Transcript of Dr. Fei-Fei Li Interview on Huberman Lab

You and I are both vision students and scientists. It is estimated half of the cortical activities in human brain is involved in visual function. Children were first visual before they were verbal in development. So vision really to this day plays a central role in both the evolution of animal intelligence as well as in the daily life of human life.

Now, in parallel, vision as a discipline or as an area of artificial intelligence really played a pivotal role in what we see as this modern AI moment in a couple of ways. First of all is the algorithms, the neural network algorithms. Neural network algorithms were first— computer scientists started dabbling in that in the early 1950s. And Andrew, you might remember what’s happening on the neuroscience side in the early 1950s is that neuroscientists like Hubel and Wiesel were starting to record visual cells in mammalian brain and starting to realize there is a hierarchical structure of nervous cells that stack against each other and pass neural information across this hierarchy. And it goes from you know, collecting light from retina all the way to recognizing there is a shape in front of you. And that very neural architecture that we see in mammalian brain is also part of the inspiration of neural network algorithm.

Now, today’s neural network algorithm runs on hundreds of billions and even trillions of parameters. It has the complexity that departs from what we recorded in the mammalian brain or the visual pathway. But the origin is very close to each other. About half a century ago, a little more than half a century ago. That’s one aspect of vision’s contribution to AI.

There is another aspect of vision’s contribution to AI that is also pivotal, which is through big data. Is that that comes closer to my own work, is that AI around the century was a field of machine learning. A lot of different labs, different research scientists were trying out different algorithms. And it’s not just neural network. There are other methods, jargon words like Bayesian methods, support vector machine methods. It doesn’t matter what these methods are, but it’s an explorative phase. That we’re trying to get these algorithms to work so that we can empower the machine to read or to see.

A group of us computer vision scientists were struggling with these algorithms. And I was a very young faculty, first-year faculty, 2006 at Princeton. And my students and I are looking at these algorithms and how little data were fed into these algorithms to learn. So I turned to cognitive neuroscience literature, namely vision literature, and started to study how much humans learn, how much humans can see. And the numbers were incredible. Humans were, by age 6, can learn tens of thousands of different object categories. And the exposure to visual world is also massive, right? Babies can see the mo— most of the time the moment they’re born. So they’re inundated with this big data.

So we conjectured that the lack of data was a huge part of the reason that’s the lack of progress in AI. So we took a departure from everybody else who are really focusing only on algorithm. And said that we need data. We need data to drive these algorithms. So long story short, we led this ImageNet project that collected the first ever internet-scale large dataset for the field of artificial intelligence, but really through the field of vision because ImageNet is a collection of 15 million images. And the goal of ImageNet was to drive machines to recognize everyday objects, you know, microphones, cups, chairs. And that work converged with the advances in neural network algorithm as well as in GPU computing. And by 2012, that work, that convergence of the 3 elements of modern AI became the defining moment of what modern AI is.

From Face-Blindness to Superhuman Precision: The 2012 Turning Point

ANDREW HUBERMAN: (00:12:13 – 00:12:49): I recall somewhere around 2012, it seems there was this debate at this vision course at Cold Spring Harbor that was held every other summer. Like, could a computer learn to recognize specific faces as well as humans? Now I think most people would say, computers are actually much better at it than humans are, even though you have these super, super recognizer people who are exceptional at this. Could you tell us how is it that this technology went from a state basically where it would confuse you and maybe a cousin or even someone that looks somewhat like you?

DR. FEI-FEI LI: (00:12:49 – 00:12:50): A kooky.

ANDREW HUBERMAN: (00:12:50 – 00:12:56): Or to the point where, to the point where now it is exquisitely precise.

DR. FEI-FEI LI: (00:12:56 – 00:16:57): How do we get here? I want to definitely double, triple click on the convergence of this technology. I think around the second decade of 21st century.

So like you said, around 2012, the huge convergence was the capability of GPU computing, which basically accelerated or parallelized computing so that you can have more flops going through algorithms, right? You need that speed. Then you also have a— after many decades of research, neural network algorithm is getting more mature. You know, starting, as we said, 1950s, people start to create these very simple algorithms that behave similarly to neurons but much simpler. Neurons, as you know, are very complex. But here the idea is that you have one unit of node that takes some input and outputs another input. And within it is just a function, a very simple function. So you stack them together. That’s what neural network is. But by the time it’s in the— after, you know, around 2010-ish, the maturity of these algorithms have gotten to a level that it’s becoming really good. But also last but not the least, the recognition of big data. Internet definitely fueled that. It made data more available. But the reckoning moment of, wow, big data needs to be part of that equation. We need to use big data to drive these algorithms, to learn these patterns. So this convergence of these 3 things really set off the revolution of AI.

The specific moment is also worth mentioning because you mentioned face recognition, is this ImageNet challenge my lab put forward that starting 2010, after we collected this humongous dataset, we at that point GPU was not yet mature and we put out a public challenge for the research community for multiple years in a row. And invited people to solve this major computer vision problem called object recognition. The task was very easy. We have a dataset of 1,000 different categories of objects, and this dataset is more than 1 million images large. It’s what we call the testing dataset. And the task for the algorithm is, I’ll show you a picture, you have to name the main objects inside. And if you guess right, you get a point. If you guess wrong, you don’t get a point.

So that ImageNet challenge, we later, a couple of years later, benchmarked human performance by a very smart graduate student at Stanford, and that was roughly 4%. So random chance will be 1 over 1,000. Right. So 4% for humans is not that bad. The first few years, machines were not as good as humans. The turning point was 2012, the convergence of neural network, ImageNet dataset, and GPU. Even that year, even though the error rate was cut to— by the way, the human performance error rate was 4%. Sorry, I need to correct that. The error rate was cut down to the teens. It wasn’t where human performance was.

ANDREW HUBERMAN: (00:16:57 – 00:17:00): So this is looking at images and assigning a—

DR. FEI-FEI LI: (00:17:00 – 00:17:02): 1 out of 1,000 labels.

ANDREW HUBERMAN: (00:17:02 – 00:17:03): Got it.

DR. FEI-FEI LI: (00:17:03 – 00:17:38): Yeah. But 2012 was so momentous that year because the error rate from previous algorithm dropped a lot by this neural network algorithm. And we know in the research community when something this drastic happens, it means an inflection point. But it It still took another 3 years, I remember, by 2012, 2016, for the algorithm to beat humans in naming 1,000 objects.

ANDREW HUBERMAN: (00:17:38 – 00:17:53): Could I ask you where this 4% error is coming from in this very smart graduate student? Is it that they don’t recognize the objects or it’s a recognition against time pressure? Like they have to, they’re being fed images fast enough that occasionally they do an incorrect assignment.

DR. FEI-FEI LI: (00:17:53 – 00:18:34): I don’t think the time pressure was the main issue, even though for a graduate student to do this, I don’t think they want to do this forever. Um, but I think, you know, the, the human brain, as you know, has limited memory, whether it’s long-term or short-term, right? So retaining the patterns of 1,000 object classes, even if some classes you’re, you’re familiar, is, is not that easy, you know. So, so I think there is the confusion. And, and also, for example, different species of dogs gets really close, and that, that’s a challenge.

A Word From Our Sponsors

ANDREW HUBERMAN: (00:18:35 – 00:21:43): I’d like to take a quick break and acknowledge our sponsor, Lingo. Lingo is an everyday wearable that tracks your glucose 24/7. Glucose drives a lot of key processes that support energy, body composition, and long-term health. When glucose is constantly spiking and crashing, that’s where we can start to see metabolic dysfunction, and over time, that can even progress to prediabetes.

Right now, about 115 million adults in the US have prediabetes. Most don’t know it, and a higher percentage of men have it than women do. Often, there aren’t clear symptoms of prediabetes early on, so people don’t tend to look into it. But the fact is that metabolic health is shaping how your body functions every day, Whether you feel it or not. Tracking your glucose with Lingo can help you see how food, activity, and stress impact your glucose throughout the day.

I personally have used Lingo, and it’s been an invaluable tool for improving my metabolic health. If you would like to try Lingo, Huberman Lab listeners in the US and UK can save 10% on a 4-week plan. Just visit hellolingo.com/huberman for more information. Terms and conditions apply. Again, that’s hellolingo.com/huberman.

Huberman. Today’s episode is also brought to us by Wealthfront. In today’s financial landscape of constant market shifts and chaotic news, it’s easy to feel uncertain about how to save and invest your money. Wealthfront is the solution that helps you take control of your money while managing risk. For nearly a decade, I’ve trusted Wealthfront to navigate this volatility. With the Wealthfront Cash Account, I can earn 3.3% annual percentage yield, or APY, on my cash from program banks. And I know my money is growing until I’m ready to spend it or invest it.

One of the features I love about Wealthfront is that I have access to instant, no-fee withdrawals to eligible accounts 24/7. That means I can move my money where I need it without waiting. And when I’m ready to transition from saving to investing, Wealthfront lets me seamlessly transfer my funds into one of their expert-built portfolios.

For a limited time, Wealthfront is offering the Huberman Lab audience an exclusive 0.75% APY boost over the base rate for 3 months, meaning you can get up to 4.05% variable APY on up to $150,000 in deposits. Over 1 million people already trust Wealthfront to save more, earn more, and build long-term wealth with confidence. If you’d like to try Wealthfront, you can go to wealthfront.com/huberman to receive the boost offer and start earning 4.05% variable APY today. That’s wealthfront.com/huberman to get started. This is a paid testimonial of Wealthfront. Client experiences will vary. Wealthfront Brokerage is not a bank. The base APY is as of January 30th, 2026, and subject to change. For more information, please see the episode description.

Beyond Vision: Speech, Sound, and Natural Language

ANDREW HUBERMAN: I can see the rationale for doing this in the vision domain, but has a similar thing been explored with hearing with sounds. I mean, it’s, you know, as humans, we are amazing at recognizing speech inflection, emotional tone, things like that. But if I had to discriminate, you know, even 15 different sound frequencies, I can tell you as a non-musician, it would be very difficult for me.

DR. FEI-FEI LI: (00:21:43 – 00:23:37): Absolutely. I think that what you see is the floodgate got open and every sub-area of AI, whether it’s speech recognition, sound recognition, natural language, processing, which is more than recognition, vision, all areas got really a boost in terms of the technology. We have colleagues at Stanford who are studying whale sound, right? Whale songs using machine learning and AI now.

And speech recognition is another area that did so well in the early days of this AI revolution. And of course the technology continues to advanced. By the time the transformer paper was published around 2016, 2017, it quickly showed that it is even more powerful than the early ImageNet, AlexNet algorithm. There, it was not the field of computer vision that made the next big progress. It’s the field of natural language processing. So because The recipe hasn’t changed. Now we have an even more powerful neural network algorithm called Transformer, but we have even more data on the internet from, at least more readily available data on the internet in the form of texts. And now we have more powerful GPUs. So companies like OpenAI and Google quickly rallied beyond this this very important technology. And it still took about 5 years from 2017 to 2022 to get to the ChatGPT moment in natural language. But that’s yet another step forward.

How a Child Learns “Cat”: Object Recognition and Context

ANDREW HUBERMAN: (00:23:37 – 00:25:26): So I think for people who are not computer scientists nor neuroscientists, the natural human experience will perhaps resonate with them. And maybe I can just frame my question through that lens.

