Manolis Kellis: Human Genome and Evolutionary Dynamics | Lex Fridman Podcast #113

The following is a conversation with Manolis Kelis. He's a professor at MIT and head of the MIT Computational Biology Group. He is interested in understanding the human genome from a computational, evolutionary, biological and other cross disciplinary perspectives. He has more big impactful papers and awards than I can list, but most importantly, he's a kind, curious, brilliant human being and just someone I really enjoy talking to. His passion for science and life in general is contagious. The hours honestly flew by and I'm sure we'll talk again on this podcast soon. Quick summary of the ads 3 sponsors blinkist 8 sleep and masterclass please consider supporting this podcast by going to blinkist.com lex8sleep.com lex and signing up at masterclass.com lex click the links, buy the stuff, get the discount. It's the best way to support this podcast. If you enjoy this thing, subscribe on YouTube, review it with five stars on Apple Podcast, support it on Patreon, or connect with me on Twitter lexfriedman. As usual, I'll do a few minutes of ads now and never any ads in the middle that can break the flow of the conversation. This episode is supported by Blinkist, my favorite app for learning new things. Get it@blinkist.com lex for a seven day free trial and 25% off afterwards. Blinkist takes the key ideas from thousands of nonfiction books and condenses them down into just 15 minutes that you can read or listen to. I'm a big believer in reading at least an hour every day. As part of that, I use Blinkist every day to try out a book I may otherwise never have a chance to read. And in general, it's a great way to broaden your view of the idea landscape out there and find books that you may want to read more deeply. With Blinkist, you get unlimited access to read or listen to a massive library of condensed nonfiction books. Go to blinkist.com lex to try it free for seven days and save 25% off your new subscription. That's blinkist.com lex Blinkist spelled B L I N K I S T. This show is also sponsored by Eight Sleep and its Pod Pro mattress that you can check out at eightsleep.com lex to get $200 off it controls temperature with a nap. It can cool down to as low as 55 degrees on each side of the bed. Separately, research shows that temperature has a big impact on the quality of our sleep. Anecdotally has been true. For me, it's truly been a game changer. I love it. The Pod Pro is packed with sensors that track heart rate, heart rate, variability and respiratory rate, showing it all in their app. The app's health metrics are amazing, but the cooling alone is honestly worth the money. Check it out@ asleep.com lex to get $200 off this show is also sponsored by Masterclass. Sign up@masterclass.com lex to get a discount and to support this podcast. When I first heard about Masterclass, I thought it was too good to be true. I for 180 bucks a year you get an all access pass to watch courses from to list some of my favorites. Chris Hadfield on Space Exploration, Neil DeGrasse Tyson on Scientific thinking and communication. Will Wright, one of my favorite game designers. Carlos Santana, one of my favorite guitar players, Garry Kasparov of course, the greatest chess player of all time. I'm not biased. Daniel Negrano on poker and many more. Chris Hadfield explaining how rockets work and the experience of being launched into space alone is worth the money. By the way, you can watch it on basically any device. Once again, sign up@masterclass.com lex to get a discount and to support this podcast. And now, here's my conversation with Manolis Kelis. What to you is the most beautiful aspect of the human genome?

Don't get me started.

We got time.

The first answer is that the beauty of genomes transcends humanity. So it's not just about the human genome. Genomes in general are amazingly beautiful. And again, I'm obviously biased. So in my view, the way that I like to introduce the human genome and the way that I like to introduce genomics to my class is by telling them, you know, we're not the inventors of the first digital computer, we are the descendants of the first digital computer. Basically, life is digital. And that's absolutely beautiful about life. The fact that at every replication step, you don't lose any information because that information is digital. If it was analog, if it was just brought in concentrations, you'd lose it. After a few generations, it would just dissolve away. And that's what the ancients didn't understand about inheritance. The first person to understand digital inheritance was Mendel, of course, and his theory, in fact stayed in a bookshelf for like 50 years while Darwin was getting famous about natural selection. But the missing component was this digital inheritance, the mechanism of evolution that Mendel had discovered. So that aspect, in my view, is the most beautiful aspect, but it transcends all of life.

And can you elaborate maybe the inheritance.

Part.

What was the key thing that.

The Ancients didn't understand so the very theory of inheritance as discrete units, you know, throughout the life of Mendel and well after his writing, people thought that his PE experiments were just a little fluke, that they were just a little exception that would normally not even apply to humans. That basically what they saw is this continuum of eye color, this continuum of skin color, this continuum of hair color, this continuum of height. And all of these continuums did not fit with a discrete type of inheritance that Mendel was describing. But what's unique about genomics and what's unique about the genome is really that there are two copies and that you get a combination of these. But for every trait there are dozens of contributing variables. And it was only Ronald Fisher in the 20th century that basically recognized that even five Mendelian traits would add up to a continuum like inheritance pattern. And he wrote a series of papers that still are very relevant today about sort of this Mendelian inheritance of continuum like traits. And I think that that was the missing step in inheritance. So well before the discovery of the structure of DNA, which is again, another amazingly beautiful aspect. The double helix, what I like to call the most noble molecule of our time, is, you know, holds within it the secret of that discrete inheritance. But the conceptualization of discretene, you know, elements is something that precedes that.

