Of imitation learning for a self-driving car, but most of those are incredibly boring.

What I like is select 10% of them that are kind of the most informative.

And with just that, I would probably reach the same.

So it's a weak form of active learning, if you want.

Yes, but there might be a much stronger version.

That's what, and that's an open question if it exists.

The question is how much stronger it can get.

Elon Musk is confident, I talked to him recently.

He's confident that large scale data and deep learning can solve the autonomous driving problem.

What are your thoughts on the limitless possibilities of deep learning in this space?

Well, it's obviously part of the solution.

I mean, I don't think we'll ever have a self-driving system, or at least not in the foreseeable future, that does not use deep learning.

So in the history of sort of engineering, particularly sort of AI-like systems, There's generally a first phase where everything is built by hand.

Then there is a second phase, and that was the case for autonomous driving, you know, 20, 30 years ago.

There's a phase where there's a little bit of learning is used, but there's a lot of engineering that's involved in kind of, you know, taking care of corner cases and putting limits, etc.

Because the learning system is not perfect.

And then as technology progresses, we end up relying more and more on learning.

That's the history of character recognition, it's the history of speech recognition, now computer vision, natural language processing.

And I think the same is going to happen with autonomous driving, that currently the methods that are closest to providing some level of autonomy, some decent level of autonomy, where you don't expect a driver to kind of do anything, is where you constrain the world.

So you only run within 100 square kilometers or square miles in Phoenix, where the weather is nice and the roads are wide, which is what Waymo is doing.

You completely over-engineer the car with tons of lidars and sophisticated sensors that are too expensive for consumer cars, but they're fine if you just run a fleet.

And you engineer the hell out of everything else.

You map the entire world, so you have a complete 3D model of everything.

So the only thing that the perception system has to take care of is moving objects and construction and things that weren't in your map.

And you can engineer a good SLAM system and all that stuff.

So that's the current approach that's closest to some level of autonomy.

But I think eventually, the long-term solution is going to rely more and more on learning and possibly using a combination of self-supervised learning and model-based reinforcement or something like that.

But ultimately learning will be not just at the core, but really the fundamental part of the system.

Yeah, it already is, but it'll become more and more.

What do you think it takes to build a system with human level intelligence?

You talked about the AI system in the movie, Her being way out of reach, our current reach.

This might be outdated as well, but- It's still way out of reach.

What would it take to build her, do you think?

So I can tell you the first 2 obstacles that we have to clear, but I don't know how many obstacles there are after this.

So the image I usually use is that there is a bunch of mountains that we have to climb, and we can see the first 1, but we don't know if there are 50 mountains behind it or not.

And this might be a good sort of metaphor for why AI researchers in the past have been overly optimistic about the result of AI.

You know, for example, Newell and Simon, right, wrote the general problem solver and they call it a general problem solver.

And of course, the first thing you realize is that all the problems you want to solve are exponential, and so you can't actually use it for anything useful.

Yeah, so yeah, all you see is the first peak.

So what are the first couple of peaks for her?

So the first peak, which is precisely what I'm working on, is self-supervised learning.

How do we get machines to learn models of the world by observation, kind of like babies and like young animals.

So we've been working with, you know, cognitive scientists.

So this Emmanuel Dupou, who is at FAIR in Paris, is a half-time, is also a researcher in French university.

And he has this chart that shows how many months of life baby humans can learn different concepts.

And you can measure this in various ways.

So things like distinguishing animate objects from inanimate objects, you can tell the difference at age 2, 3 months.

Whether an object is going to stay stable, is going to fall, you know, about 4 months, you can tell.

You know, there are various things like this.

And then things like gravity, the fact that objects are not supposed to float in the air, but are supposed to fall, you learn this around the age of 8 or 9 months.

If you look at a lot of, you know, eight-month-old babies, you give them a bunch of toys on their high chair, first thing they do is they throw them on the ground and they look at them.

It's because they're learning about, actively learning about gravity.

So, they're not trying to annoy you, but they need to do the experiment, right?

So, how do we get machines to learn like babies, mostly by observation with a little bit of interaction and learning those models of the world?

Because I think that's really a crucial piece of an intelligent autonomous system.

So if you think about the architecture of an intelligent autonomous system, it needs to have a predictive model of the world.

So something that says, here is a world at time t, here is a state of the world at time t plus 1 if I take this action.

And it's not a single answer, it can be a-

Yeah, well, but we don't know how to represent distributions in high dimensional spaces so it's going to be something weaker than that okay yeah but with some some representation of uncertainty if you have that then you can do what optimal control theorists call model predictive control which means that you can run your model with a hypothesis for a sequence of action and then see the result.

