#110 – Jitendra Malik: Computer Vision

Jitendra Malik is a professor at Berkeley and one of the seminal figures in the field of computer vision, the kind before the deep learning revolution, and the kind after. He has been cited over 180,000 times and has mentored many world-class researchers in computer science. Support this podcast by supporting our sponsors: – BetterHelp: http://betterhelp.com/lex – ExpressVPN: https://www.expressvpn.com/lexpod If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter https://twitter.com/lexfridman, LinkedIn https://www.linkedin.com/in/lexfridman/, Facebook https://www.facebook.com/lexfridman, Medium https://medium.com/@lexfridman, or YouTube https://www.youtube.com/lexfridman where you can watch the video versions of these conversations. If you enjoy the podcast, please rate it 5 stars on Apple Podcasts https://podcasts.apple.com/us/podcast/artificial-intelligence/id1434243584, follow on Spotify https://open.spotify.com/show/2MAi0BvDc6GTFvKFPXnkCL, or support it on Patreon https://www.patreon.com/lexfridman. Here’s the outline of the episode. On some podcast players you should be able to click the timestamp to jump to that time. OUTLINE: 00:00 – Introduction 03:17 – Computer vision is hard 10:05 – Tesla Autopilot 21:20 – Human brain vs computers 23:14 – The general problem of computer vision 29:09 – Images vs video in computer vision 37:47 – Benchmarks in computer vision 40:06 – Active learning 45:34 – From pixels to semantics 52:47 – Semantic segmentation 57:05 – The three R’s of computer vision 1:02:52 – End-to-end learning in computer vision 1:04:24 – 6 lessons we can learn from children 1:08:36 – Vision and language 1:12:30 – Turing test 1:16:17 – Open problems in computer vision 1:24:49 – AGI 1:35:47 – Pick the right problem

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Lex Fridman?

  The following is a conversation with Jitendra Malik, a professor at Berkeley and one of the seminal figures in the field of computer vision, the kind before the deep learning revolution and the kind after. He has been cited over 180,000 times and has mentored many world-class researchers in computer science. Quick summary of the ads. Two sponsors, one new one, which is BetterHelp, and an old goodie, ExpressVPN. Please consider supporting this podcast by going to betterhelp.com/lex and signing up at expressvpn.com/lexpod. Click the links, buy the stuff.

• 0:43Lex Fridman?

  It really is the best way to support this podcast and the journey I'm on. If you enjoy this thing, subscribe on YouTube, review it with five stars on Apple Podcasts, support it on Patreon, or connect with me on Twitter at LexFriedman, however the heck you spell that. 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 show is sponsored by BetterHelp, spelled H-E-L-P, help. Check it out at betterhelp.com/lex. They figure out what you need and match you with a licensed professional therapist in under 48 hours. It's not a crisis line, it's not self-help, it's professional counseling done securely online.

• 1:28Lex Fridman?

  I'm a bit from the David Goggins line of creatures, as you may know, and so have some demons to contend with, usually on long runs or all nights working, forever and possibly full of self-doubt. It may be because I'm Russian, but I think suffering is essential for creation. But I also think you can suffer beautifully in a way that doesn't destroy you. For most people, I think a good therapist can help in this, so it's at least worth a try. Check out their reviews, they're good, it's easy, private, affordable, available worldwide. You can communicate by text any time and schedule weekly audio and video sessions.

• 2:09Lex Fridman?

  I highly recommend that you check them out at betterhelp.com/lex. This show is also sponsored by ExpressVPN. Get it at expressvpn.com/lexpod to support this podcast and to get an extra three months free on a one-year package. I've been using ExpressVPN for many years. I love it. I think ExpressVPN is the best VPN out there. They told me to say it, but it happens to be true. It doesn't log your data. It's crazy fast and is easy to use. Literally just one big sexy power on button. Again, for obvious reasons, it's really important that they don't log your data. It works on Linux and everywhere else too.

• 2:54Lex Fridman?

  But really, why use anything else? Shout out to my favorite flavor of Linux, Ubuntu MATE 2004. Once again, get it at expressvpn.com/lexpod to support this podcast and to get an extra three months free on a one-year package.

• 3:13Lex Fridman?

  And now here's my conversation with Jitendra Malik.

• 3:35Lex Fridman?

  In 1966, Seymour Papert at MIT wrote up a proposal called the Summer Vision Project to be given, as far as we know, to 10 students to work on and solve that summer. So that proposal outlined many of the computer vision tasks we still work on today. Why do you think we underestimate? And perhaps we did underestimate and perhaps still underestimate how hard computer vision is.

• 4:03Speaker 01

  Because most of what we do in vision, we do unconsciously or subconsciously.

• 4:08Lex Fridman?

  In human vision.

• 4:09Speaker 01

  In human vision. So that gives us this, that effortlessness gives us the sense that, oh, this must be very easy to implement on a computer. Now, uh, this is why the early researchers in AI got it so wrong.

• 4:27Speaker 01

  However, if you go into neuroscience or psychology of, of human vision, then the complexity becomes very clear. The fact is that a very large part of the, the cerebral cortex is devoted to visual processing. I mean, and this is true in other primates as well. So, once we looked at it from a neuroscience or psychology perspective, it, it becomes quite clear that the problem is very challenging and it will take some time.

• 4:56Lex Fridman?

  You said the higher level parts are the harder parts?

• 5:01Speaker 01

  I think vision appears to, to be easy because, uh, most of what visual processing is subconscious or unconscious. Uh, so we underestimate the difficulty. Whereas, uh, when you are, uh, like proving a mathematical theorem or playing chess, the difficulty is much more evident. So, because it is your conscious brain which is processing, uh, various aspects of the problem solving behavior. Whereas in vision, uh, all this is happening, but it's not in your awareness. It's in your, it's operating below that.

• 5:43Lex Fridman?

  But it's, it still seems strange. Yes, that's true, but it seems strange that as computer vision researchers, for example, the community broadly is, time and time again, makes the mistake of, um, thinking the problem is easier than it is. Or maybe it's not a mistake. We'll talk a little bit about autonomous driving, for example, how hard of a vision task that is. It, it, do you think, I mean, what, is it just human nature? Or is there something fundamental to the vision problem that we, we underestimate? We're still not able to be cognizant of how hard the problem is.

• 6:23Speaker 01

  Yeah, I think in the early days it could have been excused because in the early days, all aspects of AI were regarded as too easy. But I think today it is much less excusable. And I think why people fall for this is because of what I call the fallacy of the successful first step. There are many problems in vision where getting fifty percent of the solution you can get in one minute, getting to ninety percent can take you a day, getting to ninety nine percent may take you five years, and ninety nine point nine nine percent may be not in your lifetime.

• 7:07Lex Fridman?

  I wonder if that's the unique to vision. It seems that language people are not so confident about, so natural language processing. People are a little bit more cautious about our ability to to solve that problem. I think for language, people intuit that we have to be able to do natural language understanding. For vision, it seems that we're not cognizant or we don't think about how much understanding is required. It's probably still an open problem. But in your sense, how much understanding is required to solve vision? Like this, put another way, how much something called common sense reasoning is required

• 7:52Lex Fridman?

  to really be able to interpret even static scenes?

• 7:56Speaker 01

  Yeah, so vision operates at all levels. And there are parts which are which can be solved with what we could call maybe peripheral processing. So in the in the human vision literature, there used to be these terms sensation, perception, and cognition, which roughly speaking referred to like the front end of processing, middle stages of processing, and higher level of processing. And I think they made a big deal out of out of this, and they wanted to just study only perception and then dismiss certain certain problems as being quote cognitive. But really, I think these are artificial divides.

• 8:40Speaker 01

  The problem is continuous at all levels and there are challenges at all levels. The techniques that we have today, they work better at the lower and mid levels of the problem. I think the higher levels of the problem, quote, the cognitive levels of the problem are there. And we, in many real applications, we have to confront them. Now, how much that is necessary will depend on the application. For some problems, it doesn't matter. For some problems, it matters a lot. So I am, for example, a pessimist on fully autonomous driving in the near future.

• 9:21Speaker 01

  And the reason is because I think there will be that 0.01% of the cases where quite sophisticated cognitive reasoning is called for. However, there are tasks where you can, first of all, they are much more, they are robust. So in the sense that error rates, error is not so much of a problem. For example, let's say you're doing image search. You're trying to get images based on some, some description, some visual description.

• 9:59Speaker 01

  We are very tolerant of errors there, right? I mean, when Google image search gives you some images back and a few of them are wrong, it's okay. It doesn't hurt anybody. There's no, there's not a matter of life and death. But making mistakes when you are driving at 60 miles per hour and you could potentially kill somebody is much more important.

• 10:23Lex Fridman?

  So just for the, for the fun of it, since you mentioned, let's go there briefly about autonomous vehicles. So one of the companies in the space, Tesla, is with Andre Karpathy and Elon Musk are working on a system called Autopilot, which is primarily a vision-based system with eight cameras and basically a single neural network, a multitask neural network. They call it Hydranet, multiple heads. So it does multiple tasks, but is forming the same representation at the core.

• 10:56Lex Fridman?

  Do you think driving can be converted in this way to purely a vision problem? And then solved with learning?

• 11:06Lex Fridman?