So when a child learns that there’s something called a kitty cat, they go, oh, cat. Then they usually drop the kitty part. They say kitty, and then they learn cat. And if they have enough interactions with a cat, they’ll realize what a cat is, even if they see it from the side, from the back. And eventually, if they see a tail that looks a little bit like a cat and it’s you know, behind some books, you say, what is that? They’re very likely to say cat, even if they’ve also seen foxes and other animals with tails, just based on their experience. They’re making a probability judgment. And that’s essentially what AI can do. That’s essentially what machine learning can do.

But it seems to me that there’s a key moment that had to happen in the progression of, you know, from calculators to the AI we have now to be able to see an image of a tail and make the reasonable assumption that it’s most likely a cat if it’s indoors or something like that, because foxes generally aren’t indoors, this sort of thing. So at what point did machine learning and AI gain the ability to do kind of contextual learning and come up with the most likely assignment of what something is? Because it’s one thing to show apples and bananas and oranges, they’re all fruit. Okay, you could distinguish them, you could distinguish those from cars and trucks, et cetera. But this object constancy piece, that if something is moving, you’re only getting a partial image, This isn’t what most people will think of in terms of intelligence, but it’s part of what makes our brains and the brains of other animals, but especially our brains, so remarkable and why we consider ourselves probably the smartest species on Earth. And if not the smartest, then certainly the best at technology development.

DR. FEI-FEI LI: (00:25:27 – 00:25:27): Yeah.

ANDREW HUBERMAN: (00:25:27 – 00:25:34): So when did AI achieve this and how was that scripted into these computers to allow them to do that?

DR. FEI-FEI LI: (00:25:34 – 00:28:40): So let’s just take the problem very— you have described it so well, this problem of seeing a glimpse of a cat and be able to recognize cat, right? Or assign a high likelihood there is a cat.

The interesting thing is, Andrew, generations of machine learning computer scientists have tried this problem. So before today that machines could reliably do it, there were different algorithms. You know, you can imagine a common sense way of thinking about this is, oh, maybe we should recognize all the furniture to know it’s indoors, so it’s unlikely to be a fox. So though there are rules like that, that it was built into previous generations of algorithms, there are also rules like, well, let’s only— instead of guess it’s a cat, let’s only guess 1 out of the 10 potential animals, you know, cat being one of them. That limits the search or guess space, and that would help. So many ideas were tried.

So when was the moment it became much more reliable? It’s this current era when the huge data that these algorithms have learned— let’s take Gemini or GPT— have learned really created the capability in the machine’s learned space, so much knowledge, so much pattern that when presented with this more or less maybe a newish photo of a cat’s tail sticking outside of a bookshelf, that pattern activated the learned, what we call learned weights or learned parameters that put put the machine’s assessment or guess of this object closer to what it has seen, which is likely to be a cat tail or just tail because there’s just so much data.

This is where, Andrew, as neuroscientists, I think we depart from a human brain because that child who learns about when you say kitty cat, will not have the chance to download the internet of images of cat. They likely have seen 3 cats, 10 cats at most, but yet they’re able to identify that tail as a cat tail instead of a fox tail through a different kind of learning pathway. These are the mysteries we haven’t fully solved. But I do want to point out that departure between today’s AI algorithm that is learned with a humongous amount of data versus how humans have evolved.

From Stills to Motion: How AI Learned to Animate a Cat

ANDREW HUBERMAN: (00:28:41 – 00:30:18): If we continue to ascend the hierarchy from simple object recognition to what you and I would call higher-order brain functions, like moving more towards what most people— they hear the word intelligence and they just think, oh, it must be some higher-order thing. Creativity, imagination. Let’s go to a middle step and then a much further step out.

So staying with the cat example, if a computer or a child learns to recognize a cat through the tail, the whole thing, whatever, and they’ve seen a cat move, it’s a very new world at that point for that brain, that child, or that computer, because now they know that the cat generally moves in the direction of its head, not its tail. These are simple learning rules, right? It might go after mice, but it might run from dogs. Maybe yes, maybe no, and on and on. And so it seems that the next layer up in terms of quote-unquote intelligence is to assign likelihoods of direction to move, directions not to move, other objects that that object is likely to interact with. This all sounds very basic to people, but like, this is how brains learn and this is how machines learn.

So when was the next sort of big inflection in terms of like giving a computer, AI, a picture of a cat and saying, animate this cat for me, make it move like a cat, without giving it any specific instructions about how to move its limbs, et cetera. But I would imagine that was a pretty quick but a remarkably important transformation in this whole thing that we call AI, because that’s what a brain does.

DR. FEI-FEI LI: (00:30:19 – 00:33:34): Yeah. So it’s really funny you asked this, and you put it beautifully. I never thought it to put it in this way for a public audience. But that moment came when video became part of the training data.

So see, again, I’m going back to the training data. So around 2023, very shortly after ChatGPT moment, multiple research teams start to put video into the training data. Of course, I’m not going to get into the nuance of the algorithm. There is a little bit of changes and variations. So remember January 2024, Sora was released, and that’s where people see a video can be generated. Literally what you just said. People can then type and say a cat running towards a mouse, and then a few-second clip would be generated. There would be a cat moving its leg in a plausible way, running towards a mouse. At that time, there were still mistakes. Still, even today, it’s not perfect, but things have gotten a lot better. But that opened the floodgate of video generation, as you described it.

So what happened there? What happened there is actually not as revolutionary as you I think, because the bottom line is it’s still data. As a scientist, I can tell you there are all kinds of algorithm tweaks and changes and improvements and all that. But overall, if you zoom out, it’s still part of this great neural network era, right? But what happened is that we are now able to process video data in a way, again, some clever engineering, tokenize it, whatever you call it. And now we can generate these short clips of videos, which is frames put together that look like plausible cat movement.

Now you might ask, does the algorithm know the muscle structure of a cat’s legs so that when the algorithm shows that the cat is moving in a plausible way with the paws, you know, in a sequence? I would say the algorithm doesn’t. But what it does have is so many videos, especially cat on the internet, so many videos of cat. So it learned what it should look like. So in a way, humans do that. Most of us without education would not know how muscles move. In cats. I still don’t know. You know, our colleagues in medical school might know, but we have just got so used to seeing cats moving this way that we have a plausible idea of how cats move. So that is similar. That’s how similar AI is. It’s the statistics. It’s the large amount of data that showed you what is the plausible generation of cat movements. Yeah.

ANDREW HUBERMAN: (00:33:34 – 00:33:42): So when people have heard almost certainly that the brain is a prediction machine, it’s a learning machine, this is exactly what you’re referring to.

DR. FEI-FEI LI: (00:33:42 – 00:33:43): Yeah.

ANDREW HUBERMAN: (00:33:44 – 00:34:09): Let’s go to a really far out there aspect of brain function that we know exists in humans, which is thoughts and creativity. Now, there are probably rules for thoughts and creativity, they’re a little bit harder to tack down than, um, examples from the visual system. Like, if it’s a tail and it’s indoors, it’s likely a cat, this kind of thing. But they’re there. The rules are there.

The Limits of the Internet: Abstraction, Emotion, and Creativity

DR. FEI-FEI LI: (00:34:10 – 00:34:27): If you use apple as an example, we could have gone from low level, seeing an apple, to mid-level, seeing apple always drop, not fly off. At the highest level, what is the equation that governs the apple’s movement.

ANDREW HUBERMAN: (00:34:27 – 00:37:08): Right. So that’s ascending to like a higher order, more reductionist analysis. What do you think about the idea that while AI is indeed intelligent, it can do things that brains can do, maybe even— well, certainly things that individual human brains can’t do. We know this by virtue of beating humans at chess and this sort of thing.

The idea right now, as I understand it, is that AI is trained on the internet. Images, discussions, videos, songs. But that’s not all of human cognition, right? So are there aspects of AI that are— whether or not it’s ChatGPT or it’s Claude, or even the most powerful not-yet-released machine learning and AI tools— that don’t have access to features of human brain function yet because they’ve never been uploaded to the internet, at least not in a way that the AI can pull out. So, for instance, you know, you could put a symphony there, and it follows certain rules of music and mathematics and sound like that. That makes sense. But you have thoughts all day long, and I have thoughts all day long that don’t quite mesh with language in a way that I can just type them out on the internet.

Stay with me here. I know this is a long question, but I feel like this is the one thing you are perfectly poised to answer, and I’ve been waiting to ask you this for a year and a half since I saw you in Utah. In the world of art, we have this thing called abstraction, right? And occasionally somebody will come up with a painting or a drawing that it doesn’t look like anything specific. This happens in music too, where you just feel something like there’s like a fundamental rule or an emotion associated with it, like they’ve tapped into some aspect of brain function, but you can’t say what it is.

I feel like this is the sort of thing that is complicated for AI, or for me to understand how AI could do, because you can put that piece of art into AI and say, you know, what fundamental feature of human, uh, experience does this reveal? And it only has access to what’s on the internet. So how can you capture a complex constellation of feelings and experience with AI? That seems to be the gap for me, and I’m sure we’ll get there with AI. But I’m not seeing from neuroscience to AI in any kind of direct way, the same way we could ratchet through visual motion, sadness, happiness. You could pull out a lot of things, but it’s hard to get to these higher-order abstract representations that can’t be spoken or written down or drawn. If I just say, give me your example of whatever nostalgia for your childhood home, you could write about it, but those are just words. It’s not I can’t understand your experience at a first-person level.

DR. FEI-FEI LI: (00:37:09 – 00:41:48): Totally. Andrew, I know you put a lot of thoughts into this question, and I think it’s a very important question. And let’s peel this one step at a time. First of all, TL;DR, short answer is I agree with you that we do have to be very careful recognizing what AI can do. Is likely to do, not conjecturing over 100 years or whatever. I recognize what you just said are these extremely nuanced, personalized, hard to characterize, or not even captured human cognitive behaviors. And because they were not captured, then they were not uploaded on the internet. And we don’t have— today’s AI doesn’t have a way to do that.

So when you call internet, which is the source of AI’s data, let’s be very clear. What is internet? Internet is not some random thing. Internet is the biggest collection of human behavior in multimodal forms. Let’s break it down further. Internet has the world’s population typing on it for many, many— at this point, multiple decades. That typing is a sensing mechanism that captured everything from teenager chitchat all the way to deep scientific articles who did— got digitized and get uploaded, right? So that capturing human language is what internet is super good at.

Then internet captures images. How? Because we now have digital cameras that’s so prevalent in smartphones and digital cameras so that humans love taking photos from, you know, the cat in your house to selfies to beautiful, you know, BBC-captured photos. Those also got uploaded in our digital sphere. On top of that, there’s videos. Videos now has sound, has movements. That also got uploaded to our digital sphere. On top of that, there’s music. We’re not even getting into the legal discussion of copyrights, but let’s just table that aside. I’m just talking about the forms of data. The speeches, and singing and music and orchestra, that also got uploaded into the digital sphere.

So now we have created this humongous library of human knowledge in words, human behavior in videos, human expressions, or even nature’s whatever in sound. And now AI gets trained on that. That is why it’s so powerful. This is why, especially in the words front, that AI can recognize patterns, can synthesize patterns because so much of this is already there.

