?

    8. MM

      Oh, that's a good question. I haven't really thought about that. But, um, I think both I would guess 'cause, it's because I think you, I think intelligence, it's so hard to, to separate it from our f- self, our f- desire for self-preservation, our emotions, our, all that non-rational stuff that kinda gets in the way of l- logical thinking. (laughs) Because we, the way, you know, if we're talking about human intelligence or human level intelligence, whatever that means, uh, a huge part of it is social. That, you know, we, we were evolved to be social and to deal with other people, and that's just so ingrained in us, um, that it's hard to separate intelligence from that. I, I think, you know, AI f- for the last 70 years or however long it's been around, it, it has largely been separated. There's this idea that there's like, it's kind of very, uh, cartesian. There's this, you know, thinking thing-

    9. LF

      Yeah.

    10. MM

      ... that we're trying to create, but we don't care about all this other stuff.

    11. LF

      Mm-hmm.

    12. MM

      And I think the other stuff is very fundamental.

    13. LF

      So, the, there's an idea that things like emotion get in the way of intelligence.

    14. MM

      A- as opposed to being an integral part of it.

    15. LF

      Integral part of it. So-

    16. MM

      Yeah.

    17. LF

      ... I mean, I'm Russian, so romanticize the notions of emotion and suffering and all that kinda, uh, fear of mortality, those kinds of things. So, uh...

    18. MM

      (laughs)

    19. LF

      (laughs) In AI, uh, especially, sort of have, uh-

    20. MM

      By the way, did you see that, there was this recent thing going around the internet, uh, this, some, some, I think he's a Russian or some Slavic had, had written this thing ab- sort of anti the idea of super intelligence?

    21. LF

      Mm.

    22. MM

      I forgot, maybe he's Polish. Anyway, so he had all these arguments, and one w- one was the argument from Slavic pessimism. (laughs)

    23. LF

      (laughs)

    24. MM

      Which, my favorite. (laughs)

    25. LF

      Uh, do you remember what the argument is? Just-

    26. MM

      Uh, it's like, nothing ever works.

    27. LF

      Yeah. (laughs)

    28. MM

      (laughs) Everything's, everything sucks. (laughs)

    29. LF

      (laughs) So, what, what do you think is the role, like, that's such a fascinating idea that, uh, the l- what we perceive as sort of the limits of human, of the human mind, which is emotion and, uh, fear and all those kinds of things are integral to intelligence. Could y- could you, um, elaborate on that? Like, what, why is that important do you think, for human level intelligence?

    30. MM

      At least for the way that humans work, it's a big part of how, it affects how we perceive the world.

Episode duration: 1:52:39

Install uListen for AI-powered chat & search across the full episode — Get Full Transcript

Transcript of episode ImKkaeUx1MU

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

iOSAndroidClaudeChrome