  Or even more specifically, in the current approach, what do you think about what Tesla Autopilot team is doing?

• 11:14Speaker 01

  So the way I think about it is that there are certainly subsets of the visual-based driving problem which are quite solvable. So for example, driving in freeway conditions is quite a solvable problem. I think there were demonstrations of that going back to the 1980s by someone called Ernst Stickmans in Munich.

• 11:39Speaker 01

  In the 90s, there were approaches from Carnegie Mellon, there were approaches from a team at Berkeley. In the 2000s, there were approaches from Stanford and so on. So autonomous driving in certain settings is very doable. The challenge is to have an autopilot work under all kinds of driving conditions. At that point, it's not just a question of vision or perception, but really also of control and dealing with all the edge cases.

• 12:11Lex Fridman?

  So where do you think most of the difficult cases? To me, even the highway driving is an open problem because it applies the same 50-90, 95-99 rule, or the first step, the fallacy of the first step, I forget how you put it, we fall victim to. I think even highway driving has a lot of elements because to solve autonomous driving, you have to completely relinquish the help of a human being. You're always in control, so you're really going to feel the edge cases. So I think even highway driving is really difficult. But in terms of the general driving task, do you think vision is the fundamental problem? Or is it also your action, the interaction with the environment?

• 13:00Lex Fridman?

  The ability to, and then like the middle ground, I don't know if you put that under vision, which is trying to predict the behavior of others, which is a little bit in the world of understanding the scene, but it's also trying to form a model of the actors in the scene and predict their behavior.

• 13:19Speaker 01

  Yeah, I include that in vision because to me, perception blends into cognition and building predictive models of other agents in the world, which could be other agents, could be people, other agents, could be other cars, that is part of the task of perception. Because perception always has to not tell us what is now, but what will happen. Because what's now is boring, it's done, it's over with. Okay.

• 13:46Lex Fridman?

  Yeah.

• 13:47Speaker 01

  We care about the future because we act in the future.

• 13:50Lex Fridman?

  And we care about the past in as much as it informs what's going to happen in the future.

• 13:56Speaker 01

  So I think we have to build predictive models of behaviors of people and those can get quite complicated. So, I mean, I've seen examples of this in, actually, I mean, I own a Tesla and it has various safety features built in. And what I see are these examples where, let's say, there is some skateboarder. I mean, and I don't want to be too critical because obviously, these are, the systems are always being improved and any specific criticism I have, maybe the system six months

• 14:38Speaker 01

  from now will not have that particular failure mode. So it had the wrong response and it's because it couldn't predict what this skateboarder was going to do. Okay. And because it really required that higher level cognitive understanding of what skateboarders typically do as opposed to a normal pedestrian. So what might have been the correct behavior for a pedestrian, a typical behavior for a pedestrian was not the typical behavior for a skateboarder.

• 15:15Speaker 01

  Right?

• 15:16Lex Fridman?

  Yeah.

• 15:17Speaker 01

  And so therefore, to do a good job there, you need to have enough data where you have pedestrians, you also have skateboarders, you've seen enough skateboarders to see what what kinds of patterns of behavior they have. So it is it is in principle with enough data, that problem could be solved. But I think our current systems, computer vision systems, they need far, far more data than humans do for learning those same capabilities.

• 15:51Lex Fridman?

  So say that there is going to be a system that solves autonomous driving. Do you think it will look similar to what we have today? But have a lot more data, perhaps more compute, but the fundamental architectures involved, like neural, well, in the case of Tesla Autopilot, is neural networks. Do you think it will look similar in that regard, and it'll just have more data?

• 16:14Speaker 01

  That's a scientific hypothesis as to which way is it going to go. I will tell you what I would bet on. So and this is my general philosophical position on how these learning systems have been. What we have found currently very effective in computer vision with in the deep learning paradigm is sort of tabula rasa learning and tabula rasa learning in a supervised way with lots and lots of.

• 16:45Lex Fridman?

  What's tabula rasa learning?

• 16:46Speaker 01

  Tabula rasa in the sense that blank slate. We just have the system which is given a series of experiences in this setting and then it learns there. Now, if let's think about human driving, it is not tabula rasa learning. So at the age of sixteen in high school.

• 17:06Speaker 01

  A teenager goes into driver ed class, right? And now, at that point, they learn. But at the age of sixteen, they are already visual geniuses. Because from zero to sixteen, they have built a certain repertoire of vision. In fact, most of it has probably been achieved by age two, right? In this period of age, up to age two, they know that the world is three dimensional. They know how objects look like from different perspectives. They know about occlusion. They know about common dynamics of humans and other bodies. They have some notion of intuitive physics.

• 17:49Speaker 01

  So they have built that up from their observations and interactions in early childhood and of course reinforced through their growing up to age sixteen. So then, at age sixteen, when they go into driver ed, what are they learning? They are not learning afresh the visual world. They have a mastery of the visual world. What they are learning is control, okay? They are learning how to be smooth about control, about steering and brakes and so forth. They are learning a sense of typical traffic situations. Now, that education process can be quite short because they are coming in as visual geniuses.

• 18:34Speaker 01

  And of course, in their future, they are going to encounter situations which are very novel, right? So during my driver ed class, I may not have had to deal with a skateboarder. I may not have had to deal with a truck driving in front of me who is from, where the back opens up and some junk gets dropped from the truck and I have to deal with it, right? But I can deal with this as a driver, even though I did not encounter this in my driver ed class. And the reason I can deal with it is because I have all this general visual knowledge and expertise.

• 19:11Lex Fridman?

  And do you think the learning mechanisms we have today can do that kind of long-term accumulation of knowledge? Or do we have to do some kind of, you know, the work that led up to expert systems with knowledge representation, you know, the broader field of what of artificial intelligence worked on this kind of accumulation of knowledge? Do you think neural networks can do the same?

• 19:39Speaker 01

  I think I don't see any in-principle problem with neural networks doing it. But I think the learning techniques would need to evolve significantly. So the current, the current learning techniques that we have are supervised learning. You're given lots of examples, X I Y I pairs, and you you learn the functional mapping between them. I think that human learning is far richer than that. It includes many different components. There are, there is, a child explores the world and sees us. For example, a child takes an object and manipulates it

• 20:23Speaker 01

  in his hand and therefore gets to see the object from different points of view. And the child has commanded the movement. So that's a kind of learning data, but the learning data has been arranged by the child. And this is a very rich kind of data. The child can do various experiments with the world. So, so there are many aspects of a sort of human learning, and these have been studied in, in child development by psychologists. And they, what they tell us is that supervised learning is a very small part of it. There are many different aspects of learning.

• 21:05Speaker 01

  And what we would need to do is to develop models of all of these and then train our systems in that, with that kind of protocol.

• 21:19Lex Fridman?

  So new, new methods of learning, that's some of which might imitate the human brain. But you also, in your talks, have mentioned sort of the compute side of things, the, in terms of the difference in the human brain or referencing Marvick, Hans Marvick. The, so. Do you think there's something interesting valuable to consider about the difference in the computational power of the human brain versus the computers of today? In terms of instructions per second?

• 21:52Speaker 01

  Yes, so if we go back, so this is a point I've been making for 20 years now.

• 21:58Lex Fridman?

  Yeah.

• 21:58Speaker 01

  And I think, once upon a time, the way I used to argue this was that we just didn't have the computing power of the human brain. Our computers were not quite there. And, I mean, there is a well-known trade-off, which we know that neurons are slow compared to transistors, but we have a lot of them and they have a very high connectivity. Whereas in silicon, you have much faster devices. Transistors switch at on the order of nanoseconds, but the connectivity is usually smaller.

• 22:37Lex Fridman?

  Right.

• 22:38Speaker 01

  At this point in time, I mean, we are now talking about 2020. We do have, if you consider the latest GPUs and so on, amazing computing power. And if we look back at Hans Moravec's type of calculations, which he did in the 1990s, we may be there today in terms of computing power comparable to the brain, but it's not of the same style. Right. It's of a very different style.

• 23:07Speaker 01

  So, I mean, for example, the style of computing that we have in our GPUs is far, far more power-hungry than the style of computing that is there in the human brain or other biological entities.

• 23:22Lex Fridman?

  Yeah, and that, the efficiency part is, we're gonna have to solve that in order to build actual real-world systems of large scale.

• 23:32Lex Fridman?

  Let me ask sort of the high level question, taking a step back. How would you articulate the general problem of computer vision? Does such a thing exist? So if you look at the computer vision conferences and the work that's been going on, it's often separated into different little segments, breaking the problem of vision apart into segmentation, 3D reconstruction, object detection, I don't know, image captioning, whatever. There's benchmarks for each. But if you were to sort of philosophically say, what is the big problem of computer vision? Does such a thing exist?

• 24:11Speaker 01

  Yes, but it's not in isolation. So if we have to... So for all intelligence tasks, I always go back to sort of biology or humans. And if we think about vision or perception in that setting, we realize that perception is always to guide action. Perception for a biological system does not give any benefits unless it is coupled with action. So we can go back and think about the first multicellular animals which arose in the Cambrian era, you know, 500 million years ago.