But the thing that you just talked about, that when let’s say Picasso had that incredibly profound thought about that particular way of expressing that, that portrait of the, of the young woman, that thought has never been captured. In fact, as neuroscientists, if I ask you which brain area did that thought come from, you don’t know, right? Is it Broca? Is it V1? Is it motor? Is it prefrontal? We don’t know. Maybe it’s diffused everywhere because that thought is so personalized, so special. You can call it creativity, you can You can call it emotion, you can call it whatever you want. You can call it Cat231, whatever name you can give it. That thought is not captured. Therefore, it’s not on the internet. Therefore, AI has not seen it. So that is where humans still remain so unique.

But we also need to give credit to AI because AI has learned so many things. It can combine information in highly creative way. Did you remember move 37?

Move 37: A Different Kind of Creativity

ANDREW HUBERMAN: (00:41:48 – 00:41:50): This is in AlphaGo, right? Right. Yeah.

DR. FEI-FEI LI: (00:41:50 – 00:45:48): Move 37 has symbolized AI’s creativity. I think it’s both true but can be taken out of context because that was a game when AlphaGo was playing Lee Sedol. And I think it’s a 3rd game out of the 5 games that AlphaGo as a computer algorithm made a move that the human masters of Go never thought about. And that is an incredible move, right? Because if really humans collectively, these are the masters, never thought about it.

But if you really go deep into what AI did there, it was because first of all, Go is a highly mathematical game. It has very clear mathematical objective, very clear mathematical rules in terms of move. So when AI having the bigger compute, and ways to retain how many moves it can remember. It was able to do things that human brains don’t typically do. So is that called creativity? I think it is. But we do have to recognize that’s a special kind of creativity.

I was talking to an incredible mathematician of our time, and I was asking him about the unsolved problem of mathematics and how AI can contribute to that. And he was very positive. He said there are many problems in today’s mathematics, as hard as they are, even as, say, a field medalist, I probably have forgotten there are known methods in math that can solve these problems because I have a human brain. I don’t remember and I don’t know all of math’s, you know, solutions in the past hundreds of years, even if I were a field gold medalist. So AI can help us to solve these problems. But as a mathematician, he was also telling me, he said, I don’t know if AI can solve all of math problems because some of these math problems require solutions that have not been invented, that will push creativity to a whole different level.

And this is where, you know, I’m— we should be curious. Is it going to be human creativity, or AI would go through its iterations of improvement and get to a point of creativity that humans don’t have? Or is it a combined creativity? My current conjecture is hybrid, is that humans working alongside AI would help us to solve these problems whose solutions have yet to be invented.

And then And what you said, especially you touched on emotion, is even more personalized. This is not necessarily logic. This is not necessarily deductive reasoning. This is maybe, Andrew, you look at this cup and say it’s a gray cup. What if it evoked an emotion in me, a childhood moment that a gray cup might mean something that only me and my best friend share? That is a completely inaccessible piece of information in my brain that is never uploaded on the internet. And no matter how mighty AI is today, cannot access that. So that my reaction to this cup and potentially what I would do with it because of that piece of memory can be completely different. You can call it creativity. You can call it expression. You can call it storytelling. You can call it in many ways, but that’s where AI cannot access.

Brain-Computer Interfaces and the Future of Self-Knowledge

ANDREW HUBERMAN: (00:45:49 – 00:48:12): I feel like at some point in the not too distant future, computers will have access to our brain activity in non-invasive ways. So, you know, like I might even imagine in 5, 10 years I’m wearing something on my head right now. You can’t see it. It’s a very, very fine hairnet. Hairnet makes it sound like it was whatever, like some electrodes that are just there on the outside of my skull, not bothering me, sensing my activity inside the brain, maybe also sensing my heart rate, autonomic activity, how alert I am, and comparing that, yes, to what I’m saying and what I’m doing.

This is all totally within reach and is going to happen. You and I both know this, and it’s probably already starting to scare people. But let’s keep it benevolent, right? There’s this world where a computer that I own And I’m not worried about data getting out or anything like that. We can manage that problem. Is sensing all these aspects of me and is picking up on the fact that, yes, what I say might be important, but there are aspects of my internal state and brain activity that I’m not even aware of. Yeah. And I can decide to collaborate with this aspect of me and say, let’s, let’s come up with a really interesting picture that I’ve never seen before, but comes from some experience of mine that’s important. Important based on whatever. Like, and, and it could reveal that to me because it has access to my— to unconscious features of my brain activity.

I think this is very likely to happen in, in the not too distant future. And perhaps if people thought about it within the bubble of their own experience, like, this isn’t immediately going to the internet, or it’s not going to be used against them, you’re actually learning about yourself. Of course. And, and I feel most people have an inherent interest in what’s going on for them, also with other people, thank goodness. But they’re, I think, amazing. I would love to know why I trip up in certain ways and don’t have the best day, or why some days I have the best day, or where ideas come from in me, what states I could kind of elaborate on. But I’m not going to know how to do that except, okay, 1 cup of coffee, good. 1.5, a little better. 2 is too much. Right now, if you think about how primitively we go about this, it’s kind of crazy. It’s crazy. And everyone has a different method, and we all try and get this right. And then you’ve aged enough by the time you get it right that then you have to update it. And like, we’re probably not getting the most out of our biology and our brains at all right now.

AI as an Amplifier of Human Agency

DR. FEI-FEI LI: (00:48:13 – 00:49:41): No, we’re not. And this is why I keep saying, this is why it bothers me when people talk about AI. Some people make it sound like it’s replacing humanity, but what we really— what you describe is about enhancing and augmenting humanity, right? This is where it doesn’t even have to go as sci-fi as a smart hairnet accessing your brainwaves. Just AI learning your patterns of writing can already help you to be, you know, a better communicator, a more effective communicator, a more efficient communicator. And that is an empowering capability that we could unleash in today’s AI.

I think one of the most important things, Andrew, that as a neuroscientist and also faculty, we know is agency is so important for humanity. You know, that boils down to motivation, agency, and dignity at every individual level. And I think we need to recognize that we need to think about AI as a tool that helps us us in our agency. It does— it should not take away our agency. And people who lead in today’s AI should not try to talk like that this, this work will take away agency from people.

ANDREW HUBERMAN: (00:49:42 – 00:50:28): Yeah, I think people who are very familiar with the technology, whether it’s computers or it’s biology or any technology, cars for that matter, We, they become such nerds of that thing that we forget that it can be scary to people and that the languaging around it is essential. It is. And I remember a time in the early ’90s, I’m sure you remember this too, when genetic testing was viewed as this thing like, would you want to have it? Would you want to do a blood test? Because, oh my goodness, you might see something that could really scare you. And that discussion is happening now around self-elected MRIs and things like that, none of which people have to do. But I come from the stance like more information is better. But I’ve come to understand that not everyone feels that way. Some people don’t want to know. They don’t want to know.

The Right to Choose: Genetic Testing and the Ethics of Communicating Science

DR. FEI-FEI LI: (00:50:28 – 00:51:23): Yeah, but they should have the choice. In the meantime, we should have enough public education and communication to let people know the pros and cons, but not to deny them the choice and also not to take away, you know, and say, well, since you don’t understand this, let me decipher you what’s good. That is not good, you know? And the rhetoric around AI right now is getting really skewed because people who know what this is tend to talk down at the public. It tend to talk, whether the motivation is a positive one or negative one, there is a rhetoric of, You guys don’t know what this is, and I will tell you, and I will make you whether happy, safe, whatever it is, and I would decide for you. These are not healthy and not helpful.

ANDREW HUBERMAN: (00:51:24 – 00:52:03): Yeah, I agree. And I think, you know, one of the reasons for starting this podcast was to showcase the scientists and physicians who really have a benevolence about them, and they have no interest in dumbing things down, but they do have an interest in people understanding things. Yes. And many people would feel that health information is among the more important things to understand. Absolutely. Well, thankfully, you’re breaking the mold of the phenotype you just described. And there are a few others, but you’ve been doing this at the highest levels, really encouraging people to think about the collaboration that is AI, the agency that exists, and whether to use it or not to use it, and so forth.

DR. FEI-FEI LI: (00:52:03 – 00:53:04): One of the agency I do think is important for individual humans, whether you’re a student, a teacher, a doctor, policymaker is learn about this. Not necessarily learn about how to code. I don’t think it’s necessary depending on your job, right? So for example, if you’re an artist or if you’re a teacher or doctor, you don’t necessarily need to code. But learn about what this technology is. Learn about how you can use it yourself to empower yourself, your learning or your work or your expression. By learning By learning, one feels more in control. By learning, you’re less scared of trying. And by learning, you retain that agency and that dignity because at the end of the day, no matter how advanced technology is or medicine is, as humans, we want that benevolence that helps us to live better, keep our dignity, and make our community better.

A Word From Our Sponsors

ANDREW HUBERMAN: (00:53:05 – 00:56:38): I’d like to take a quick break and acknowledge our sponsor, AG1. I’m excited to share that AG1 has just launched their newest formulation, AG1 Pro. AG1 Pro takes the clinically backed AG1 formula, which is a blend of vitamins, minerals, probiotics, and adaptogens, and adds 3 important new ingredients: creatine monohydrate, calcium HMB, and zinc carnosine.

Each serving has 5 grams of creatine monohydrate to support muscle strength and performance, as well as brain health, calcium HMB to support muscle recovery and reduce muscle breakdown, and zinc carnosine to support and improve the lining of your gut. All 3 of these ingredients have compelling science to support them, and therefore I love seeing them added to the existing AG1 formula. As most of you know, I’ve been taking AG1 every day for nearly 14 years now. I started taking it long before I even knew what a podcast was. It’s a great product, and it’s now made even better with the new AG1 Pro formula.

Formula. If you would like to try AG1 Pro, you can go to drinkag1.com/huberman to get a special offer. AG1 is giving away a free bottle of omega-3 coenzyme Q10 with your first subscription. Again, go to drinkag1.com/huberman to get a free bottle of omega-3 coenzyme Q10 with your first AG1 subscription.

Today’s episode is also brought to us by Element. Element is an electrolyte drink that has everything you need and nothing you don’t. That means the electrolytes— sodium, magnesium, and potassium— all in the correct ratios, but no sugar. Proper hydration is critical for brain and body function. Even a slight degree of dehydration can diminish your cognitive and physical performance. It’s also important that you get adequate electrolytes. The electrolytes— sodium, magnesium, and potassium— are vital for the functioning of all cells in your body, especially your neurons or your nerve cells. Drinking LMNT makes it very easy to ensure that you’re getting adequate hydration and adequate electrolytes.

My days tend to start really fast, meaning I have to jump right into work or right into exercise. So to make sure that I’m hydrated and I have sufficient electrolytes, when I first wake up in the morning, I drink 16 to 32 ounces of water with an LMNT packet dissolved in it. I also drink LMNT dissolved in water during any kind of physical exercise that I’m doing, especially on hot days when I’m sweating a lot and losing water and electrolytes. LMNT has a bunch of great tasting flavors. In fact, I love them all. I love the watermelon, the raspberry, the citrus, and I really love the lemonade flavor. So if you’d like to try Element, you can go to drinkelement.com/huberman to claim a free Element sample pack with any purchase. Again, that’s drinkelement.com/huberman to claim a free sample pack.