Now, what you need, the other thing you need, is some sort of objective that you want to optimize.

Am I reaching the goal of grabbing this object?

So there is some sort of objective that you have to minimize.

And so in your head, if you have this model, you can figure out the sequence of action that will optimize your objective.

That objective is something that ultimately is rooted in your basal ganglia, at least in the human brain, that's what it is.

Basal ganglia computes your level of contentment or miscontentment, I don't know if that's a word, unhappiness.

And so your entire behavior is driven towards kind of minimizing that objective, which is maximizing your contentment, computed by your basal ganglia.

And what you have is an objective function, which is basically a predictor of what your basal ganglia is going to tell you.

So you're not going to put your hand on fire because you know it's going to burn and you're going to get hurt.

And you're predicting this because of your model of the world and your sort of predictor of this objective, right?

So if you have those 3 components, you have 4 components.

You have the hardwired contentment objective computer, if you want, calculator.

1 is the objective predictor, which basically predicts your level of contentment.

And there's a third module I didn't mention, which is a module that will figure out the best course of action to optimize an objective given your model.

Call this a policy network or something like that, right?

Now, you need those 3 components to act autonomously intelligently.

And you can be stupid in 3 different ways.

You can be stupid because your model of the world is wrong.

You can be stupid because your objective is not aligned with what you actually want to achieve.

And then the third way you can be stupid is that you have the right model, you have the right objective, but you're unable to figure out a course of action to optimize your objective given your model.

Some people who are in charge of big countries actually have all 3 that are wrong.

Okay, so if we think about this agent, if we think about the movie Her, you've criticized the art project that is Sophia the Robot.

And what that project essentially does is uses our natural inclination to anthropomorphize things that look like human and give them more.

Do you think that could be used by AI systems, like in the movie Her?

So do you think that body is needed to create a feeling of intelligence?

Well, if Sophia was just an art piece, I would have no problem with it, but it's presented as something else.

If creators of Sophia could change something about their marketing or behavior in general, what would it be?

I mean, don't you think, here's a tough question.

So Sophia is not, the general public feels that Sophia can do way more than she actually can.

And the people who created Sophia are not honestly, publicly communicating, trying to teach the public.

Don't you think the same thing is scientists in industry and research are taking advantage of the same misunderstanding in the public when they create AI companies or publish stuff.

I mean, there is no sense of, there's no desire to delude.

There's no desire to kind of over-claim with something that's done.

Right, you publish a paper on AI that, you know, has this result on ImageNet, you know, it's pretty clear.

I mean, it's not even interesting anymore.

But, you know, I don't think there is that.

I mean, the reviewers are generally not very forgiving of unsupported claims of this type.

But there are certainly quite a few startups that have had a huge amount of hype around this that I find extremely damaging and I've been calling it out when I've seen it.

So yeah, but to go back to your original question, like the necessity of embodiment.

I think, I don't think embodiment is necessary.

So I don't think we're going to get machines that really understand language without some level of grounding in the real world.

And it's not clear to me that language is a high enough bandwidth medium to communicate how the real world works.

So grounding means that, so there is this classic problem of common sense reasoning, you know, the Winograd Schema, right?

And so I tell you the trophy doesn't fit in the suitcase because it's too big, or the trophy doesn't fit in the suitcase because it's too small.

And the it in the first case refers to the trophy in the second case to the suitcase.

And the reason you can figure this out is because you know where the trophy and the suitcase are, you know when it's supposed to fit in the other 1, and you know the notion of size and the big object doesn't fit in a small object, unless it's a TARDIS, you know, things like that, right?

So you have this knowledge of how the world works, of geometry and things like that.

I don't believe you can learn everything about the world by just being told in language how the world works, I think you need some low-level perception of the world, you know, be it visual touch, you know, whatever, but some higher bandwidth perception of the world.

So by reading all the world's text, you still may not have enough information.

There's a lot of things that just will never appear in text and that you can't really infer.

So I think common sense will emerge from, certainly a lot of language interaction, but also with watching videos or perhaps even interacting in virtual environments and possibly, you know, robot interacting in the real world.

But I don't actually believe necessarily that this last 1 is absolutely necessary.

But I think there's a need for some grounding.

But the final product doesn't necessarily need to be embodied.

It just needs to have an awareness, a grounding.

But it needs to know how the world works to have, you know, to not be frustrated, frustrating to talk to.