• 24:52Speaker 01

  And these animals could move and they could see in some way. And the two activities help each other. Because how does movement help? Movement helps that because you can get food in different places. But you need to know where to go. And that's really about perception or seeing. I mean, vision is perhaps the single most perception sense. But all the others are equally are also important. So perception and action kind of go together. So earlier, it was in these very simple feedback loops, which were about finding food or avoiding becoming food

• 25:36Speaker 01

  if there's a predator running, trying to, you know, eat you up and so forth. So we must, at the fundamental level, connect perception to action. Then, as we evolved, perception became more and more sophisticated because it served many more purposes. And so today we have what seems like a fairly general purpose capability, which can look at the external world and build a model of the external world inside the head. We do have that capability. That model is not perfect. And psychologists have great fun in pointing out the ways in which

• 26:17Speaker 01

  the model in your head is not a perfect model of the external world. They create various illusions to show the ways in which it is imperfect. But it's amazing how far it has come from a very simple perception-action loop that exists in, you know, an animal 500 million years ago. Once we have these very sophisticated visual systems, we can then impose a structure on them. It's we as scientists who are imposing that structure, where we have chosen to characterize this part of the system as this quote module of object detection, or quote this module of 3D reconstruction.

• 27:02Speaker 01

  What's going on is really all of these processes are running simultaneously. And they are running simultaneously because originally their purpose was in fact to help guide action.

• 27:18Lex Fridman?

  So as a guiding general statement of a problem, do you think we can say that the general problem of computer vision,

• 27:28Lex Fridman?

  you said in humans it was tied to action. Do you think we should also say that ultimately the goal, the problem of computer vision is to sense the world in a way that helps you act in the world?

• 27:44Speaker 01

  Yes, I think that's the most fundamental, that's the most fundamental purpose. We have by now hyper-evolved. So we have this visual system which can be used for other things. For example, judging the aesthetic value of a painting. And this is not guiding action. Maybe it's guiding action in terms of how much money you will put in your auction bid, but that's a bit stretched. But the basics are, in fact, in terms of action. But we have, we evolved really this hyper, we have hyper evolved our visual system.

• 28:25Lex Fridman?

  Actually, just to, sorry to interrupt, but perhaps it is fundamentally about action. You kind of jokingly said about spending, but perhaps the capitalistic drive that drives a lot of the development in this world is about the exchange of money. And the fundamental action is money. If you watch Netflix, if you enjoy watching movies, you're using your perception system to interpret the movie. Ultimately, your enjoyment of that movie means you'll subscribe to Netflix. So the action is this, this extra layer that we've developed in modern society perhaps is fundamentally tied to the action of spending money.

• 29:04Speaker 01

  Well, certainly with respect to, you know, interactions with firms. So, so in this homo economicus role, when you're interacting with firms, it does become, it does become that.

• 29:18Lex Fridman?

  What else is there?

• 29:22Lex Fridman?

  No, it was a rhetorical question. Okay. So to linger on the division between the static and the dynamic, so much of the work in computer vision, so many of the breakthroughs that you've been a part of have been in the static world, in looking at static images. And then you've also worked on, starting, but at a much smaller degree, the community's looking at dynamic, at video, at dynamic scenes. And then there is robotic vision, which is dynamic, but also where you actually have a robot in the physical world interacting based on that vision.

• 30:00Lex Fridman?

  Which problem is harder?

• 30:04Lex Fridman?

  The intuitive, sort of the trivial first answer is, well, of course, one image is harder. But if you look at a deeper question there, are we, what's the term? Cutting ourselves at the knees or like making the problem harder by focusing on images?

• 30:25Speaker 01

  That's a fair question. I think sometimes we can simplify a problem so much that we essentially lose part of the juice that could enable us to solve the problem. And one could reasonably argue that, to some extent, this happens when we go from video to single images. Now, historically, you have to consider the limits imposed by the computation capabilities we had. So if we, many of the choices made in the computer vision community

• 31:04Speaker 01

  through the 70s, 80s, 90s can be understood as choices which were forced upon us by the fact that we just didn't have access to compute, enough compute.

• 31:19Lex Fridman?

  Not enough memory, not enough hard drive.

• 31:21Speaker 01

  Exactly, not enough, not enough compute, not enough storage. So, so think of these choices. So one of the choices is focusing on single images rather than video. Okay, clear question, storage and compute.

• 31:36Speaker 01

  We had to focus on, we did, we used to detect edges and throw away the image. Right? So you have an image, which I say 256 by 256 pixels. And instead of keeping around the grayscale value, what we did was we detected edges, find the places where the brightness changes a lot. So now that's, and now, and then throw away the rest. So this was a major compression device. And the hope was that this makes it, that you can still work with it. And the logic was humans can interpret a line drawing.

• 32:10Speaker 01

  And, uh, and, uh, yes, and this will save us a computation. So many of the choices were dictated by that. I think today, uh, we are no longer detecting edges, right? We process images with conv nets because we don't need to. We don't have that, those compute restrictions anymore. Now, video is still understudied because video compute is still quite challenging. If you are a university researcher, I think video computing is not so challenging. If you are at Google or Facebook or Amazon.

• 32:46Lex Fridman?

  Still super challenging. I've, I just spoke with the VP of Engineering, Google head of YouTube Search and Discovery, and they still struggle doing stuff on video. It's very difficult except doing, except using techniques that are essentially the techniques you used in the nineties, some very basic computer vision techniques.

• 33:05Speaker 01

  No, I, that, that's when you want to do things at scale. So if, if you want to operate at the scale of all the content of YouTube, it's very challenging. And there are similar issues in Facebook. But as a researcher, you, you have, you have more, uh, you know, opportunities.

• 33:23Lex Fridman?

  You can train large, you know, networks with relatively large, uh, video data sets. Yeah.

• 33:27Speaker 01

  Yes. So I think that this is part of the reason why we have so emphasized static images. I think that this is changing and over the next few years, I see a lot more progress happening in, in video. So I, I have this generic statement that, uh, to me, video recognition feels like ten years behind object recognition. And you can quantify that because you can take some of the challenging video data sets and their performance on action classification is like, say, thirty percent, which is kind of what we used to have around, uh, two thousand nine in object detection, you know? It's like about ten years behind.

• 34:11Speaker 01

  And, uh, whether it'll take ten years to catch up is a different question. Hopefully it will take less than that.

• 34:19Lex Fridman?

  Let me ask a similar question I've already asked, but once again, so for dynamic scenes, do you think, do you think some kind of injection of knowledge bases and reasoning is required to help improve like action recognition?

• 34:37Lex Fridman?

  Like if, if, if, um,

• 34:41Lex Fridman?

  if we solve the general action recognition problem, what do you think the solution would look like? Is another way to put it.

• 34:48Speaker 01

  So I, I completely agree that knowledge is called for and that knowledge can be quite sophisticated. So the way I would say it is that perception blends into cognition and cognition brings in issues of memory and

• 35:06Speaker 01

  this notion of a schema from psychology, which is, let me use the classic example, which is you go to a restaurant, right? Now there are things that happen in a certain order. You walk in, somebody takes you to a table, waiter comes, gives you a menu, takes the order, food arrives, eventually a bill arrives, et cetera, et cetera. This is a classic example of AI from the 1970s. It was called, there was the term frames and scripts and schemas. These are all quite similar ideas. Okay. And there, in the seventies, the way the AI of the time dealt with it was by hand coding this.

• 35:51Speaker 01

  So they hand coded in this notion of a script and the various stages and the actors and so on and so forth. And use that to interpret, for example, language. I mean, if there's a description of a, of a story involving some people eating at a restaurant, there are all these inferences you can make because you know what happens typically at a restaurant. So I think this kind of, this kind of knowledge is absolutely essential. So I think that when we are going to do long form video understanding, we are going to need to do this. I think the kinds of technology that we have right now

• 36:33Speaker 01

  with 3D convolutions over a couple of second of clip or video, it's very much tailored towards short term video understanding, not that long term understanding. Long term understanding requires a notion of, this notion of schemas that I talked about, perhaps some notions of goals, intentionality, functionality and so on and so forth. Now, how will we bring that in? So we could either revert back to the seventies and say, okay, I'm going to hand code in a script, or we might try to learn it.

• 37:13Speaker 01

  So I tend to believe that we have to find learning ways of doing this. Because I think learning ways land up being more robust. And there must be a learning version of the story because children acquire a lot of this knowledge by sort of just observation. So at no moment in a child's life does a it's possible, but I think it's not so typical that somebody that a mother coaches a child through all the stages of what happens in a restaurant. They just go as a family, they they go to the restaurant, they eat, come back, and the child goes through ten such experiences.

• 37:55Speaker 01

  And the child has has got a schema of what happens when you go to a restaurant. So we somehow need to we need to provide that capability to our systems.

• 38:05Lex Fridman?

  You mentioned the following line from the end of the Alan Turing paper, Computing Machinery and Intelligence, that many people, like you said, many people know and very few have read, where he proposes the Turing test. This is, this is how you know, because it's towards the end of the paper. Instead of trying to produce a program to simulate the adult mind, why not rather try to produce one which simulates the child's? So that's a really interesting point. And if I think about the benchmarks we have before us, the the tests of our computer vision systems, they're often kind of trying to get to the adult. So what kind of benchmarks should we have?

• 38:48Lex Fridman?

  What kind of tests for computer vision do you think we should have that mimic the child's? In computer vision.