The idea that technologies can be connectors as opposed to separators, I think, has to sit at the center of the discussion. Yes. And we all know who they are, that there, there are several of them, but the big names in this field, you know, they are also in a developmental process where they’re learning how to be public-facing, and it happens very fast. Like, you know, the microscope is on them and the cameras are on them. And so every subtle dysfunction is magnified. So I like to think that they will mature quickly enough to realize that, and I think they are, that some are, that the public needs to hear the correct, the true message, but in a way that makes them understand. That’s the kind of dirty secret of medicine and academia that you break this mold. I like to think I break this mold is that there’s a power in not sharing how things work, but it doesn’t serve anybody well. At the end of the day, like you pull back the veil and let people in and people feel safer.

DR. FEI-FEI LI: (00:56:38 – 00:57:32): Yeah, there is a power in not sharing. There’s also a power to say, just trust me, I will tell you. And neither as educators, that is, we don’t go to our lectures and say, just trust me, you know, 2 2 4. We actually say, here’s how you break it down and learn about it so next time you can do it yourself, right? I also think that especially your podcast is so important as part of public communication and education of knowledge. I also think that we need to hear voices of different background, right? So because there are plenty of scholars, technologists, builders, thinkers out there who have been dealing with AI, using AI, thinking hard about how to use AI to empower people.

AI and the Future of Health Discovery

ANDREW HUBERMAN: (00:57:33 – 01:00:29): And these voices are so important. Well, certainly I’ll take names of people to host in addition to you. But since you’re here, I’m going to go next to something that I think most everybody would agree would be a wonderful thing if it existed. And it’s already starting to happen, which is the use of AI to augment health discovery, treatment of disease, and so on.

So using the AlphaGo example from before, and people surely still remember the cat example, those just follow certain rules. AlphaGo is a very complicated set of rules, but if you learn them, there’s a constrained set of rules. With the cat, it seems unconstrained, like infinite possibilities, but it’s constrained enough that machines and humans can learn it really well. When you start getting into medicine, there are rules of medicine, there are rules of science. You have a question, you pose a hypothesis, you test the hypothesis, you try and rule out your hypothesis, and so on, like the scientific method. And in medicine, every field has its methods. We observe, we observe disease, we observe who recovers, we have a case report, we do a randomized controlled trial. So there are rules, and the internet knows these rules. So LLMs can be used to mine health information very well because there are constrained rules.

Rules. But I think you and I both know, because I also consider you a biologist, that the rules of biology are still revealing themselves to us. Which is not to say that the dermatologists, neurosurgeons, and oncologists don’t know what they’re doing, but they’re doing what they’re doing within a constrained set of rules that they learned. And even if they continue to learn and update them, it’s every month it seems now that a discovery comes out that violates the rule. Like, I that action potentials are unitary. They always look the same. You either fire or not. But there was a paper not but 12 years ago that showed that the shape of an action potential can vary quite a lot. It was published in Nature. Everyone saw it, and then no one wanted to deal with it. It’s just too much. It changes the rule. Yeah, neurons are supposed to be either graded or all-or-one, and the all— I mean, it’s in every single textbook. So now if I take a bunch of neural activity and I give it the rule, oh, well, you know, action potentials can be big, they can be small in the same neuron, it completely confuses everything we understand about neuroscience, and it just— our understanding of the brain just breaks down to zero. Yeah. But if you gave AI the rule that it could be, you know, 100 different shapes of this signal, well, AI could probably do a lot more than even the very, very best graduate student at, dare I say, Stanford, or to be fair, MIT or Caltech. I don’t think And I think it can do it, and it can do it, like, in the duration of this question, which admittedly is a bit long. So I’d like to get your thoughts on how is it that humans in healthcare, the general public, and AI can collaborate to help solve disease and ideally come up with new rules for discovery so that we can finally understand our biology at a level that can really change the course of humanity for the better.

DR. FEI-FEI LI: (01:00:30 – 01:03:02): Yeah. No, Andrew, this is probably— perhaps you touched one of the most exciting usage of AI, discovery. And in the case of biomedicine, scientific discovery directly connects to human health and diseases.

I think we’re ready for a complete rewriting of how scientific discovery can be done because for ages, I don’t even know how long, it relies on smart humans retaining what they have learned from other smart humans. Doing things at the speed of our own muscles, I guess, you know, most likely. Of course, there’s like supercolliders and all that. But by and large, the ways of doing scientific discovery, human brain or scientist brain are the only central character in this process. Now we have a new tool. Whose brain that can retain humongous amount of information, can help us synthesize knowledge, can go across disciplines in ways that you and I cannot go. So for example, we happen to be both in the vision neuroscience AI domain. I know nothing about, you know, olfactory, zero. Like I don’t even know how to spell most of probably these words that our colleagues know, right? So it’s so hard. Hard for our brain. But now we have a tool that can break open.

So I think that we need to change. We need to use this tool. We absolutely— I was just thinking 150 or I don’t know exactly when years ago, electricity changed everything in our life, right? I’m sure that’s a moment we were thinking about how the changes, the opportunities, the scary moment. I think we have to come to reckon that scientific discovery is one of the most exciting opportunities for AI and for health, right? How information can be synthesized, how information can be presented not only to clinicians but also to patients, and how patients can participate in that process from diagnosis to treatment is also— there is just so much we can do now.

Diagnosing Vertigo: A Personal Story

ANDREW HUBERMAN: (01:03:03 – 01:03:21): Yeah, I mean, AI— I won’t say AI is better than all doctors, but AI was able to disambiguate vertigo from low blood pressure for me a few months back. And one of the people who got it wrong is an ENT who works on the vestibular system.

DR. FEI-FEI LI: (01:03:21 – 01:03:24): What information did you provide? Just your subjective feelings?

ANDREW HUBERMAN: (01:03:24 – 01:03:58): My subjective experience over a day or two. Okay. Um, turns out it was a medication that a doctor had prescribed me that I had a, like, a mild but adverse event. And it’s a weird thing to step and feel like the whole world’s dropping down, right? And then kind of spinning. And I thought, oh my goodness, this, like, feels like vertigo. But I remember dizzy and lightheaded are different. So I started, like, looking into that. And then, and, um, sure enough, it was a, it was a blood pressure issue. It brought my blood pressure, excuse me, down too low. And, but I consulted, we know some smart doctors None of these were at Stanford. I will say that. This is the truth.

DR. FEI-FEI LI: (01:03:58 – 01:04:00): But it was just remarkable.

ANDREW HUBERMAN: (01:04:01 – 01:04:35): And when I ran it back to them, they were like, that’s really incredible. You know, had you not been on the phone with me and in my clinic, I would have been able to do some additional testing, to be fair. But this was zero cost. It took a morning to know if I drank some electrolytes at what I would have thought would be excessive level, that by 2 hours later, I would be fine. Now, of course, there’s the possibility of a placebo effect here, but 2 hours later I was fine. And so it’s also very consoling to the patient to have this. And so it’s not to say don’t go to a doctor, but it, it’s incredible. I mean, this exists now.

Her Father’s Liver Surgery and the da Vinci Robot

DR. FEI-FEI LI: (01:04:36 – 01:07:22): Yeah, the doctor can use this tooling. By the way, I have a very interesting example. You know that we have to reschedule this, uh, our conversation because my father was going through a surgery, right, at Stanford, uh, with an incredible surgeon. But the surgery was done by a robot, the da Vinci robot system, because it was a liver surgery and the surgeon, incredible surgeon, was driving the robot. So it was a deep human-machine collaboration.

After the surgery, I asked the surgeon, I said, do you imagine if, say you’ve done a million, which is impossible for a surgeon, but a human surgeon, but let’s collect all of human surgeons’ for this liver, this type of liver surgery data, can we possibly train an automatic AI to do this? The answer was not clear. So we went a little bit down a rabbit hole because liver is a very complicated organ. It’s extremely vascular. It has a lot of vessels and everybody’s liver is very different. So given the reality of how many patients undergo liver surgery per year, even if you aggregate the world’s liver patient surgeries, you might not have enough data to train these algorithms.

So this speaks of a very important fact that AI learns from patterns. When the patterns are not abundant, Then we have to be careful. We have to know how to use AI or how not to use AI. You know, in this case, having a human collaborating with the robot is way better than an underlearned robot doing the surgery by itself. But the same issue might be true for surgeons because how many surgeries a surgeon can get trained on? So these are opportunities that humans and AI can totally collaborate with and might reveal the best result right now. The future remains to be seen. Can we create an artificial simulation of a liver that we can now train infinite possibilities? These are all incredibly open scientific possibilities that is waiting ahead of us. But then there are situations like your situation where the vertigo versus low blood pressure probably have been reported so many times that in the database there’s enough of that that AI has learned that. So we can then now take advantage of that for people who don’t have immediate access to doctors. Amazing.

ANDREW HUBERMAN: (01:07:23 – 01:07:25): Is your father’s surgery went okay? It did.

DR. FEI-FEI LI: (01:07:25 – 01:07:38): It actually— happy to hear that. He lost 10x less blood. Than a typical surgery, thanks to the laparoscopic capability of a robot surgery.

Intuition, Motivation, and What Machines Still Can’t Access

ANDREW HUBERMAN: (01:07:39 – 01:10:07): I’d like to talk a little bit about some features that we think are uniquely human that may or may not be. You’ll tell me. These are genuine questions, not loaded questions. And then I’d also like to get educated on how AI is structured to allow these things to happen.

For instance, intuition. We all like to think of intuition as this mystical, very, like, it certainly is powerful, but this thing that we own that no one can take from us, that can’t be mimicked kind of thing. But I could also break intuition down to be, well, it’s my experience over time. It’s a data set coupled to some bodily and brain sensations and some prediction cues. Like, the last time I felt this, this happened. The last 2 times I felt that, things didn’t work out that way, so I’m going to go this way. I mean, could assign these rules to a computer.

But there are other aspects of our deeper self, if I can refer to them that way. Like, we don’t know where intuition is mapped in the body. Yeah, could do an imaging experiment, but you’re not going to collect all the neurons and hormones and everything simultaneously. So we don’t really have like a location or even a network to point to. Like, things like creativity, intuition, premonition, the idea that, you know, you really sense something is coming on, but it hasn’t happened yet. What sorts of rules can AI get that could give it these sorts of capabilities?

And here I want to talk about it in the context, if you will, of energy. So whatever this thing is, it’s like mitochondria driving cells more around one thing versus another, the same way fear or happiness would, right? We were just talking about energy. But within AI systems— and I’m not a computer scientist— within AI systems and GPUs, can we actually allocate more energetic flow through particular learning rules? So we could tell maybe someday, you know, based on everything you know about my sister, who I love, you know, what is your intuition about how our brother-sister relationship will evolve over time? And what is your sense about what would be great for us to do perhaps for our birthdays this year that’s different than before? Giving— and it only has access to the internet, can it actually become sort of mind-like or mind-body-like and come up with a sort of sense of what might actually be worthwhile? Or does it just need more and more prompts? Like, it’s just gonna keep asking me questions, so I’m actually doing the work.