And you talked about emotions being important.

Well, so, you know, I talked about this, the basal ganglia as the, you know, the thing that calculates your level of mixed contentment, And then there is this other module that sort of tries to do a prediction of whether you're gonna be content or not.

So fear, for example, is an anticipation of bad things that can happen to you, right?

You have this inkling that there is some chance that something really bad is gonna happen to you and that creates fear.

When you know for sure that something bad is gonna happen to you, you kind of give up, right?

So the punchline is, we're not gonna have autonomous intelligence without emotions.

So you mentioned very practical things of fear, but there's a lot of other mess around it.

But they are kind of the results of drives.

Yeah, there's deeper biological stuff going on.

There's a fascinating stuff that ultimately connects to our brain.

You know, I think the first 1 we'll create will probably not be that smart.

So you would have to ask her a question to know she's not that smart.

Well, what's a good question to ask, you know, to be

And if she answers, oh, it's because the leaves of the tree are moving and that creates wind, she's onto something.

And if she says, that's a stupid question, she's really onto something.

No, and then you tell her, actually, here is the real thing, and she says, oh, yeah, that makes sense.

So questions that reveal the ability to do common sense reasoning about the physical world.

Yeah, and you know, some level of causal inference.


Speaker 0: Of imitation learning for a self-driving car, but most of those are incredibly boring. What I like is select 10% of them that are kind of the most informative. And with just that, I would probably reach the same. So it's a weak form of active learning, if you want.

Speaker 1: Yes, but there might be a much stronger version.

Speaker 0: Yeah, that's right.

Speaker 1: That's what, and that's an open question if it exists. The question is how much stronger it can get. Elon Musk is confident, I talked to him recently. He's confident that large scale data and deep learning can solve the autonomous driving problem. What are your thoughts on the limitless possibilities of deep learning in this space?

Speaker 0: Well, it's obviously part of the solution. I mean, I don't think we'll ever have a self-driving system, or at least not in the foreseeable future, that does not use deep learning. Let me put it this way. Now, how much of it? So in the history of sort of engineering, particularly sort of AI-like systems, There's generally a first phase where everything is built by hand.

Then there is a second phase, and that was the case for autonomous driving, you know, 20, 30 years ago. There's a phase where there's a little bit of learning is used, but there's a lot of engineering that's involved in kind of, you know, taking care of corner cases and putting limits, etc. Because the learning system is not perfect. And then as technology progresses, we end up relying more and more on learning. That's the history of character recognition, it's the history of speech recognition, now computer vision, natural language processing.

And I think the same is going to happen with autonomous driving, that currently the methods that are closest to providing some level of autonomy, some decent level of autonomy, where you don't expect a driver to kind of do anything, is where you constrain the world. So you only run within 100 square kilometers or square miles in Phoenix, where the weather is nice and the roads are wide, which is what Waymo is doing. You completely over-engineer the car with tons of lidars and sophisticated sensors that are too expensive for consumer cars, but they're fine if you just run a fleet. And you engineer the hell out of everything else. You map the entire world, so you have a complete 3D model of everything.

So the only thing that the perception system has to take care of is moving objects and construction and things that weren't in your map. And you can engineer a good SLAM system and all that stuff. So that's the current approach that's closest to some level of autonomy. But I think eventually, the long-term solution is going to rely more and more on learning and possibly using a combination of self-supervised learning and model-based reinforcement or something like that.

Speaker 1: But ultimately learning will be not just at the core, but really the fundamental part of the system.

Speaker 0: Yeah, it already is, but it'll become more and more.

Speaker 1: What do you think it takes to build a system with human level intelligence? You talked about the AI system in the movie, Her being way out of reach, our current reach. This might be outdated as well, but- It's still way out of reach. It's still way out of reach. What would it take to build her, do you think?

Speaker 0: So I can tell you the first 2 obstacles that we have to clear, but I don't know how many obstacles there are after this. So the image I usually use is that there is a bunch of mountains that we have to climb, and we can see the first 1, but we don't know if there are 50 mountains behind it or not. And this might be a good sort of metaphor for why AI researchers in the past have been overly optimistic about the result of AI. You know, for example, Newell and Simon, right, wrote the general problem solver and they call it a general problem solver.

Speaker 1: General problem solver. Okay.

Speaker 0: And of course, the first thing you realize is that all the problems you want to solve are exponential, and so you can't actually use it for anything useful. But you know.

Speaker 1: Yeah, so yeah, all you see is the first peak. So what are the first couple of peaks for her?