• 38:55Speaker 01

  Yeah, I I think we should have those and we don't have those today. And I think the part of the the challenge is that we should really be collecting data of the type that a child that a child experiences. Right, so that gets into issues of, you know, privacy and so on and so forth. But there are attempts in this direction to sort of try to collect the kind of data that a child encounters growing up. So what's the child's linguistic environment? What's the child's visual environment? So if we could collect that kind of data and then develop learning schemes based on that data, that would be one way to do it.

• 39:42Speaker 01

  I I I think that's a very promising direction myself. There might be people who would argue that we could just short circuit this in some way. And sometimes we have imitated. We have not, we have had success by not imitating nature in detail. So the usual example is airplanes, right? We don't build flapping wings. So, yes, that's that's one of the points of debate. In my mind, I I I would I would bet on this this learning like a child approach.

• 40:22Lex Fridman?

  So one of the fundamental aspects of learning like a child is the interactivity. So the child gets to play with the data set it's learning from.

• 40:31Speaker 01

  Yes.

• 40:32Lex Fridman?

  So it gets to select. I mean, you can call that active learning. You can, in the machine learning world, you can call it a lot of terms. What are your thoughts about this whole space of being able to play with the data set or select what you're learning?

• 40:46Speaker 01

  Yeah, so I think that I I believe in that. And I think that we could achieve it in in two ways, and I think we should use both. So one is actually real robotics, right? So real, you know, physical embodiments of agents who are interacting with the world and they have a physical body with dynamics and mass and moment of inertia and friction and all the rest. And you learn your body. The robot learns its body by doing a series of. Actions. The second is that simulation environments.

• 41:28Speaker 01

  So I think simulation environments are getting much, much better. In my, in my life, in Facebook AI research, our group has worked on something called Habitat, which is a simulation environment, which is a visually photorealistic environment of, you know, places like houses or interiors of various urban spaces and so forth. And as you move, you get a picture, which is a pretty accurate picture. So I, I can now, you can imagine that subsequent generations of these simulators

• 42:10Speaker 01

  will be accurate, not just visually, but with respect to, you know, forces and masses and haptic interactions and so on. And then, then we have that environment to play with. I think that, that, let me state one reason why I think this active, being able to act in the world is important. I think that this is one way to break the correlation versus causation barrier. So this is something which is of a great deal of interest these days. I mean, people like Judea Pearl have talked a lot about that we are neglecting causality

• 42:52Speaker 01

  and he describes the entire set of successes of deep learning as just curve fitting, right? Because it's, but I, I don't quite agree.

• 43:02Lex Fridman?

  He's a troublemaker, he is.

• 43:03Speaker 01

  But causality is important, but causality is not, is not like a single silver bullet. It's not like one single principle. There are many different aspects here. And one of the ways in which, one of our most reliable ways of establishing causal links, and this is the way, for example, the, the medical community does this is randomized controlled trials. So you have, you, you pick some situation and now in some situation you perform an action and for certain others you don't, right? So, so you have a controlled experiment. Well, the child is in fact performing controlled experiments all the time, right?

• 43:46Lex Fridman?

  Right, right.

• 43:47Speaker 01

  Okay.

• 43:48Lex Fridman?

  Small scale.

• 43:49Speaker 01

  And in a small scale and, but, but that is a way that the child gets to build and refine its causal models of the world. And my colleague Alison Gopnik has, together with a couple of authors, co-authors, has this book called The Scientist in the Crib, referring to his children. So I like, the part that I like about that is the scientist wants to do, wants to build causal models and the scientist does control experiments. And I think the child is doing that. So to enable that, we will need to have these, these active experiments. And I think this could be done, some in the real world and some in simulation.

• 44:32Lex Fridman?

  So you have hope for simulation?

• 44:33Speaker 01

  I have hope for simulation.

• 44:35Lex Fridman?

  That's an exciting possibility if we can get to not just photorealistic, but what's that called? Life realistic?

• 44:42Speaker 01

  Yeah.

• 44:43Lex Fridman?

  Simulation. So you don't see any fundamental blocks to why we can't eventually simulate the, the principles of what it means to exist in the world as a physical agent?

• 44:56Speaker 01

  I, I don't see any fundamental problems there. I mean, and, and look, the computer graphics community has come a long way.

• 45:02Lex Fridman?

  Right.

• 45:03Speaker 01

  So the, in the early days back, going back to the eighties and nineties, they were, they were focusing on visual realism, right? And then they could do the easy stuff, but they couldn't do stuff like hair or fur and so on. Okay, well, they managed to do that. Then they couldn't do physical actions, right? Like there's a bowl of glass and it falls down and it shatters. But then they could start to do pretty realistic models of that. And so on and so forth. So the graphics people have shown that they can do this forward direction, not just for optical interactions, but also for physical interactions. So I think, of course, some of that is very compute intensive, but I think by and by we will find ways of

• 45:49Speaker 01

  making our models ever more realistic.

• 45:52Lex Fridman?

  You break vision apart into, in one of your presentations, early vision, static scene understanding, dynamic scene understanding, and raise a few interesting questions. I thought I could just throw some at you to see if you want to talk about them. So early vision, so it's, what is it that you said, sensation, perception and cognition. So is this a sensation?

• 46:16Speaker 01

  Yes.

• 46:18Lex Fridman?

  What can we learn from image statistics that we don't already know? So at the lowest level, what, what can we make from just this, the statistics, the basics, the variations in the raw pixels, the textures and so on?

• 46:36Speaker 01

  Yeah. So what we seem to have learned is, is that there's a lot of redundancy in these images. And as a result, we are able to do a lot of compression. And this compression is very important in biological settings, right? So you might have 10 to the eight photoreceptors and only 10 to the six fibers in the optic nerve. So you have to do this compression by a factor of 100 is to one. And, and so there are analogs of that which are happening in, in our neural, artificial neural network.

• 47:12Lex Fridman?

  As the early layer. So you think

• 47:13Speaker 01

  At the early layers.

• 47:14Lex Fridman?

  There's a lot of compression that can be done in the beginning.

• 47:18Speaker 01

  Yeah.

• 47:18Lex Fridman?

  Just, just the statistics.

• 47:19Speaker 01

  Yeah.

• 47:22Lex Fridman?

  How much?

• 47:24Speaker 01

  Well, I so I mean, the the way to think about it is is just how successful is image compression, right? And we we and there are and and that's been done with older technologies, but it can be done with there are several companies which are trying to use sort of these more advanced neural network type techniques for compression, both for static images as well as for for video. One of my former students has a company which is trying to do stuff like this. And I think I think that they are showing quite interesting results. And I think that that's all the success of that's really about image statistics and

• 48:09Speaker 01

  video statistics.

• 48:10Lex Fridman?

  But that's still not doing compression of the kind. When I see a picture of a cat, all I have to say is it's a cat. That's another semantic kind of compression.

• 48:19Speaker 01

  Yeah, so this is this is at the lower level, right? So we are we are, as I said, yeah, that's focusing on low level statistics.

• 48:27Lex Fridman?

  So to linger on that for a little bit, you mentioned, how far can bottom up image segmentation go? And in general, what you mentioned that the central question for scene understanding is the interplay of bottom up and top down information. Maybe this is a good time to elaborate on that, maybe define what is what is bottom up, what is top down in the context of computer vision?

• 48:54Speaker 01

  Right. That's so today what we have are are very interesting systems because they work completely bottom up. How are they?

• 49:04Lex Fridman?

  What does bottom up mean, sorry?

• 49:05Speaker 01

  So bottom up means, in this case, means a feed forward neural network.

• 49:08Lex Fridman?

  So starting from the raw pixels.

• 49:10Speaker 01

  Yeah, they start from the raw pixels and they they end up with some something like cat or not a cat, right? So our our systems are running totally feed forward. They're trained in a very top down way. So they're trained by saying, OK, this is a cat, this is a cat, this is a dog, this is a zebra, et cetera.

• 49:31Speaker 01

  And I'm not happy with either of these choices fully. We have gone into, because we have completely separated these processes, right? So there is a, so I would like the, the process, the, so what do we know compared to biology? So in biology, what we know is that the processes in, at test time, at runtime, those processes are not purely feed forward, but they involve feedback. So, and they involve much shallower neural networks. So the kinds of neural networks we are using in computer vision, say a ResNet fifty, has fifty layers.

• 50:13Speaker 01

  Well, in, in the brain, in the visual cortex, going from the retina to IT, maybe we have like seven. Right? So they are far shallower, but we have the possibility of feedback. So there are backward connections. And this might enable us to, to deal with the more ambiguous stimuli, for example. So the, the biological solution seems to involve feedback. The solution in, in artificial vision seems to be just feed forward, but with a much deeper network. And the two are functionally equivalent, because if you have a feedback network, which just has like three rounds of feedback, you can just unroll it and make it three times the depth

• 50:58Speaker 01

  and create it in a totally feed forward way. So this is something which, I mean, we have written some papers on this theme, but I really feel that this should, this theme should be pursued further.

• 51:12Lex Fridman?

  Some kind of recurrence mechanism.

• 51:14Speaker 01

  Yeah. Okay, the other, so that's, so I, so I want to have a little bit more top down in the, at test time. Okay, then at training time, we make use of a lot of top down knowledge right now. So basically, to learn to segment an object, we have to have all these examples of, this is the boundary of a cat, and this is the boundary of a chair, and this is the boundary of a horse, and so on. And this is too much top down knowledge. How do hu- humans do this? We manage to, we manage with far less supervision. And we do it in a sort of bottom-up way. Because, for example, we're looking at a video stream and the horse moves.