DR. FEI-FEI LI: (01:10:07 – 01:12:27): Such an interesting question, Andrew. So, um, I do want to separate intuition from creativity for the sake of argument here, and maybe we’ll come back to merging.

So let’s talk about this intuition of giving my sibling love, what’s going to happen, right? Is it really intuition? So today when you go to an AI chatbot, you’re going to prompt, you know, I’m a Stanford professor and a neuroscientist. Give me this information. That is already called context. I don’t know if you call it intuition, but because you gave that piece of information, The AI’s answer for you is already gonna be different. If I type that I’m a 14-year-old teenager, you know, loving race cars, even if we ask the same question, it’ll have customized answer. That is a mathematical, I wouldn’t call it energy. I wanna be, that is just a mathematical fact of how these, these algorithms take this context and tailor the outputs. And it’s called context. It’s not that deep in computer science. That’s one type of intuition that is fairly shallow because you already are able to use language to describe it. Or you can say, I’ll upload an image that also is already expressible, and then AI The deeper intuition you just said is like, you don’t even know where they come from, right? Like, is it because I smell something? Is it hormones? Is it, you know, the mixture of mood? Is it my breakfast? That intuition, what would AI do with it? That is what I would say is inaccessible. There’s no sensory apparatus yet that can glean that data. Data and feed it to not only AI, cannot even feed it to, you know, for example, sometimes as a couple you might have moments that you’re just rubbing each other in the wrong way.

ANDREW HUBERMAN: (01:12:27 – 01:12:28): Never. No, I’m just kidding.

DR. FEI-FEI LI: (01:12:28 – 01:14:23): Yeah, of course. If you’re really familiar with each other, you kind of can sense it, but you can’t quite tell. Maybe you just leave quietly, leave that person alone. So that means whatever that intuition that person has they could not even express it in words or a gesture to give it to another person to use as a piece of information. So when you cannot even access that, neither a human, a different human, nor a machine can do anything about it because there’s no access to that. Highly individualized intuition. There’s no technology that can do that till you say we put brainwave collectors or, you know, skin conductance sensors. I mean, by the time we do those, maybe they become accessible.

So we have to recognize. So what I’m trying to say here is it’s not, what’s not very deep is the data accessibility. Accessible, either through language or through picture or through imaging or through brainwaves, whatever it is, it needs to be an accessible piece of information. If it’s accessible, then if we have collected enough of that, you can train machines with, or if a machine is well trained, it can, like you said, in a private way, forget about privacy, uh, uh, a breach, but in a private way, the machine can probably use it. What I’m trying to do, Andrew, here is not to make it sound mystical, but try to give it a scientific process to describe if it were to happen, how would that happen?

ANDREW HUBERMAN: (01:14:24 – 01:16:13): Yeah, because, um, pattern recognition based on big data sets and rules get us a long way is what I’m hearing. And earlier we were talking about where doctors fail and robots and machines perhaps do better, or they collaborate to do better than either one alone.

As a neuroscientist, you spend a lot of time looking at cells at some point in your career. And it’s amazing how the electrophysiologists for decades, if not longer, you develop an intuition. I’m not really a physiologist, but I learned to recognize cells based on, like, kind of these things that were not written up in any papers. But, like, if there was kind of a, like, a, like, a straighter edge along this thing and it had a certain shape and roundness, like, I tell you right now, that’s a transient off-alpha cell in the retina. Eventually, we, we developed genetic labels to reveal that that was true in every case. But then you also saw some that didn’t fit the rule. Machines can learn that, computers can learn that. And with all that information from all those papers, now we have a pretty good parts The retina. Cool. That works. And then you can apply rules like they fire this way, they fire that way. Okay, I’m good with all of that.

What I think I was trying to get to with intuition, and I probably didn’t give the best example, is like, what are some internal states of humans that are really hard to imagine machines could recapitulate, but perhaps they can? Like motivation. Do machines, do robots get motivated? We have rules of motivation. Like when I’m really motivated to do something, we call that urgency, a state of urgency, and I might move faster to do it. Less activation energy. You say, let’s go. I stand up a little bit faster. Machines could like go quicker in a certain direction, but can you say, hey, I want you to seek this out, but with a heightened level of urgency? Or are they just constrained by the mathematical rules they can work with?

DR. FEI-FEI LI: (01:16:13 – 01:19:19): So you could build this in the mathematics. So certain things Whether you call it motivation or in machine learning world we call them objective functions, you can build certain things into math. For example, now you go to say ChatGPT, it has different mode like think deeper mode or like give me a quick answer mode. If you don’t know how this works, you’re like, oh, this is interesting. One has more urgency that gives me a quicker answer. Answer. The other one has to go deeper into the search, right? And take longer to gimme the answer. So as a human, if you anthropomorphize it too much, you might call it urgency or motivation. But the truth is this is just a different kind of objective for the algorithm. You can say, well, The, the one that thinks quicker has a time limit or token limit. The one that thinks slower can activate a different part of the model that would take longer. So it becomes actually mathematically very dry and not that deep. But for a human, you can call that motivation or urgency.

But let’s go deeper because you are asking something deeper than, than that, right? Is that There are cognitive states that humans, you truly just, whether it’s motivation or urgency or fear or love, that is very hard to access and express. And do machines have it today? No. Let’s make it very clear. We tend to imagine that the machines feel or they’re not. They don’t have that data. They don’t have that mathematical objective function. So they can say, when the machine says, I’m sorry you’re so sick today, it’s very different from how your friend says it to you. Because the machine said that because it has learned through pattern. When someone tells it, I’m sick, you should say, I’m sorry you’re sick, instead of, I’m so glad you’re sick, because that data exists. Whereas your friend who hears that, they genuinely want your well-being. They love you. They want— they don’t want to see you suffer. They have that empathetic feel of, oh wow, if you’re in pain, I’ve experienced pain. So that’s— it’s not mirror neuron, but it’s at least a memory of what pain means. The machine doesn’t have any of that.

So we do need to make sure we differentiate that. So a lot of what drives humans what ticks human, what triggers human, doesn’t exist in today’s machine. We operate fundamentally different from today’s AI, and we have to recognize that, respect that. And this is where public communication is so important. We cannot confuse the public about this.

A Word From Our Sponsors

ANDREW HUBERMAN: (01:19:20 – 01:21:34): I’d like to take a quick break to acknowledge one of our sponsors, David. David makes protein bars unlike any Their newest bar, the Bronze Bar, has 20 grams of protein, only 150 calories, and 0 grams of sugar. I have to say, these are the best-tasting protein bars I’ve ever had, and I’ve tried a lot of protein bars over the years.

These new David bars have a marshmallow base and they’re covered in chocolate coating, and they’re absolutely incredible. I of course eat regular whole foods. I eat meat, chicken, fish, eggs, fruits, vegetables, etc. But I also make it a point to eat 1 or 2 David bars per day as a snack. Snack, which makes it easy to hit my protein goal of 1 gram of protein per pound of body weight. And that allows me to take in the protein I need without consuming excess calories.

I love all the David Bronze Bar flavors, including cookie dough, caramel chocolate, double chocolate, peanut butter chocolate. They all actually taste like candy bars. Again, they’re amazing, but again, they have no sugar and they have 20 grams of protein with just 150 calories. If you’d like to try David, you can go to davidprotein.com/huberman. Huberman. Right now, David is offering a deal where if you buy 4 cartons, you get the 5th carton for free. You can also find David on Amazon or in stores such as Target, Walmart, and Kroger. Again, to get the 5th carton for free, go to davidprotein.com/huberman.

Faces, Text, and the Question of Synthetic Presence

I feel like people assume there’s an emotion, a person, or whatever inside of the AI chatbot because we’re so language-oriented. It’s talking to us, it’s writing things to me, and we do that more now than we did 30 years ago. Yeah, certainly we’ve gotten very accustomed to receiving communications in fairly deprived language. Texts are not like extensive prose. Language has changed, modes of communication have changed— more deprived as opposed to more enriched. Yeah, but at some point soon, I’m guessing faces are going to start to enter the picture, uh, no pun intended. Um, like, how far off are we from— like, if you were I were to text the other person, oh, see you on campus for coffee next week at this time. How soon is it that that text is going to be actually a photo or video-like image of you just talking to me, telling me that? I mean, this would be trivial to do nowadays.

DR. FEI-FEI LI: (01:21:34 – 01:22:31): The technology is there, but we have to now look, zoom out a little bit and think about the social parameters, the legal implications. I mean, humans are capable of doing a lot of things with our tools. Rules, but we don’t do all of them. For example, today any car manufacturer can say every Friday the brake doesn’t work. This is a trivial technology. There’s a clock in the car’s computer and it just turns off the brake every Friday. But we don’t do that because it has deeply bad implications to our human society. That’s where rules comes in, laws come in, social norm comes in, morality comes And I think this is where we exit the pure technical discussion of AI and need to enter the social discussion of AI.

Governance, Norms, and the Multi-Stakeholder Challenge

ANDREW HUBERMAN: (01:22:31 – 01:24:25): Well, let’s do that, because one thing that I know about biologists or technologists is they like to go fast because it’s exciting. It’s the next edge, right? I remember long ago I had a friend who was studying viruses and ways of putting These weren’t infectious disease viruses. These were viral vectors for getting genes expressed as experimental tools in animals. But there came the opportunity to actually put the rabies virus, a modified rabies virus, into Drosophila, into fruit flies.

Oh my God. Now that’s fine and good, in my opinion, if you are absolutely certain, 100% certainty, that that is a nonfunctional version of the rabies virus, because you can put other cargo in there and do all sorts of important experiments on, believe it or not, disease and things like that. But if there’s just one fruit fly that somehow is an escaper and you get the actual rabies virus, there’s the potential it mates with another and then they eventually find the others. I don’t know if this would be a dominant or recessive situation, but now you have fruit flies with rabies and those things move really fast. So there’s a reason why you don’t do that experiment. Yeah, but it was exciting for them to think about And then they got denied, right? For good reason. Yes. I was grateful, right? Go to any biology department, you’re gonna see some fruit flies flying around. They love vinegar, by the way, you know, so they’re coming to your salad.

But the point here is that technologists love to go fast. They love sensing that next edge of things. So how is it that between government, the general public, technologists— and now I’m just leaving out biology here, and medicine. How is it that that conversation can occur in a way that’s going to satisfy each of those groups enough, not hold us back? Because we’re also supposedly in an AI race right now, so that warrants going faster, not slower. How do you think about this?

DR. FEI-FEI LI: (01:24:26 – 01:27:29): I mean, Andrew, this is why I returned from Google 8 years ago back to Stanford and started the Human-Centered AI Institute. These are profound societal questions we had to face. And back in 2018, there was no ChatGPT. But as an AI scientist, I knew that this is only going to accelerate. This is why I went to my colleagues and university leadership and said, let’s put a framework. But it’s not just my framework or Stanford’s framework.

The entire society in every way needs to wake up to the social implication, as we have done this in human history, whether it was cars or airplanes or biotech, is that it’s multidimensional with multi-stakeholders, right? There is the professional norm. For example, you guys as biologists don’t sneak into the lab and try to put rabies into Drosophila or fruit flies because that’s a professional norm and you’re ethical training. There is industry rules, for example, IRBs. Every human subject experiment today on university campuses are subject to the IRB regulatory framework so that we can look at this. And then there are laws and regulatory laws depending on if it’s applied to humans versus crops or, you know, so AI has to go through the same, right?