So even though it's discrete, when it materializes itself into actual traits that we see, it can be continuous, it can basically arbitrarily rich and complex.

So if you have five genes that contribute to human height, and there aren't five, there's a thousand. If there's only five genes and you inherit some combination of them and everyone makes you 2 inches taller or 2 inches shorter, it'll look like a continuum trait, a continuous trait. But instead of five, there are thousands. And every one of them contributes to less than 1 millimeter. We change in height more during the day than each of these genetic variants contributes. So by the evening, you're shorter than you woke up with.

Isn't that weird then, that we're not more different than we are? Why are we all so similar if there's so much possibility to be different?

Yeah. So there are selective advantages to being medium. If you're extremely tall or extremely short, you run into selective disadvantages. So you have trouble breathing, you have trouble running, you have trouble sitting if you're too tall, if you're too short, you might, I don't know, have other selective pressures are acting against that. If you look at natural history of human population, there's actually selection for height in northern Europe and selection against height in southern Europe. So there might actually be advantages to actually being not super tall. And if you look across the entire human population for many, many traits, there's a lot of push towards the middle balancing. Selection is the usual term for selection that sort of seeks to not be extreme and to sort of have a combination of alleles that sort of, you know, keep recombining. And if you look at, you know, mate selection, super, super tall people will not tend to sort of marry super, super tall people. Very often you see these couples that are kind of compensating for each other. And the best predictor of the kid's age is very often just take the average of the two parents and then adjust for sex and boom, you get it. It's extremely heritable.

Let me ask you kind of took a step back to the genome outside of just humans, but is there something that you find beautiful about the human genome specifically?

So I think the genome, if more people understood the beauty of the human genome, there would be so many fewer wars, so much less anger in the world. I mean, what's really beautiful about the human genome is really the variation that teaches us both about individuality and about similarity. So any two people on the planet are 99.9% identical. How can you fight with someone who's 99.9% identical to you? It's just counterintuitive. And yet any two siblings of the same parents differ in millions of locations. So every one of them is basically two to the million unique from any pair of parents, Let alone any two random parents on the planet. So that's, I think, something that teaches us about sort of the nature of humanity in many ways, that every one of us is as unique as any star and way more unique in actually in many ways. And yet we're all brothers and sisters. Yeah.

Just like stars, Most of it is just fusion reactions.

Yeah. You only have a few parameters to describe stars.

Yeah, exactly.

You know, mass, size, initial size, and know, stage of life. Whereas for humans, it's, you know, thousands of parameters scattered across our genome. So the other thing that makes humans unique, the other things that makes inheritance unique in humans is that most species inherit things vertically. Basically, instinct is a huge part of their behavior. The way that, you know, I mean, with my kids, we've been watching this nest of birds with two little eggs outside our window for the last few months, for the last few weeks as they've been growing. And there's so much behavior that's hard coded. Birds don't just learn as they grow. They don't, you know, there's no culture like a bird that's born in Boston will be the same as a bird that's born in California. So there's not as much inheritance of ideas, of customs. A lot of it is hard coded in their genome. What's really beautiful about the human genome is that if you take a person from today and you place them back in ancient Egypt, or if you take a person from ancient Egypt and you place them here today, they will grow up to be completely normal. That is not genetics. This is the other type of inheritance in humans. So on one hand we have genetic inheritance, which is vertical from your parents down. On the other hand, we have horizontal inheritance, which is the ideas that are built up at every generation are horizontally transmitted. And the huge amount of time that we spend in educating ourselves, a concept known as neoteny. Neo for newborn and then teni for holding. So if you look at humans, I mean, the little birds that were, you know, eggs two weeks ago, and now one of them has already flown off, the other one's ready to fly off in two weeks. They're ready to just fend for themselves. Humans, 16 years, 18 years, 24.

Getting out of college, I'm still learning. So. So that's so fascinating. This picture of a vertical and the horizontal, when you talk about the horizontal, is in the realm of ideas.

Exactly.

Okay, so it's the actual social interactions.