• 51:59Speaker 01

  And that enables me to say that all these pixels are together.

• 52:04Lex Fridman?

  Yeah.

• 52:04Speaker 01

  So the gestalt psychologists used to call this the principle of common fate. So there was a bottom-up process by which we were able to segment out these objects. And we have totally focused on this top-down training signal. So in my view, we have currently solved it in machine vision, this top-down, bottom-up interaction. But I don't find the solution fully satisfactory. And I would rather have a bit of both in, at both stages.

• 52:37Lex Fridman?

  For all computer vision problems, not just segmentation.

• 52:41Speaker 01

  And, and, and, and the question that you can ask is, so for me, I'm inspired a lot by human vision and I care about that. You could be a, just a hard-boiled engineer, not give a damn. So to you, I would then argue that you would need far less training data if you could make my research agenda, you know, fruitful.

• 53:05Lex Fridman?

  Okay, so then maybe taking a step into segmentation, static scene understanding. What is the interaction between segmentation and recognition? You mentioned the movement of objects. So for people who don't know computer vision, segmentation is this weird activity that we, that computer vision folks have all agreed is very important. Of drawing outlines around objects versus a bounding box or, and then classifying that object. What's, what's the value of segmentation? What is it as a problem in computer vision? How is it fundamentally different from detection, recognition and the other problems?

• 53:48Speaker 01

  Yeah, so I think, so, so segmentation enables us to say.

• 53:56Speaker 01

  That some set of pixels are an object without necessarily even being able to name that object or knowing properties of that object.

• 54:05Lex Fridman?

  Oh, so you mean segmentation purely as the act of separating an object

• 54:12Speaker 01

  From its background.

• 54:12Lex Fridman?

  A blob of that's united in some way from its background.

• 54:18Speaker 01

  Yeah, so entitification, if you will, making an entity out of it.

• 54:22Lex Fridman?

  Entitification.

• 54:23Speaker 01

  Yeah.

• 54:23Lex Fridman?

  Beautifully.

• 54:24Speaker 01

  So I think that we have that capability and that enables us to, as we are growing up, to acquire names of objects with very little supervision. So suppose the child, let's posit that the child has this ability to separate out objects in the world. Then when the mother says, pick up your bottle, or the cat's behaving funny today.

• 54:58Speaker 01

  The word cat suggests some object and then the child sort of does a mapping.

• 55:03Lex Fridman?

  Right.

• 55:04Speaker 01

  Right? The mother doesn't have to teach specific object labels by pointing to them.

• 55:12Speaker 01

  Weak supervision works in the context that you have the ability to create objects. So I think that, so to me, that's a very fundamental capability. There are applications where this is very important, for example, medical diagnosis. So in medical diagnosis, you have some brain scan. I mean, this is some work that we did in my group where you have CT scans of people who have had traumatic brain injury and what the radiologist needs to do is to precisely delineate various places where there might be bleeds, for example. And there's, there are clear needs like that.

• 55:57Speaker 01

  So there's certainly very practical applications of computer vision where segmentation is necessary. But philosophically, segmentation enables the task of recognition to proceed with much weaker supervision than we require today.

• 56:15Lex Fridman?

  And you think of segmentation as this kind of task. That takes on a visual scene and breaks it apart into into interesting entities.

• 56:25Speaker 01

  Yeah.

• 56:25Lex Fridman?

  That might be useful for whatever the task is.

• 56:28Speaker 01

  Yeah. And and it is not semantics free. So I think I I I mean, it it blends into it involves perception and cognition. It is not, it is not, I I think the mistake that we used to make in the early days of computer vision was to treat it as a purely bottom-up perceptual task. It is not just that. Because we do revise our notion of segmentation with more experience, right? Because, for example, there are objects which are non-rigid, like animals or humans. And I think understanding that all the pixels of a human are one entity is actually quite a challenge.

• 57:11Speaker 01

  Because the parts of the human, they can move independently. The human wears clothes, so they might be differently colored. So it's all sort of a challenge.

• 57:22Lex Fridman?

  You mentioned the three Rs of computer vision are recognition, reconstruction, reorganization. Can you describe these three Rs?

• 57:31Speaker 01

  Sure.

• 57:31Lex Fridman?

  And how they interact?

• 57:32Speaker 01

  Yeah, so so recognition is the easiest one because that's what I think people generally think of as computer vision achieving these days, which is labels. So is this a cat? Is this a dog? Is this a chihuahua? I mean, you know, it could be very fine-grained, like, you know, a specific breed of a dog or a specific species of bird, or it could be very abstract, like animal.

• 58:04Lex Fridman?

  But given a part of an image or a whole image, say, put a label on that.

• 58:08Speaker 01

  Yeah.

• 58:09Lex Fridman?

  That's recognition.

• 58:10Speaker 01

  So that's that's recognition. Reconstruction is essentially.

• 58:17Speaker 01

  You can think of it as inverse graphics. I mean, that's one way to think about it. So graphics is you have some internal computer representation. And you have a computer representation of some objects arranged in a scene. And what you do is you produce a picture. You produce the pixels corresponding to a rendering of that scene. So let's do the inverse of this. We are given an image and we try to, we say, oh, this image arises from some objects in a scene looked at with a camera from this viewpoint.

• 58:59Speaker 01

  And we might have more information about the objects like their shape, maybe their textures, maybe, you know, color, et cetera, et cetera. So that's the reconstruction problem. In a way, you are in your head creating a model of the external world.

• 59:17Lex Fridman?

  Right.

• 59:18Speaker 01

  Okay, reorganization is to do with essentially finding these entities. So it's organization. The word organization implies a structure. So in perception, in psychology, we use the term perceptual organization. That the world is not just, an image is not just seen as, is not internally represented as just a collection of pixels, but we make these entities. We create these entities, objects, whatever you want to call them.

• 59:55Lex Fridman?

  And the relationship between the entities as well? Or is it purely about the entities?

• 59:59Speaker 01

  It could be about the relationships, but mainly we focus on the fact that there are entities.

• 1:00:04Lex Fridman?

  So I'm trying to pinpoint what the organization means.

• 1:00:09Speaker 01

  So organization is that instead of like a uniform grid, we have this structure of objects.

• 1:00:19Lex Fridman?

  So segmentation is a small part of that.

• 1:00:22Speaker 01

  So segmentation gets us going towards that.

• 1:00:25Lex Fridman?

  Yeah. And you kind of have this triangle where they all interact together.

• 1:00:30Speaker 01

  Yes.

• 1:00:31Lex Fridman?

  So how do you see that interaction? In sort of reorganization is, yes, defining the entities in the world. The recognition is labeling those entities. And then reconstruction is what, filling in the gaps?

• 1:00:51Speaker 01

  Well, to, for example, see, impute some 3D objects corresponding to each of these entities. That would be part of.

• 1:01:01Lex Fridman?

  Adding more information that's not there in the raw data.

• 1:01:05Speaker 01

  Correct. I mean, I started pushing this kind of a view in the, around 2010 or something like that. Because at that time, in computer vision, the distinction that people were just working on many different problems, but they treated each of them as a separate, isolated problem. With each with its own data set, and then you try to solve that and get good numbers on it. So I wasn't, I didn't like that approach because I wanted to see the connection between these. And if people divided up vision into, into various modules, the way they would do it is as low level, mid level, and high level vision.

• 1:01:51Speaker 01

  Corresponding roughly to the psychologist's notion of sensation, perception, and cognition. And I didn't, that didn't map to tasks that people cared about. Okay, so therefore I tried to promote this particular framework as a way of considering the problems that people in computer vision were actually working on. And trying to be more explicit about the fact that they actually are connected to each other. And I was at that time just doing this on the basis of information flow. Now, it turns out in the last five years or so, in the post the deep learning revolution,

• 1:02:34Speaker 01

  that this this architecture has turned out to be very conducive to that. Because basically, in these neural networks, we are trying to build multiple representations.

• 1:02:50Speaker 01

  There can be multiple output heads sharing common representations. So in a certain sense, today, given the reality of what solutions people have to this, I I I do not need to preach this anymore. It is it is just there, it's part of the solution space.

• 1:03:09Lex Fridman?

  So, speaking of neural networks, how much of this problem of computer vision, of reorganization, recognition, can be reconstruction?

• 1:03:26Lex Fridman?

  How much of it can be learned end to end, do you think? Sort of set it and forget it, just plug and play, have a giant data set, multiple perhaps, multimodal, and then just learn the entirety of it?

• 1:03:42Speaker 01

  Well, so I I think that currently what that end to end learning means nowadays is end to end supervised learning. And and that, I would argue, is too narrow a view of the problem. I would, I like this child development view, this lifelong learning view, one where there are certain capabilities that are built up, and then there are certain capabilities which are built up on top of that. So that's that's what I I believe in. So I think.

• 1:04:19Speaker 01

  End to end learning in the supervised setting for a very precise task to me is

• 1:04:30Speaker 01

  kind of is a sort of a limited view of the of the learning process.

• 1:04:35Lex Fridman?