Speaker 0: So the first peak, which is precisely what I'm working on, is self-supervised learning. How do we get machines to learn models of the world by observation, kind of like babies and like young animals. So we've been working with, you know, cognitive scientists. So this Emmanuel Dupou, who is at FAIR in Paris, is a half-time, is also a researcher in French university. And he has this chart that shows how many months of life baby humans can learn different concepts.

And you can measure this in various ways. So things like distinguishing animate objects from inanimate objects, you can tell the difference at age 2, 3 months. Whether an object is going to stay stable, is going to fall, you know, about 4 months, you can tell. You know, there are various things like this. And then things like gravity, the fact that objects are not supposed to float in the air, but are supposed to fall, you learn this around the age of 8 or 9 months.

If you look at a lot of, you know, eight-month-old babies, you give them a bunch of toys on their high chair, first thing they do is they throw them on the ground and they look at them. It's because they're learning about, actively learning about gravity.

Speaker 1: Gravity, yeah.

Speaker 0: Okay. So, they're not trying to annoy you, but they need to do the experiment, right? So, how do we get machines to learn like babies, mostly by observation with a little bit of interaction and learning those models of the world? Because I think that's really a crucial piece of an intelligent autonomous system. So if you think about the architecture of an intelligent autonomous system, it needs to have a predictive model of the world.

So something that says, here is a world at time t, here is a state of the world at time t plus 1 if I take this action. And it's not a single answer, it can be a-

Speaker 1: Yeah, it can be a distribution, yeah.

Speaker 0: Yeah, well, but we don't know how to represent distributions in high dimensional spaces so it's going to be something weaker than that okay yeah but with some some representation of uncertainty if you have that then you can do what optimal control theorists call model predictive control which means that you can run your model with a hypothesis for a sequence of action and then see the result. Now, what you need, the other thing you need, is some sort of objective that you want to optimize. Am I reaching the goal of grabbing this object? Am I minimizing energy? Am I whatever, right?

So there is some sort of objective that you have to minimize. And so in your head, if you have this model, you can figure out the sequence of action that will optimize your objective. That objective is something that ultimately is rooted in your basal ganglia, at least in the human brain, that's what it is. Basal ganglia computes your level of contentment or miscontentment, I don't know if that's a word, unhappiness. Okay.

Speaker 1: Yeah. Yeah.

Speaker 0: Discontentment. Discontentment. And so your entire behavior is driven towards kind of minimizing that objective, which is maximizing your contentment, computed by your basal ganglia. And what you have is an objective function, which is basically a predictor of what your basal ganglia is going to tell you. So you're not going to put your hand on fire because you know it's going to burn and you're going to get hurt.

And you're predicting this because of your model of the world and your sort of predictor of this objective, right? So if you have those 3 components, you have 4 components. You have the hardwired contentment objective computer, if you want, calculator. And then you have the 3 components. 1 is the objective predictor, which basically predicts your level of contentment.

1 is the model of the world. And there's a third module I didn't mention, which is a module that will figure out the best course of action to optimize an objective given your model. Okay? Call this a policy network or something like that, right? Now, you need those 3 components to act autonomously intelligently.

And you can be stupid in 3 different ways. You can be stupid because your model of the world is wrong. You can be stupid because your objective is not aligned with what you actually want to achieve. Okay. In humans, that would be a psychopath.

And then the third way you can be stupid is that you have the right model, you have the right objective, but you're unable to figure out a course of action to optimize your objective given your model. Okay. Some people who are in charge of big countries actually have all 3 that are wrong.

Speaker 1: All right. Which countries, I don't know. Okay, so if we think about this agent, if we think about the movie Her, you've criticized the art project that is Sophia the Robot. And what that project essentially does is uses our natural inclination to anthropomorphize things that look like human and give them more. Do you think that could be used by AI systems, like in the movie Her?

So do you think that body is needed to create a feeling of intelligence?

Speaker 0: Well, if Sophia was just an art piece, I would have no problem with it, but it's presented as something else.

Speaker 1: Let me add that comment real quick. If creators of Sophia could change something about their marketing or behavior in general, what would it be? What's...

Speaker 0: I'm just about everything.

Speaker 1: I mean, don't you think, here's a tough question. Let me, so I agree with you. So Sophia is not, the general public feels that Sophia can do way more than she actually can. And the people who created Sophia are not honestly, publicly communicating, trying to teach the public. But here's a tough question.