We need to have our professional norms. We need to have education. Computer scientists are not educated in ethics and societal studies. You know, they’re starting to. I mean, this is why a number of universities, including Stanford, are feverishly putting that part of curriculum into our education. Now, those are the norms and education, but we also should work with the government and different kinds of governments and society have different kinds of norms and traditions and heritage and look at where the regulatory measure should apply. AI, for example, crossing biology, FDA. I think that’s a very important area to look at how AI should be used to help but also guardrail so that we can avoid harm. What I would not like to see is one person or a few people coming from industry and telling everybody what to do. I think that would be dangerous because market forces are different from societal norms and culture and heritage are different from education and ethics. And, and these are multi-stakeholder problems to solve together.

Calculators, Smartphones, and the Fear of Losing the Next Generation

ANDREW HUBERMAN: (01:27:29 – 01:29:36): I love that answer, and it’s something that’s very, very timely right now. Um, this aspect of our conversation is surely going to expand over time, but you bullseyed it. I’d like to get your thoughts on how the human brain is being shaped on machines and how machines are being shaped our understanding of the human brain.

So first question first, many people, parents and kids, are thinking, oh, like, my kid is never going to learn anything now. They’re just going to look everything up on a chatbot. But if you look back in the history of learning, similar arguments were made about calculators and computers and the typewriter and on and on. However, it is an interesting question that this hardware that we have in our heads evolved to process physical things in the world— light, sound, it smells, And then it got this really cool piece up front, the prefrontal cortex, that can learn learning rules and can update those learning rules. So like, if anything, we were gifted with a learning-to-learn machine and updating learning. So that’s how kids can adjust and use LLMs.

So I, as a generation that grew up with the personal computer showed up— granted, I grew up in Palo Alto. It was like, here’s Pong and there’s the Apple IIe. And like, we had And I think, oh cool, like the brain can mature around technology, collaborate with technology in a way that I think my life has been greatly enriched by it. But I think the smartphone and perhaps the camera smartphone combination, as people like Jonathan Haidt have pointed out, have created a situation where most people, like they love these technologies for the ease and convenience, but we’re all a little bit more aware now or a lot more aware that we’re giving up something too. Yeah. And that there are traps that people, in particular young people, can fall down. Yeah. So what is the very optimistic, meh, and very pessimistic view in your, in your mind, if 3, if 3 flavors actually exist there, of how young brains can be enriched, are unaffected, or can be harmed by AI as it exists now?

The Danger of Taking Away Agency — And the Danger of Denying the Tools

DR. FEI-FEI LI: (01:29:36 – 01:33:09): Let’s just kind of stay with what we Great question, Andrew, and the answer almost falls out of our previous conversations because you used the word motivation and I was using the word agency.

The absolute bad outcome is that our young generation, their agency and human-level motivation of learning and living is taken away by tools. So doom scrolling, passive watching of Shorts, all this are not helping agency, human agency. Learning fundamentally, respecting the hardware you’re talking about takes time, takes effort, sometimes takes some pain. That is just how our brain is. It doesn’t matter how transistors move. Move. Our neurons move in certain ways. Our chemistry, our hormones move in certain ways. So for young generation, no matter how the society will be different, jobs will be different, our human body needs to go through a deeply developmental phase where learning needs to happen. And that agency of learning, that motivation of learning cannot be taken away by anybody, should not be taken away by humans, nor should it be taken away by machines. That would be my concern, which is that if AI is not used right, the agency and motivation is taken away. Then we are left with generations or generations to come who have not properly developed the brain.

The other kind of danger is in the name of agency and motivation, the tools are denied. To our students because we’re worried you cheat. We’re worried you only got your answer from ChatGPT. That is very bad as well, because with the proper agency, proper motivation, proper ways of using this tool, we can go a lot deeper with AI than we have ever learned. I was just thinking about— I was a pre-med student for a while. Man, organic chemistry was hard. You know, I remember remembered trying to learn the molecules, their orientations. But the TA hours are too short, or it overlaps with my other class, and my professors only have a certain number of office hours. It was just a struggle to learn that, right? If today I were to have an AI companion, I would ask so many questions about organic chemistry because I know where I’m stuck, right? I have the motivation to learn. I needed guidance, that would be such a powerful tool for me to learn. So that we should not deny students from. So both things worry me, is either denying the tool or taking away agency and motivation.

Of course, the flip side is, is great, is let’s find a way to keep our children and students’ motivation and agency. Let’s find a way to give them the access and the right way of using these tools, then this generation, this coming generation, and many generations to come will be way smarter than us because they are superpowered.

ANDREW HUBERMAN: I love that answer.

ANDREW HUBERMAN: (01:33:09 – 01:33:14): Um, I have great faith in neuroplasticity and the younger generations too.

DR. FEI-FEI LI: (01:33:14 – 01:33:18): Even our own. I know we’re old, but not so old.

ANDREW HUBERMAN: (01:33:18 – 01:33:21): Let’s give ourselves some credit. Plasticity does exist throughout the life cycle.

DR. FEI-FEI LI: (01:33:21 – 01:33:28): Even our own Neuroplasticity, right? Like, I, I find AI a great tool for my learning.

ANDREW HUBERMAN: (01:33:29 – 01:33:45): I mean, for me, it’s been a remarkable discovery of what it can do. Yeah. But I, I tend to approach it from the position of consumer if I know nothing about something, and from the position of creator if I have some, uh, knowledge set inside of whatever it is I’m asking.

DR. FEI-FEI LI: (01:33:46 – 01:34:17): Well, I actually have another thing. A Stanford undergrad taught me something last year. And I realized before ChatGPT, sometimes I get lazy. I— if I have a question, I ask the person I think is smart next to me. Now I realize I should not ask lazy questions because it’s so much easier to get information before you spend somebody else’s time to ask something that’s, that’s, that’s too lazy. And AI is forcing me not to be too easy.

Prompting as a Skill, and the Socratic Method

ANDREW HUBERMAN: (01:34:17 – 01:34:25): How essential is the specificity of the prompt to getting the best information out of AI? Prompting is very important.

DR. FEI-FEI LI: (01:34:25 – 01:34:41): And that’s a skill, right? That is a skill. This is why public education is so important. This is why education is so important. I would love to see our schools, K-12, teaching prompting. Here’s a quiz. Who is humanity’s best prompter?

ANDREW HUBERMAN: (01:34:43 – 01:34:44): I’m gonna flunk this quiz.

DR. FEI-FEI LI: (01:34:45 – 01:35:02): Socrates, if he were alive, because that is the method of prompting, right? Think about it. What is Socrates’ method? It’s prompting and seeking truth by asking questions. And we should go back and teaching kids that.

Embodied AI, Speech, and Dr. Eddie Chang’s Work

ANDREW HUBERMAN: (01:35:02 – 01:36:46): And taking a walk while you have those discussions. Yes. Which actually is a good transition perhaps to this notion of embodied AI. You know, it’s a world apart to attach a face speaking to hearing words.

Um, my good childhood friend, um, who I hope you’ll meet soon because you both would benefit from the conversation so much, and I just want to be a fly on the wall, um, Dr. Eddie Chang, chair of neurosurgery, bioengineer, and he studies speech and language. He and others have figured out the transformation of neural activity to control of the larynx and pharynx, and he’s brought people essentially out of locked-in syndrome so they can speak. For the first time in 10 years, he has this patient who was sadly paralyzed, and he could speak through a computer. He has others, many examples of these, in fact.

But the incredible thing is, when he started putting an iPad next to this person, who is one woman in particular who’s wheelchair-bound, they had a video of her at her wedding, so they knew her voice, they knew her emotive patterns, they knew a bit about how she moved her body as well. And she now speaks through an iPad next to her frozen real face, but she can interact with the world and it can interact with her in a completely different level of depth than if it were just a microphone, the sort of Stephen Hawking thing. And it’s constantly being updated through machine learning. What— and now also paying attention to the people she’s speaking to and their responses. I mean, this is, this is embodiment. Yes, it’s on a 2D flat 2D screen, admittedly, but this is like an exponential leap over just robot sound or even accurate sound alone.

DR. FEI-FEI LI: (01:36:46 – 01:37:19): It’s not just embodiment of people, it’s embodiment also embodied AI goes into robotics, right? The next frontier of AI, as I have been saying, is beyond language because again, humans develop first preverbal, Firstly, evolution took, you know, 500 million years without verbal communication. And also the world would, in the right version, would be a lot better place with robots helping humans.

ANDREW HUBERMAN: (01:37:19 – 01:37:42): Could you give me some examples? I love this idea, but again, I realize I’m probably a little too deep into the technology rabbit hole and it’s probably scaring some people. We’ve got self-driving cars. Actually, the Waymo always stops for me and my puppy, my beautiful little six-month-old puppy. How could you not stop when he wants to cross the street? A lot of people won’t stop; they’ll almost run us over in the morning.

DR. FEI-FEI LI: (01:37:43 – 01:37:46): The Waymo is very respectful. Waymo has to learn the rules, right?

ANDREW HUBERMAN: (01:37:46 – 01:37:59): Exactly, exactly. So there’s benevolence there that doesn’t always exist in humans. But where do you think this is going to show up first? And what’s it going to look like? Like if we zoom out 12 months from now?

Robots in Everyday Life: Caregiving, Safety, and the Shape of the Future

DR. FEI-FEI LI: (01:38:00 – 01:40:44): 12 months is a little bit too fast for robotics. 2 years, 3 years? I would say if we zoom out 30 years. 30 years. Okay. I’m not saying that’s the first time robots hit the street. We already have robotic cars. I’m just saying it, it takes longer for especially a hardware also involved technology to manifest.

But I would say hopefully in you and my lifetime, I would love to see robots being part of our society, helping us. For example, I’m a single grown-up child taking care of 2 very advanced-aged and very sick parents. And they happen not to speak English either. The amount of work I do is incredible, right? So I would love to have help. It doesn’t take away family’s responsibility. It doesn’t take away love. It doesn’t take away the necessary communication. But the physical labor would really— certain part, I would love to get help.

We live in the state of California. What is the one thing we all experience? Traffic. Traffic, high taxes. Certain part of California doesn’t have traffic, but wildfires. Oh wow, yes, right. Who is fighting these wildfires? Putting humans in danger of rescue, natural disaster, is not a great idea, right? So my family and my parents happen to have, have enough means, but I was just thinking, an elderly living How do they go get grocery? How do they go get medicine? Now, there might be some shipping we are starting to see, but what if they want to go, you know, um, for a walk or want to go to a park? So there are just so many things that— oh, by the way, you’re in the School of Medicine. We don’t have access of caretakers. We have a shortage of caretakers. Our nurses are deeply fatigued and overworked. I was literally in the hospital with my dad for the past month and just watching the amount of work nurses do. We know that on a given shift, nurses walk miles to fetch things, get medicine. There’s just so many— can you imagine robots helping, right? Like, so there are just So many ways that our society can be structured and can benefit from help.