  Got it. So if we think about beyond purely supervised, look back to children. You mentioned six lessons that we can learn from children of be multimodal, be incremental, be physical, explore, be social, use language. Can you speak to these, perhaps picking one that you find most fundamental to our

• 1:04:59Speaker 01

  Yeah.

• 1:04:59Lex Fridman?

  time today?

• 1:05:00Speaker 01

  Yeah, so I mean, I should say to give due credit, this is from a paper by Smith and Gasser. And it reflects essentially, I would say, common wisdom among child development people. It's just that these are, this is not common wisdom among people in computer vision and AI and machine learning. So I view my role as trying to

• 1:05:29Lex Fridman?

  Bridge the two worlds.

• 1:05:31Speaker 01

  Bridge the two worlds. So, so let's take an example of a multimodal. I like that. So multimodal, a canonical example is a child interacting with with an object. So then the child, so the child holds a ball and plays with it. So at that point, it's getting a touch signal. So the touch signal is

• 1:05:56Speaker 01

  is getting as the notion of 3D shape, but it is sparse. And then the child is also seeing a visual signal, right? And and these two, so imagine these are two in totally different spaces, right? So one is the space of receptors on the skin of the fingers and the thumb and the palm. Right? And then these map on to these neuronal fibers are getting activated somewhere. Right? These lead to some activation in somatosensory cortex. I mean, a similar thing will happen if we have a robot hand. Okay, and then we have the pixels corresponding to the visual view. But we know that they correspond to the same object.

• 1:06:41Speaker 01

  Right? So that's a very, very strong cross calibration signal. And it is self supervisory, which is beautiful, right? There's nobody assigning a label. The mother doesn't have to come and assign a label. The child doesn't even have to know that this object is called a ball. Okay, but the child is learning something about the three dimensional world. From this signal.

• 1:07:07Speaker 01

  I think tactile and visual, there is some work on. There is a lot of work currently on audio and visual. Okay, and audio visual, so there is some event that happens in the world. And that event has a visual signature. And it has a auditory signature. So there is this glass bowl on the table and it falls and breaks and I hear the smashing sound and I see the pieces of glass. Okay, I built that connection between the two. Right? We have people, I mean, this has become a hot topic in computer vision in the last couple of years. There are problems like separating out multiple speakers. Right? Which was a classic problem in in audition.

• 1:07:52Speaker 01

  They call this the problem of source separation or the cocktail party effect and so on. But just try to do it visually. When you also have, it becomes so much easier and so much more useful.

• 1:08:07Lex Fridman?

  So the multimodal. I mean, there's so much more signal with multimodal and you can use that for some kind of weak supervision as well.

• 1:08:17Speaker 01

  Yes, because they are occurring at the same time in time. So you have time which links the two, right? So at a certain moment, T1, you've got a certain signal in the auditory domain and a certain signal in the visual domain. But they must be causally related.

• 1:08:32Lex Fridman?

  Yeah, it's an exciting area, not well studied yet.

• 1:08:35Speaker 01

  Yeah, I mean, we have a little bit of work at this, but but but so much more needs to be done. So so so so this, this is, this is a good example. Be physical, that's to do with like some one thing we talked about earlier, that that there's an embodied world.

• 1:08:53Lex Fridman?

  To mention language, use language. So Noam Chomsky believes that language may be at the core of cognition, at the core of everything in the human mind. What is the connection between language and vision to you? Like, what's more fundamental? Are they neighbors? Is one the parent and the child? The chicken and the egg?

• 1:09:15Speaker 01

  Oh, it's very clear. It is vision which is the parent.

• 1:09:17Lex Fridman?

  The fundamental.

• 1:09:18Speaker 01

  Which is the fundamental ability. Okay.

• 1:09:21Lex Fridman?

  Wait, wait, wait.

• 1:09:22Speaker 01

  So.

• 1:09:24Lex Fridman?

  It comes before. You think vision is more fundamental than language?

• 1:09:28Speaker 01

  Correct. And you can think of it either in phylogeny or in ontogeny. So phylogeny means if you look at evolutionary time, right? So we have vision that developed 500 million years ago. Okay, then something like when we get to maybe like 5 million years ago, you have the first bipedal primate. So when we started to walk, then the hands became free. And so then manipulation, the ability to manipulate objects and build tools and so on and so forth.

• 1:10:02Lex Fridman?

  So you said 500, 1,000 years ago?

• 1:10:04Speaker 01

  No, no, sorry. The first multicellular animals which you can say had some intelligence arose 500 million years.

• 1:10:14Lex Fridman?

  Million.

• 1:10:15Speaker 01

  Okay, and now let's fast forward to say the last 7 million years, which is the development of the hominid line, right? Where from the other primates, we have the branch which leads on to modern humans. Now, there are many of these hominids, but the ones which, you know, people talk about Lucy because that's like a skeleton from 3 million years ago and we know that Lucy walked. Okay, so at this stage you have that the hand is free for manipulating objects. And then the ability to manipulate objects, build tools, and the brain size grew in this era.

• 1:11:00Speaker 01

  So, okay, so now you have manipulation. Now, we don't know exactly when language arose.

• 1:11:06Lex Fridman?

  But after that.

• 1:11:07Speaker 01

  But after that. Because no apes have, I mean, so, I mean Chomsky is correct in that, that it is a uniquely human capability.

• 1:11:16Lex Fridman?

  Right.

• 1:11:17Speaker 01

  And we, primates, other primates don't have that. But, so it developed somewhere in this era, but it developed, I would, I mean, argue that it probably developed after we had this stage of humans, I mean, the human species is already able to manipulate and hands-free, much bigger brain size.

• 1:11:42Lex Fridman?

  And for that, there's a lot of vision has already had, had to have developed.

• 1:11:48Speaker 01

  Yeah.

• 1:11:49Lex Fridman?

  So the sensation and the perception may be some of the cognition.

• 1:11:53Speaker 01

  Yeah, so we, we, so those, so that, so the world, so there, so, so these ancestors of ours, you know, three, four million years ago, they had, they had spatial intelligence. So they knew that the world consists of objects. They knew that the objects were in certain relationships to each other. They had observed causal interactions among objects. They could move in space, so they had space and time and all of that. So, language builds on that substrate. So, language has a lot of, I mean, I mean, the, all human languages have constructs

• 1:12:37Speaker 01

  which depend on a notion of space and time. Where did that notion of space and time come from? It had to come from perception and action in the world we live in.

• 1:12:47Lex Fridman?

  Yeah, well, you've referred to the spatial intelligence.

• 1:12:50Speaker 01

  Yeah.

• 1:12:51Lex Fridman?

  Yeah. So, to linger a little bit, we mentioned Turing and his mention of we should learn from children. Nevertheless, language is the fundamental piece of the test of intelligence that Turing proposed.

• 1:13:07Speaker 01

  Yes.

• 1:13:08Lex Fridman?

  What do you think is a good test of intelligence? Are you, what would impress the heck out of you? Is it fundamentally natural language? Or is there something in vision?

• 1:13:19Speaker 01

  I, I think, I, I wouldn't, I, I don't think we should have, create a single test of intelligence. So just like I don't believe in IQ as a single number, I think generally there can be many capabilities which are correlated perhaps.

• 1:13:39Speaker 01

  So I think that there will be accomplishments, which are visual accomplishments, accomplishments which are accomplishments in manipulation or robotics, and then accomplishments in language. I do believe that language will be the hardest nut to crack.

• 1:13:57Lex Fridman?

  Really?

• 1:13:58Speaker 01

  Yeah.

• 1:13:58Lex Fridman?

  So what's harder, to pass the spirit of the Turing test, or like whatever formulation will make it natural language, convincingly in natural language, like somebody you would want to have a beer with, hang out and have a chat with, or the general natural scene understanding? You think language is the tougher problem?

• 1:14:18Speaker 01

  I think I'm not a fan of the... I think Turing test, that Turing, as he proposed the test in 1950, was trying to solve a certain problem.

• 1:14:31Lex Fridman?

  Yeah, imitation.

• 1:14:32Speaker 01

  Yeah. And I think it made a lot of sense then. Where we are today, 70 years later, I think we should not worry about that. I mean, I think the Turing test is no longer the right way to channel research in AI, because that, it takes us down this path of this chatbot which can fool us for five minutes or whatever. Okay? I think I would rather have a list of 10 different tasks. I mean, I think there are tasks which, there are tasks in the manipulation domain, tasks in navigation, tasks in visual scene understanding, tasks in reading a story and answering questions based on that.

• 1:15:16Speaker 01

  I mean, so my favorite language understanding task would be, you know, reading a novel and being able to answer arbitrary questions from it. Okay.

• 1:15:27Lex Fridman?

  Right.

• 1:15:27Speaker 01

  I think that, to me, and this is not an exhaustive list by any means, so I would, I think that that's where we need to be going to, and each of these, on each of these axes, there's a fair amount of work to be done.

• 1:15:43Lex Fridman?

  So on the visual understanding side, in this Intelligence Olympics that we've set up?

• 1:15:48Speaker 01

  Yeah.

• 1:15:48Lex Fridman?

  What's a good test for one of many of visual scene understanding?

• 1:15:56Lex Fridman?

  Do you think such benchmarks exist? Sorry to interrupt.