Don't you think the same thing is scientists in industry and research are taking advantage of the same misunderstanding in the public when they create AI companies or publish stuff.

Speaker 0: Some companies, yes. I mean, there is no sense of, there's no desire to delude. There's no desire to kind of over-claim with something that's done. Right, you publish a paper on AI that, you know, has this result on ImageNet, you know, it's pretty clear. I mean, it's not even interesting anymore.

But, you know, I don't think there is that. I mean, the reviewers are generally not very forgiving of unsupported claims of this type. But there are certainly quite a few startups that have had a huge amount of hype around this that I find extremely damaging and I've been calling it out when I've seen it. So yeah, but to go back to your original question, like the necessity of embodiment. I think, I don't think embodiment is necessary.

I think grounding is necessary. So I don't think we're going to get machines that really understand language without some level of grounding in the real world. And it's not clear to me that language is a high enough bandwidth medium to communicate how the real world works. I think for this- Can you

Speaker 1: talk to what grounding means to you?

Speaker 0: So grounding means that, so there is this classic problem of common sense reasoning, you know, the Winograd Schema, right? And so I tell you the trophy doesn't fit in the suitcase because it's too big, or the trophy doesn't fit in the suitcase because it's too small. And the it in the first case refers to the trophy in the second case to the suitcase. And the reason you can figure this out is because you know where the trophy and the suitcase are, you know when it's supposed to fit in the other 1, and you know the notion of size and the big object doesn't fit in a small object, unless it's a TARDIS, you know, things like that, right? So you have this knowledge of how the world works, of geometry and things like that.

I don't believe you can learn everything about the world by just being told in language how the world works, I think you need some low-level perception of the world, you know, be it visual touch, you know, whatever, but some higher bandwidth perception of the world.

Speaker 1: So by reading all the world's text, you still may not have enough information.

Speaker 0: That's right. There's a lot of things that just will never appear in text and that you can't really infer. So I think common sense will emerge from, certainly a lot of language interaction, but also with watching videos or perhaps even interacting in virtual environments and possibly, you know, robot interacting in the real world. But I don't actually believe necessarily that this last 1 is absolutely necessary. But I think there's a need for some grounding.

Speaker 1: But the final product doesn't necessarily need to be embodied. No. You're saying. It just needs to have an awareness, a grounding.

Speaker 0: Right. But it needs to know how the world works to have, you know, to not be frustrated, frustrating to talk to.

Speaker 1: And you talked about emotions being important. That's a whole nother topic.

Speaker 0: Well, so, you know, I talked about this, the basal ganglia as the, you know, the thing that calculates your level of mixed contentment, And then there is this other module that sort of tries to do a prediction of whether you're gonna be content or not. That's the source of some emotion. So fear, for example, is an anticipation of bad things that can happen to you, right? You have this inkling that there is some chance that something really bad is gonna happen to you and that creates fear. When you know for sure that something bad is gonna happen to you, you kind of give up, right?

It's not fear anymore. It's uncertainty that creates fear. So the punchline is, we're not gonna have autonomous intelligence without emotions.

Speaker 1: Whatever the heck emotions are. So you mentioned very practical things of fear, but there's a lot of other mess around it.

Speaker 0: But they are kind of the results of drives.

Speaker 1: Yeah, there's deeper biological stuff going on. And I've talked to a few folks on this. There's a fascinating stuff that ultimately connects to our brain. If we create an AGI system, Sorry.

Speaker 0: Human level intelligence.

Speaker 1: Human level intelligence system. And you get to ask her 1 question. What would that question be?

Speaker 0: You know, I think the first 1 we'll create will probably not be that smart. They'd be like a 4 year old. Okay.

Speaker 1: So you would have to ask her a question to know she's not that smart. Yeah. Well, what's a good question to ask, you know, to be

Speaker 0: impressed with? What is the cause of wind? And if she answers, oh, it's because the leaves of the tree are moving and that creates wind, she's onto something.

Speaker 1: And if she says, that's a stupid question, she's really onto something.

Speaker 0: No, and then you tell her, actually, here is the real thing, and she says, oh, yeah, that makes sense.

Speaker 1: So questions that reveal the ability to do common sense reasoning about the physical world.

Speaker 0: Yeah, and you know, some level of causal inference.

Speaker 1: Causal inference. Well, it was a huge honor. Congratulations on the Turing Award. It's a pleasure. Thank you

See all Lex Fridman transcripts on Youtube

Yann LeCun: Deep Learning, ConvNets, and Self-Supervised Learning | Lex Fridman Podcast #36