ANDREW HUBERMAN: (01:40:45 – 01:41:45): Oh, I love these examples. So many spring to mind based on what you described. Crossing guards. Yeah, you imagine with video that somebody who’s homebound because of age or illness could navigate to the store and pick things off the shelf. It doesn’t have to be so disconnected that they just program and it comes back. Although that could be an option too. Too. Yeah, I think that we have to revise our notions of what this picture looks like, because I think there are a couple things about robots and computers that scare people. Um, one is that is their physical hardness, right? And so the way we share space with them is very different than the way we share space with other things. Of course, I’m not thinking, oh, like you cuddle with a, with a robot, although some people might think that. That’s not my mindset. But I am thinking like, okay, If I had a robot that could fold clothes, vacuum, water the plants, and feed my fish. Although I like to feed my fish myself. I really enjoy it. I love seeing them eat. I love being tactilely, you know, literally in touch with them. They’ll eat from my hands.

DR. FEI-FEI LI: (01:41:45 – 01:41:47): Does your puppy like your fish? He does.

ANDREW HUBERMAN: (01:41:47 – 01:41:51): He has his own fish tank. I just got him some tropical fish right in front of his little thing.

DR. FEI-FEI LI: (01:41:51 – 01:41:52): He’s taking care of them?

ANDREW HUBERMAN: (01:41:52 – 01:42:27): Well, he looks at them. He’s not equipped to take care of them yet. I don’t think it— unfortunately, there’s not enough prefrontal cortex in him. So, and he’s a, he’s a kind, but he’s a bulldog mutt. They’re not the smartest breed. They only have a few learning rules, but they’re very kind. Um, but if you want a dog that can take care of a fish tank, you probably need like a West Highland Terrier or something like that. Different, more prefrontal cortex. But the idea here is if one robot is doing one thing and another robot is doing another, it feels like a lot of hardware in my life. And I think that’s kind of how people feel. Right. But you could imagine a multimorphic robot.

DR. FEI-FEI LI: (01:42:27 – 01:42:47): Do you know Baymax? I don’t. Disney’s robot, probably 10 years ago, 15 years ago. It’s a— Google the image. This is the white medicine robot, healthcare robot that is very not fluffy. It’s very spongy. Like it’s, it’s, it feels like a big balloon.

ANDREW HUBERMAN: (01:42:47 – 01:43:00): Yeah. So you might like that. Yeah. More, more contours. Yeah. And more multitasking from the same robot. Robot feels like a world that I could adjust to more quickly than the idea of my world filled with robots.

DR. FEI-FEI LI: (01:43:00 – 01:43:53): Yes. Again, Andrew, I think as we imagine the future and we talk about how we imagine the future, I keep coming back to the word agency. Humanity should have the agency to decide how we imagine this. It cannot just be a company or or I don’t know, an investor decide that the, the world should be filled with metal-like robots, right? Like, our society should be collectively proactively imagining. And, and one thing I worry in this AI rhetoric is that the public is put in a position of being reactive when it feels some people are just deciding And the multi-stakeholders are not participating in this, designing the future together.

ANDREW HUBERMAN: (01:43:54 – 01:44:54): Like with your example of your father’s surgery, to cross the robot with the physician, right? If we cross a problem where there’s a vulnerability with a robot that clearly makes things better, the picture changes in the right direction. So I’m thinking of a few examples off the top of my head. Um, I think most people would agree that if their kids could walk themselves to school and home, it would be great. But you worry about safety. But if a robot was really a good guardian of your kid to the point where they could alert the authorities or maybe even protect, physically protect your child, that would be awesome. Give them more agency in the world. You think about, um, some of the darker but nonetheless unfortunately real predatory behavior online. Parents can only oversee their kids’ behavior so much. Kids are only aware of so much that’s But you could imagine kind of an avatar in there with you that’s really advocating for you, that can spot things and keep predators at bay.

DR. FEI-FEI LI: (01:44:54 – 01:44:56): Here you go. That’s a great startup idea.

Steve Jobs and the Search for AI’s Humanizer

ANDREW HUBERMAN: (01:44:56 – 01:46:51): Like, that would be cool. But here’s what’s missing, I think, from the picture for me. I remember seeing this incredible guy. I know people, some say he was kind of prickly, but this incredible guy walking around downtown Palo Alto when I was a postdoc and when I was a professor. I was a kid growing up working at the Palo Alto Toy and Sport World, and that was Steve Jobs.

No shoes, kind of looked like a hippie. Yes, he shouted at people at work, and, you know, probably HR wouldn’t look too kindly upon him nowadays, but he understood that these things we call computers needed to have rounded edges. Yes. They needed to fit kind of seamlessly in our pocket. They needed to have Bob Dylan on the landing page or whatever so that it softened the relationship to technology. Some people would say, well, it went too far. It was a Trojan horse, I don’t think so. Somebody who really understands human nature to allow these, like, what are clearly going to be benevolent collaborations between robots and humans to happen. Because as you pointed out, and with total respect to the technologists that have built AI and the scientists that do amazing science, there’s a hardness to either the way they’re being presented or what they’re capable of sharing that is a real separation. Yes. And I’m not a therapist, but if I could, like, wrap my arms around them, I’d be like, listen, guys, you’re the smartest people in the room. Guys and gals, to be fair. You’re the smartest people in the room. But people don’t like you because they don’t understand you. And maybe you need a collaborator to help you share your vision in a way that isn’t going to allow the press— because the media is guilty of building this chasm because it’s like these technologists, they’re coming for us. I think that’s— I think that’s a total trick of media too. That’s just to put money in their pocket. Like, there’s a lot going on right now. So who’s the Steve Jobs or the Stacy— Stacy whoever it is? I mean, could be a man, could be a woman, someone who really understands human nature.

DR. FEI-FEI LI: (01:46:51 – 01:46:57): There are many of them. There are many of us, you know. I mean, Stanford started Human-Centered AI Institute.

ANDREW HUBERMAN: (01:46:58 – 01:47:00): Well, there’s you. There’s you. Okay, but there are many.

DR. FEI-FEI LI: (01:47:01 – 01:48:58): Yeah, there are plenty of entrepreneurs who are doing incredible uh, startups on AI for drug discovery, AI for healthcare, AI for aging, AI for mental health. These people care about AI, right? There are many designers and product managers who are trying to— I do think the megaphone is too much focused on people pumping their chest and talking about tech in a certain So, you know, even this podcast is making a positive difference, I hope. I would hope so. To put that human angle, the rounded human angle, human perspective, human future into these conversations.

I don’t feel despair, Andrew. I’m an educator, I’m a builder, I’m a technologist. I see many people around me, including my entire staff, startup. They, they— these brilliant young technologists could join any startup or company they want, but they come to World Labs because they want to empower people, right? So I see many people, but I don’t think there’s enough. You’re right. I don’t think the, the public discourse is, um, is balanced right now, and there is too much extreme rhetoric either in terms of extreme doomerism and lack of safety, like it’s just freaking people out, or extreme utopian as if technology can do no wrong. And then that’s disingenuous. People would say, well, okay, you’re the haves, of course you say that. So I think we should come to the middle and talk about what this technology is, how to use it, how we can collectively have that agency to guide the future.

ANDREW HUBERMAN: (01:48:59 – 01:50:05): Yeah, one thing that was pointed out to me by one of my podcast colleagues that should have been obvious but wasn’t, and clearly this is something that you, for lack of a better word, you embody among many other things, is people don’t really want to hear stories about machines, but people love hearing that some person cured their dog’s cancer or their Their child that was experiencing crazy symptoms. They had no clue. The doctors had no clue. And their fingertips, AI solved the problem. These are the stories that really need amplification because I think that we can relate to them. And they’re beautiful stories. They’re incredible stories. But they’re not getting nearly as much attention as the other stuff. And that’s a challenge. Yeah, you know, traditional media doesn’t really care about the long arc of things. They are on a, like, a 12 to 24 hour cycle. But other names perhaps of, like, people who are really trying to, like, talk about the benevolent use of AI, these collaborations that, you know, we should be aware of?

DR. FEI-FEI LI: (01:50:05 – 01:50:11): Stanford HAI’s newsletter, our website, our seminars. We promote a lot of those works.

World Labs: Spatial Intelligence as AI’s Next Frontier

ANDREW HUBERMAN: (01:50:12 – 01:50:19): I would love to learn more about your startup? Um, because you don’t pick projects haphazardly. So what is the— what is the project?

DR. FEI-FEI LI: (01:50:19 – 01:52:36): What’s the goal? So my startup, uh, co-founded with, um, a couple of other co-founders, is called, uh, World Labs. We co-founded it at the beginning of, uh, 2024. It really is, for, for me, a kind of my life’s work. You know, we both come from vision and The recognition of there’s more beyond language intelligence is what really motivated me to think hard about what’s the next chapter of AI frontier.

And we recognize that unlocking spatial and physical intelligence is really the next chapter, that it’s not excluding languages. Of course, the language technology is incredible, is where we can devote more time to build models or build eventually products that can help unlocking capabilities in spatial intelligence, like generating 3D, 4D worlds that are deeply useful for creators, for robot training, for architecture design. Design or to enable those interactive environments, whether you’re talking about healthcare usage or education usage or robotics usage or industry usage, these capabilities go beyond language per se.

And so World Labs was founded based on that premise. We are still a young company. We’re very much a a model-focused company where we’re building this foundation model and we’re started by a lot of PhDs. But now we’re starting to build products. And so it’s still the beginning. It’s very exciting. And as a technologist, I feel deep in my heart I’m a builder. You know, it’s maybe it’s because also I’m an immigrant. So that, that rolling your sleeves up and just get in with the young generation that is so incredibly smart and just build something from scratch is just so exciting.

Mapping the Real World and the Imagined One

ANDREW HUBERMAN: (01:52:37 – 01:53:07): I recall a time not but what, 15, 20 years ago when there were cars driving around, oh yeah, taking images. Still, still driving, still driving around taking images. But I imagine that there— and there are certainly aerial views as well, but you can imagine little tiny drones like the type that could fly through a on and just kind of look at everything, or, um, so to speak, or drones picking up information about every nook and cranny of the fjords in Norway. Has that been done to sort of map the 3-dimensional world?

DR. FEI-FEI LI: (01:53:07 – 01:54:13): First of all, let’s not make it sound scary that drones are getting into people’s homes and properties. I think that the ability to capture imageries of the world is really rapidly advanced, right? Like, our cell phones are incredible sensors. They’re Not drones, but people take a lot of photos. And of course our camera technology has improved.

What World Labs is doing is not just taking real-world images. It’s, we allow people to imagine what’s in their mind’s eye. As long as you can type a sentence or show a picture or a sketch of what you imagine, we try to turn that into worlds and environments. Why is it useful? Because entertainment industry would use it. Design industry would use it. Robotics industry very much would use it for training environments and, and, and all that. So the combination of capturing what’s in the real world as well as capturing what’s in your imagined world is the new frontier.

ANDREW HUBERMAN: (01:54:14 – 01:54:48): If you don’t mind, I’d like to just take a couple of more minutes and talk about this moving from imagination to something, because this is Los Angeles. Angeles, it occurred to me that a lot of people write scripts and then they try and get them their movie made. But with AI, in theory, you could take a script and give it to AI and it could make the movie, in theory, right? Going from words to pictures to video. Um, and you could maybe edit it a little bit here and there where it needed help, of course. Has that been done? Has a successful movie been made start to finish using AI?