• 1:15:59Speaker 01

  No, there aren't any. I think essentially, to me, a really good aid to the blind. So suppose there was a blind person and I needed to assist the blind person.

• 1:16:14Lex Fridman?

  So ultimately, like we said, vision that aids in the action in the survival in this world.

• 1:16:20Speaker 01

  Yeah.

• 1:16:22Lex Fridman?

  Maybe in the simulated world.

• 1:16:26Speaker 01

  Maybe easier to measure performance in a simulated world. What we are ultimately after is performance in the real world.

• 1:16:34Lex Fridman?

  So David Hilbert in 1900 proposed 23 open problems in mathematics, some of which are still unsolved, most important, famous of which is probably the Riemann hypothesis. You've thought about and presented about the Hilbert problems of computer vision. So let me ask, what do you today? I don't know when the last year you've presented that, 2015, but versions of it. You're kind of the face and the spokesperson for computer visions. It's your job to state what the problem, the open problems are for the field. So what today are the Hilbert problems of computer vision, do you think?

• 1:17:13Speaker 01

  Let me pick one which I regard as clearly unsolved, which is what I would call long-form video understanding. So we have a video clip and we want to understand the behavior in there in terms of agents, their goals, intentionality, and make predictions about what might happen.

• 1:17:46Speaker 01

  You know, so that that kind of understanding which goes away from atomic visual action. So so in the short range, the question is, are you sitting, are you standing, are you catching a ball? Right, that we can do now. Or even if we can't do it fully accurately, if we can do it at fifty percent, maybe next year we'll do it at sixty five and so forth. But I think the long range video understanding, I don't think we we can do today.

• 1:18:18Lex Fridman?

  And that means so long.

• 1:18:20Speaker 01

  And it blends into cognition. That's the reason why it's challenging.

• 1:18:24Lex Fridman?

  And so you have to track, you have to understand entities, you have to understand the entities, you have to track them. And you have to have some kind of model of their behavior.

• 1:18:34Speaker 01

  Correct. And their and their behavior might be, these are these are agents, so they are not just like passive objects, but they're agents, so therefore they might, they would exhibit goal directed behavior. Okay, so this is this is one area. Then I will talk about, say, understanding the world in three D. Now, this may seem paradoxical because in a way we have been able to do three D understanding even like thirty years ago, right? But I don't think we currently have the richness of three D understanding in our computer vision system that we would like. Because so let me elaborate on that a bit. So currently, we have

• 1:19:17Speaker 01

  two kinds of techniques which are not fully unified. So there are the kinds of techniques from multi view geometry, that you have multiple pictures of a scene and you do a reconstruction using stereoscopic vision or structure from motion. But these techniques do not, they totally fail if you just have a single view. Because they are relying on this this multiple view geometry. Okay, then we have some techniques that we have developed in the computer vision community, which try to guess three D from single views. And these techniques are based on on a supervised learning. And they are based on having at training time, three D models of objects available.

• 1:20:03Speaker 01

  And this is completely unnatural supervision. Right, that's not, CAD models are not injected into your brain. Okay, so what would I like? What I would like would be a kind of learning as you move around the world notion of 3D. So, so we, we have our succession of visual experiences. And from those, we, so in, as part of that, I might see a chair from different viewpoints, or a table from viewpoint, different viewpoints and so on. Now, as part, that enables me to build some internal representation.

• 1:20:48Speaker 01

  And then, next time, I just see a single photograph, and it may not even be of that chair, it's of some other chair, and I have a guess of what its 3D shape is like.

• 1:20:59Lex Fridman?

  So you're almost learning the CAD model, kind of?

• 1:21:02Speaker 01

  Yeah, implicitly. I mean, I mean, the CAD model need not be in the same form as used by computer graphics programs.

• 1:21:09Lex Fridman?

  Hidden in the representation.

• 1:21:11Speaker 01

  It's hidden in the representation, the ability to predict new views, and what I would see if I went to such and such position.

• 1:21:21Lex Fridman?

  By the way, on a, on a small tangent on that, are you on comfort, are you okay? Or comfortable with neural networks that do achieve visual understanding, that do, for example, achieve this kind of 3D understanding? And you don't know how they, you don't know the, you're not able to interest, but you're not able to visualize, or understand, or interact with the representation? So the fact that they're not, or may not be explainable?

• 1:21:53Speaker 01

  Yeah, I think that's fine. To me, that is, so, so let me put some caveats on that. So it depends on the setting. So first of all, I think, the, humans are not explainable.

• 1:22:12Lex Fridman?

  Yeah, that's a really good point, yeah.

• 1:22:14Speaker 01

  So we, we, one human to another human is not fully explainable. I think there are settings where explainability matters. And these might, these are, these might be, for example, questions on medical diagnosis.

• 1:22:30Speaker 01

  So I'm in a setting where maybe the doctor, maybe a computer program has made a certain diagnosis. And then depending on the diagnosis, perhaps I should have treatment A or treatment B. Right? So now is the computer program's diagnosis based on data, which was data collected of for American males who are in their thirties and forties? And maybe not so relevant to me? Maybe it is relevant, you know, et cetera, et cetera. And I mean, in medical diagnosis, we have major issues to do with the reference class.

• 1:23:10Speaker 01

  So we may have acquired statistics from one group of people and applying it to a different group of people who may not share all the same characteristics. The data might have, there might be error bars in the prediction. So that prediction should really be taken with a huge grain of salt. And but this has an impact on what treatments should be picked, right? So so there are settings where I want to know more than just this is the answer. But what I acknowledge is that the so so so so I, in that sense, explainability and interpretability may matter.

• 1:23:52Speaker 01

  It's about giving error bounds and a better sense of the quality of the decision. Where what I where I'm willing to sacrifice interpretability is that I believe that there can be systems which can be highly performant but which are internally black boxes.

• 1:24:12Lex Fridman?

  And that seems to be where it's headed. Some of the best performing systems are essentially black boxes.

• 1:24:18Speaker 01

  Yeah.

• 1:24:19Lex Fridman?

  Fundamentally by their construction.

• 1:24:21Speaker 01

  You and I are are black boxes to each other.

• 1:24:23Lex Fridman?

  Yeah, so the nice thing about the black boxes, we are. So we ourselves are black boxes, but we're also, those of us who are charming, are able to convince others, like explain the black, what's going on inside the black box with narratives, with stories. So in some sense, neural networks don't have to actually explain what's going on inside. They just have to come up with stories, real or fake, that convince you that they know what's going on.

• 1:24:55Speaker 01

  And I'm sure we can do that. We can create those stories. Neural networks can create those stories.

• 1:25:01Lex Fridman?

  Yeah.

• 1:25:04Lex Fridman?

  And the transformer will be involved. Do you think we will ever build a system of human level or superhuman level intelligence? We've kind of defined what it takes to try to approach that, but do you think that's within our reach? The thing that we thought we could do, what Turing thought actually, we could do by year 2000, right? Do you think we'll ever be able to do?

• 1:25:28Speaker 01

  So I think there are two answers here. One answer is, in principle, can we do this at some time? And my answer is yes. The second answer is a pragmatic one. Do you think we will be able to do it in the next 20 years or whatever? And to that, my answer is no. So, of course, that's a wild guess. I think that, you know, Donald Rumsfeld is not a favorite person of mine, but one of his lines is very good, which is about known knowns, known unknowns, and unknown unknowns. So in the business we are in, there

• 1:26:08Speaker 01

  are known unknowns and we have unknown unknowns. So I think with respect to a lot of what's the case in vision and robotics, I feel like we have known unknowns. So I have a sense of where we need to go and what the problems that need to be solved are. I feel with respect to natural language, understanding and high level cognition, it's not just known unknowns, but also unknown unknowns. So it is very difficult to put any kind of a timeframe to that.

• 1:26:48Lex Fridman?

  Do you think some of the unknown unknowns might be positive in that they'll surprise us and make the job much easier? So fundamental breakthroughs?

• 1:26:57Speaker 01

  I think that is possible because certainly I have been very positively surprised by how effective these deep learning systems have been because I certainly would not have believed that in 2010. I think what we knew from the mathematical theory was that convex optimization works when there's a single global optima, then this gradient descent techniques would work. Now, these are nonlinear systems with non-convex systems.

• 1:27:33Lex Fridman?

  Huge number of variables, so over parameterized.

• 1:27:35Speaker 01

  Over parameterized. And the people who used to play with them a lot, the ones who were totally immersed in the lore and the black magic, they knew that they worked well, even though they were.

• 1:27:51Lex Fridman?

  Really? I thought like everybody would.

• 1:27:53Speaker 01

  No, the claim that I hear from my friends like Yan Le Cun and so forth.

• 1:27:58Lex Fridman?

  Oh, no, yeah.

• 1:27:59Speaker 01

  That they feel that they were comfortable with it.

• 1:28:03Lex Fridman?

  Well, he says that.

• 1:28:03Speaker 01

  But the community as a whole was certainly not. And I think we were, to me, that was the surprise, that they actually worked robustly for a wide range of problems, from a wide range of initializations and so on. And so that was, that was certainly more rapid progress than we expected. But then there are certainly lots of times, in fact, most of the history of AI is when we have made less progress at a slower rate than we expected. So we just

• 1:28:43Speaker 01

  keep going. I think what I regard as really unwarranted are these, these fears of, you know, AGI in 10 years and 20 years and that kind of stuff, because that's based on completely unrealistic models of how rapidly we will make progress in this field.

• 1:29:06Lex Fridman?

  So I agree with you, but I've also gotten a chance to interact with very smart people who really worry about existential threats of AI. And I, as an open-minded person, am sort of taking it in. Do you think if AI systems, in some way, the unknown unknowns, not superintelligent AI, but in ways we don't quite understand the nature of superintelligence, will have a detrimental effect on society? Do you think this is something we should be worried about? Or we need to first allow the unknown unknowns to become known unknowns?

• 1:29:46Speaker 01

  I think we need to be worried about AI today. I think that it is not just a worry we need to have when we get that AGI. I think that AI is being used in many systems today. And there might be settings, for example, when it causes biases or decisions which could be harmful, I mean, decisions which could be unfair to some people, or it could be a self-driving cars which kills a pedestrian. So AI systems are being deployed today, right? And they are being deployed in many different settings, maybe in medical diagnosis, maybe in a self-driving car, maybe in selecting applicants for an interview.

• 1:30:27Speaker 01

  So I would argue that when these systems make mistakes, there are consequences. And we are, in a certain sense, responsible for those consequences. So I would argue that this is a continuous effort. It is, we, and this is something that, in a way, is not so surprising. It's about all engineering and scientific progress, which great power comes great responsibility. So as these systems are deployed, we have to worry about them, and it's a continuous problem. I don't think of it as something which will suddenly happen on some day in 2079, for which I need to design some clever trick.

• 1:31:12Speaker 01

  I'm saying that these problems exist today.

• 1:31:14Lex Fridman?

  Yeah.

• 1:31:15Speaker 01

  And we need to be continuously on the lookout for. Worrying about safety, biases, risks, right? I mean, a self-driving car kills a pedestrian, and they have, right? I mean, there's Uber incident in Arizona.

• 1:31:32Lex Fridman?

  Yeah.

• 1:31:33Speaker 01

  Right? It has happened, right? This is not about AGI. In fact, it's about a very dumb intelligence, which is still killing people.

• 1:31:41Lex Fridman?

  The worry people have with AGI is the scale. And I, but I think you're 100% right is like the thing that worries me about AI today. And it's happening in a huge scale is recommender system, recommendation systems. So if you look at Twitter, Facebook or YouTube, they're controlling the ideas that we have access to, the news and so on. And that's a fundamentally machine learning algorithm behind each of these recommendations. And they, I mean, my life would not be the same without these sources of information. I'm a totally new human being and the ideas that I know are very much because of the internet, because of the algorithm that recommend those ideas.

• 1:32:26Lex Fridman?

  And so as they get smarter and smarter, I mean, that is the AGI.

• 1:32:31Speaker 01

  Yeah.

• 1:32:32Lex Fridman?

  Is that's the, the algorithm that's recommending the next YouTube video you should watch has control of millions of billions of people. That, that algorithm is already super intelligent and has complete control of the population. Not a complete, but very strong control. For now, we can turn off YouTube. We can just go have a normal life outside of that. But the more and more that gets into our life, it's that algorithm, we start depending on it in the different companies that are working on the algorithm. So I think it's, you're right, it's already, it's already there. And YouTube in particular is using computer vision, doing their hardest to try to

• 1:33:16Lex Fridman?

  understand the content of videos so they could be able to connect videos with the people who would benefit from those videos the most. And so that development could go in a bunch of different directions, some of which might be harmful.

• 1:33:32Lex Fridman?

  So, yeah, you're right, the, the threats of AI here already, we should be thinking about them. On a philosophical notion.

• 1:33:42Lex Fridman?

  If you could, personal perhaps, if you could relive a moment in your life outside of family, because it made you truly happy, or it was a profound moment that impacted the direction of your life, what moment would you go to?

• 1:34:01Speaker 01

  I don't think of single moments, but I look over the long haul. I feel that I've been very lucky, because I feel that, I think that in scientific research, a lot of it is about being at the right place at the right time. And you can, you can work on problems at a time when they're just too premature. You know, you bat your head against them and, and nothing happens because it's, the prerequisites for success are not there. And then there are times when you are in a field which is all pretty mature and you

• 1:34:42Speaker 01

  can only solve curlicues upon curlicues. I've been lucky to have been in this field, which for thirty four years, thirty five, well, actually thirty four years as a professor at Berkeley, so longer than that, which, when I started in it, was just like some little crazy, absolutely useless field. Couldn't really do anything to a time when it's really, really solving a lot of practical problems, has a lot, has offered a lot of tools for scientific research, right? Because computer vision is impactful for images in

• 1:35:25Speaker 01

  biology or astronomy and so on and so forth. And we have, so we have made great scientific progress, which has had real practical impact in the world. And I feel lucky that I, I got in at a time when the field was Very young and at a time when it is it's now mature, but not fully mature. It's mature, but not done. I mean, it's really still in a in a productive phase.

• 1:35:56Lex Fridman?

  Yeah, I think.

• 1:35:56Speaker 01

  This is a, yeah.

• 1:35:57Lex Fridman?

  Yeah, I think people 500 years from now would laugh at you calling this field mature.

• 1:36:01Speaker 01

  Yeah, that is very possible. Yeah.

• 1:36:04Lex Fridman?

  So, but you're also, lest I forget to mention, you've also mentored some of the biggest names of computer vision, computer science and AI today. There's so many questions I could ask, but it really is what what is it? How did you do it? What does it take to be a good mentor? What does it take to be a good guide?

• 1:36:27Speaker 01

  Yeah, I I think what I feel I've been lucky to have had very, very smart and hardworking and creative students. I think some part of the credit just belongs to being at Berkeley. Those of us who are at top universities are blessed because we have very, very smart and capable students coming on knocking on our door. So, so I have to be humble enough to acknowledge that. But what have I added? I think I have added something. What I have added is I think what I've always tried to teach them is a sense of picking the right problems.

• 1:37:11Speaker 01

  So I think that in science, in the short run, success is always based on technical competence. You're, you know, you're quick with math or you're whatever. I mean, there's certain technical capabilities which make for short range progress. Long range progress is really determined by asking the right questions and focusing on the right problems. And I feel that.

• 1:37:43Speaker 01

  What I've been able to bring to the table in terms of advising these students is some sense of taste of what are good problems. What are problems that are worth attacking now as opposed to waiting 10 years?

• 1:37:58Lex Fridman?

  What's a good problem? If you could summarize, is that possible to even summarize? Like, what's your sense of a good problem?

• 1:38:05Speaker 01

  I think I have a sense of what is a good problem, which is there is a British scientist, in fact, he won a Nobel Prize, Peter Medovar, who has a book on this. And basically, he calls it, "Research is the art of the soluble." So we need to sort of find problems which are

• 1:38:31Speaker 01

  which are not yet solved, but which are approachable. And he sort of refers to this sense that there is this problem which isn't quite solved yet, but it has a soft underbelly. There is some place where you can, you know, spear the beast.

• 1:38:49Lex Fridman?

  Yes.

• 1:38:49Speaker 01

  And having that intuition that this problem is ripe is a good thing, because otherwise you can just beat your head and not make progress. So I think that is important. So if I have that and if I can convey that to students, it's not just that they do great research while they're working with me, but that they continue to do great research. So in a sense, I'm proud of my students and their achievements and their great research, even 20 years after they have ceased being my student.

• 1:39:22Lex Fridman?

  So it's in part developing, helping them develop that sense that a problem is not yet solved, but is solvable.

• 1:39:29Speaker 01

  Correct. The other thing which I have, which I think I bring to the table

• 1:39:36Speaker 01

  is a certain intellectual breadth.

• 1:39:40Speaker 01

  I've spent a fair amount of time studying psychology, neuroscience, relevant areas of applied math and so forth. So I can probably help them see some connections to disparate things which they might not have otherwise. So the smart students coming into Berkeley can be very deep in the sense, they can think very deeply, meaning very hard down one particular path. But where I could help them is the shallow breadth, but whereas they would have the narrow depth. And but that's of some value.

• 1:40:26Lex Fridman?

  Well, it was beautifully refreshing just to hear you naturally jump to psychology, back to computer science in this conversation, back and forth. I mean, that's a rare quality. And I think it's certainly for students empowering to think about problems in a new way. So for that, and for many other reasons, I really enjoyed this conversation. Thank you so much. It was a huge honor. Thanks for talking today.

• 1:40:49Speaker 01

  It's been my pleasure.

• 1:40:51Lex Fridman?

  Thanks for listening to this conversation with Jitendra Malik. And thank you to our sponsors, BetterHelp and ExpressVPN. Please consider supporting this podcast by going to betterhelp.com/lex and signing up at expressvpn.com/lexpod. Click the links, buy the stuff. It's how they know I sent you, and it really is the best way to support this podcast and the journey I'm on. If you enjoy this thing, subscribe on YouTube, review it with five stars on Apple Podcasts, support it on Patreon, or connect with me on Twitter at Lex Fridman. Don't ask me how to spell that. I don't remember myself. And now let me leave you

• 1:41:34Lex Fridman?

  with some words from Prince Mishkin in The Idiot by Dostoevsky. "Beauty will save the world." Thank you for listening and hope to see you next time.