AI and the Future of Filmmaking

DR. FEI-FEI LI: (01:54:49 – 01:57:52): So this is a very nuanced topic. This is where we also get into people’s wariness of AI and creativity, when if not careful, it might sound like we’re taking away from storytellers and creators’ job, right? So, so let’s separate this job conversation from the technology conversation a little bit, even though they’re entangled.

Technology has advanced enough that taking scripts and generating shots, video shots, is getting really good. We have seen short movies, even almost feature-length films being assembled by AI tools. We have. And there are many companies, US companies, Asian companies creating technology. But what remains deeply human, and that is important, is every part of storytelling and story creation, there are humans behind it with their unique emotion, story, technique, how they see the world, how they move the cameras, how they characterize people, characters. A lot of that that is what Hollywood and novel writers is about.

So how do we meet the human need and human desire of storytelling with modern tools is actually a challenge because there is a fear very much coming from Hollywood that AI is taking over, and storytellers and actors and screenwriters, the jobs are being impacted. And I think it is. But how is it being impacted? What are we doing about it? Who is working in a, in a constructive way? You know, this is not my industry per se, but I would love to see much more nuanced work in this, and also nuanced public discussion about that.

But I do think just like healthcare, we were talking about how AI can rapidly change and disrupt the old ways of doing healthcare. I think AI is absolutely changing the way we’re doing storytelling. So one story, speaking of which, I have a co-founder whose name is Ben, and Ben and I met with Ben Affleck. So I was joking, Ben meeting Ben, who is also thinking very avant-garde about using AI tools about filmmaking, right? So having conversations between technologists and storytellers or movie makers at this moment is critical. Yeah.

ANDREW HUBERMAN: (01:57:52 – 01:58:33): I feel like in every example of technology, there’s some crossover point. Yes. When somebody who’s truly an insider embraces a technology and then it just kind of takes off, like, yeah, you know, uh, Steven Spielberg or something like that, or these probably aren’t the best examples, but like the, the Steve Jobs-Wozniak crossover, kind of a designer, technology-curious guy, and a real— forgive me to the Jobs family— but a real computer scientist, right? That It’s these collaborations are really key. Like, so you need an insider and an outsider to do it right because you have to understand both cultures. Yeah. And how to include the industry, the, the people.

DR. FEI-FEI LI: (01:58:33 – 01:58:59): So I really hope, right, because World Labs works with VFX industry as well, it’s so important for me that our customers and users feel empowered. It’s not that technology should be taking their jobs away. Technology should be making their jobs better, superpowering their creativity. And that’s how I, I see this technology, and that’s how I would like to work with the users and customers.

ANDREW HUBERMAN: (01:58:59 – 01:59:26): It’s wild to think that, you know, when I was a kid on California Avenue in Palo Alto, there was this store, Keeble & Shuchat, and it was just a photograph store and camera store. Yes. You go in there, you get your film developed, and there are all these guys behind the counter, and they tell you all this. You could rent a long-distance lens and this kind of thing. Um, none of that exists anymore. Everything went digital, you know, but there are still camera stores. So industries can morph. They don’t always get obliterated.

DR. FEI-FEI LI: (01:59:26 – 01:59:49): Yeah, it morphs. People also get reskilled, upskilled. You know, we are working with a lot of creators who are using AI tools because they see where technology is going and they want to reskill and upskill themselves. So I think moments of change is moment of both opportunity and loss. Mm-hmm. We need to really be thoughtful about that.

ANDREW HUBERMAN: (01:59:50 – 01:59:58): My last question is about the young generation. How do they feel about AI? Because there is this—

DR. FEI-FEI LI: (01:59:58 – 01:59:59): How young are you talking?

ANDREW HUBERMAN: (01:59:59 – 02:00:04): I’m talking about kids between the age of, uh, 7 and 20. Okay.

DR. FEI-FEI LI: (02:00:05 – 02:00:07): That, that’s, that’s literally my kids. Yeah.

ANDREW HUBERMAN: (02:00:07 – 02:00:40): So I might’ve asked that question for a reason. You know, how do they feel about it? Are they excited by it? Because there is this phenomenon where, like, computers come along and, you know, your handwriting teacher is getting nervous that people aren’t just typing now. They’re all writing with their fingertips and no one’s going to know how to write. And we wrote for— there’s these stories have been around for a long time about how we’re just going to dissolve into a puddle of our own neurons if we don’t embrace the past as much as the future. I like to think some of both is what’s important. But how do the kids feel? What do they think?

Teachers, Parents, and Kids: Fei-Fei’s Hope for the Next Generation

DR. FEI-FEI LI: (02:00:40 – 02:04:07): This is actually my pet project as an educator and technologist. Everywhere I go, I try to talk to, to students, parents, and teachers because I think that is the most forgotten population. Our policymakers and our technologists and our investors, they don’t talk well, teachers, parents, and students. They all have opinions and they all have kids, but they don’t talk about it.

I always have hope for kids, maybe because I’m an educator, because I think the biggest thing humanity never learns is the older generation lamenting about the future generation, as if the future generation doesn’t know anything, they’re rude, they’re, they’re forgetting the past. But if you look at arc of history of humanity, by and large, we advance for the better. Now, I’m not denying the atrocities. I’m not denying the setbacks. I’m not denying this. But, you know, humanity, they’re fundamentally— I’m an optimist in humanity, right? So, so that’s where I come from. So if you’re a total pessimist, maybe we’re already on the wrong footing.

But I look at kids They’re curious. That’s why they’re kids. They’re curious. They, of course, they get massively entertained by this technology, but they also are starting to use it. What I worry about are teachers and some parents, because I think our society today, and especially Silicon Valley, are not doing them a service. We’re forgetting about them. We are lecturing. Them. We are berating them. We are looking down at them. They are the most important people in our society. We should be talking to them. We should be uplifting them. We should be supporting them. We should be providing resources to them. K-12 teacher, or K-16 teachers, they share the most important critical burden of our society.

I’ll tell you a real story. November 2022, ChatGPT came out. Obviously, I’m an insider in terms of technology, but the first thing I did was emailing the principal of the elementary school my kid was in and said, I would like to come and guest lecture for your students and teachers. It’s not because I’m so special, it’s because I want them in real time to know what’s happening. Because nobody, nobody in Silicon Valley, no investors multibillion-dollar investment firms or multimillion-dollar, multitrillion-dollar companies, when ChatGPT came out, the first thing is, what about our teachers in the neighborhood? Nobody thinks like that. But we need to, we need to be talking to teachers. We need to show teachers. Of course they’re going to ask the question about what if kids cheat? It’s okay they ask those questions. Let’s just show them. Let’s work with them. And empower them to come up with ways to deal with that. They are smart too. They are eager to change. They’re just forgotten.

So I have hope for kids, but in order not to have a blind hope, I think we should all remember our teachers and help our teachers and parents so that we can help our kids.

ANDREW HUBERMAN: (02:04:08 – 02:04:23): I absolutely love that answer, and I And I, that sentiment is shared by many, many people listening. Um, God bless the teachers and they need help, support, and information because now they, they turn on not yours, but most podcasts.

DR. FEI-FEI LI: (02:04:23 – 02:04:40): They’re just scared. They’re so scared. They hear this doomerism. They hear the doomsday, or they say, oh, don’t worry, it’s utopian. Neither of these messages can help our teachers. If they’re not helped, our kids are not helped.

Closing Thoughts

ANDREW HUBERMAN: (02:04:40 – 02:05:31): Couldn’t agree more. Couldn’t agree more. Fei-Fei, thank you so much for taking the time out of your incredibly busy schedule. I’m so glad to hear your father’s okay. And that is also part of your schedule, taking care of your parents, kids, and all the rest, to come educate us on this thing that’s not just important, it’s a major wedge of where we’re at and where we’re headed. And I share great optimism with caution, even more so on the basis of what you shared today. And also, thank you for teaching us more neuroscience as we went along, because these machines are informed by the brain, and the brain is informed by these machines. And this is the world we’re living in. And I have great optimism, in no small part thanks to the fact that you exist in this world. And thank you for taking the time to come here to share. I know many people are very grateful, so thank you.

DR. FEI-FEI LI: (02:05:32 – 02:05:37): Thank you, Andrew, and I really appreciated this conversation. It’s a civilizational moment.

Outro and Podcast Information

ANDREW HUBERMAN: (02:05:37 – 02:08:04): Thank you for joining me for today’s discussion with Dr. Fei-Fei Li. To learn more about her work, please see the links in the show note caption. If you’re learning from and/or enjoying this podcast, please subscribe to our YouTube channel. That’s a terrific zero-cost way to support us.

In addition, please follow the podcast by clicking the follow button on both Spotify and And on both Spotify and Apple, you can leave us up to a 5-star review, and you can now leave us comments at both Spotify and Apple. Please also check out the sponsors mentioned at the beginning and throughout today’s episode. That’s the best way to support this podcast.

If you have questions for me or comments about the podcast or guests or topics that you’d like me to consider for the Huberman Lab podcast, please put those in the comment section on YouTube. I do read all the comments.

For those of you that haven’t heard, I have a new book coming out. It’s my very first book. It’s entitled Protocols: An Operating Manual for the Human Body. This is a book that I’ve been working on for more than 5 years, and that’s based on more than 30 years of research and experience. And it covers protocols for everything from sleep to exercise to stress control, protocols related to focus and motivation. And of course, I provide the scientific substantiation for the protocols that are included. The book is now available by presale. At protocolsbook.com. There you can find links to various vendors. You can pick the one that you like best. Again, the book is called Protocols: An Operating Manual for the Human Body.

And if you’re not already following me on social media, I am Huberman Lab on all social media platforms. So that’s Instagram, X, Threads, Facebook, and LinkedIn. And on all those platforms, I discuss science and science-related tools, some of which overlaps with the content of the Huberman Lab podcast. Podcast, but much of which is distinct from the information on the Huberman Lab podcast. Again, it’s Huberman Lab on all social media platforms.

And if you haven’t already subscribed to our Neural Network newsletter, the Neural Network newsletter is a zero-cost monthly newsletter that includes podcast summaries as well as what we call protocols in the form of 1 to 3 page PDFs that cover everything from how to optimize your sleep, how to optimize dopamine, deliberate cold exposure. We have a foundational fitness protocol that covers cardiovascular training and resistance training. All of that is available completely zero cost. You simply go to hubermanlab.com, go to the menu tab in the top right corner, scroll down to newsletter, and enter your email. And I should emphasize that we do not share your email with anybody.

Thank you once again for joining me for today’s discussion with Dr. Fei-Fei Li. And last but certainly not least, thank you for your interest in science.

Related Posts

• Transcript: Jeff Dean Interview on The 1% Rule for Building in AI with Diana Hu
• Transcript of Kim Dotcom Interview on Surveillance: Tucker Carlson Show
• Full Transcript: Zanny Minton Beddoes Interviews Elon Musk
• Transcript of Daniel Kokotajlo Interview: Diary Of A CEO Podcast
• Transcript: Why a Social Media Ban Won’t Save Teens – Candice Odgers