That's exactly right. That's exactly right. So basically the concept of neoteny is that you spend acquiring characteristics from your environment in an extremely malleable state of your brain and the wiring of your brain for a long period of your life. Compared to primates, we are useless. You take any primate at seven weeks, and in human, at seven weeks we lose the battle. But at 18 years, you know, all better off, like we basically, our brain continues to develop in an extremely malleable form till very late. And this is what allows education, this is what allows the person from Egypt to do extremely well now. And the reason for that is that the wiring of our brain and the development of that wiring is actually delayed. So, you know, the longer you delay that, the more opportunity you have to pass on knowledge, to pass on concepts, ideals, ideas from the parents to the child. And what's really absolutely beautiful about humans today is that that lateral transfer of ideas and culture is not just from uncles and aunts and teachers at school, but it's from Wikipedia and review articles on the web and thousands of journals that are sort of putting out information for free and podcasts and videocasts and all of that stuff where you can basically learn about any topic, pretty much everything that would be in any super advanced textbook in a matter of days instead of having to go to the Library of Alexandria and sail there to read three books and then sail for another few days to get to Athens and et cetera, et cetera, et cetera. So the democratization of knowledge and the spread, the speed of spread of knowledge is what defines I think, the human inheritance pattern.

So you sound excited about it, about it. Are you also a little bit afraid or are you more excited by the power of this kind of distributed spread of information? So you put it very kindly that most people are kind of using the Internet and looking Wikipedia, reading articles, reading papers and so on. But if we're honest, most people online, especially when they're younger, probably looking at 5 second clips on TikTok or whatever the new social network is, are you given this power of horizontal inheritance? Are you optimistic or a little bit pessimistic about this new effect of the Internet and democratization of knowledge on our. What would you call this? Would you use the term genome, by.

The way, for this? I think we use the genome to talk about DNA, but very often we say, I mean, when I'm Greek, so people ask me, hey, what's in the Greek genome? And I'm like, well yeah, what's in the Greek genome is both our genes and also our ideas and our ideals and our culture.

So the poetic meaning of the word.

Exactly, exactly, yeah, yeah. So I think that there's a beauty to the democratization of knowledge. The fact that you can reach as many people as you know any other person on the planet. And it's not who you are, it's really your ideas that matter is a beautiful aspect of the Internet. I think there's of course a danger of my ignorance is as important as your expertise. The fact that with this democratization comes the abolishment of respecting expertise just because you've spent 10,000 hours of your life studying, I don't know, human brain circuitry, why should I trust you? I'm just going to make up my own theories and they'll be just as good as yours is an attitude that sort of counteracts the beauty of the democratization. And I think that within our educational system and within the upbringing of our children, we have to not only teach them knowledge, but we have to teach them the means to get to Knowledge and that it's very similar to sort of you catch a fish for a man for one day, you fed them for one day, you teach them how to f, you fed them for the rest of their life. So instead of just gathering the knowledge they need for any one task, we can just tell them, all right, here's how you Google it, here's how you figure out what's real and what's not. Here's how you check the sources, here's how you form a basic opinion for yourself. And I think that inquisitive nature is paramount to being able to sort through this huge wealth of knowledge. So you need a basic educational foundation based on which you can then add on the sort of domain specific knowledge. But that basic educational foundation should not just be knowledge, but it should also be epistemology, the way to acquire knowledge.

I'm not sure any of us know how to do that in this modern day. We're actually learning. One of the big surprising things to me about the, the coronavirus, for example, is that Twitter has been one of the best sources of information. Basically like building your own network of experts, as opposed to the traditional centralized expertise of the WHO and the CDC or maybe any one particular respectable person at the top of a department at some kind of institution. You instead look at a, you know, 10, 20, hundreds of people, some of whom are young kids with just that, are incredibly good at aggregating data and plotting and visualizing that data. That's been really surprising to me. I don't know what to make of it. I don't know. I don't know how that matures into something stable. You know, I don't know if you have ideas like what if you were to just try to explain to your kids of how, where should you go to learn about coronavirus, what would you say?

It's such a beautiful example. And I think the current pandemic and the speed at which the scientific community has moved in the current pandemic, I think exemplifies this horizontal transfer and the speed of horizontal transfer of information. The fact that the genome was first sequenced in early January. The first sample was obtained Dec. 29, 2019, a week after the publication of the first genome sequence. Moderna had already finalized its vaccine design and was moving to production. I mean, this is phenomenal. The fact that we go from not knowing what the heck is killing people in Wuhan to wow, it's SARS COV2 and here's the set of genes, here's the genome, here's the sequence here are the polymorphisms, et cetera, in the matter of weeks is phenomenal. In that incredible pace of transfer of knowledge, there have been many mistakes. So some of those mistakes may have been politically motivated. Other mistakes may have just been innocuous errors. Others may have been misleading the public for the greater good, such as, don't wear masks because we don't want the masks to run out. I mean, that was very silly in my view, and a very big mistake. But the spread of knowledge from the scientific community was phenomenal. And some people will point out to bogus articles that snuck in and made the front page. Yeah, they did, but within 24 hours they were debunked and went out of the front page. And I think that's the beauty of science today. The fact that it's not, oh, knowledge is fixed. It's the ability to embrace that nothing is permanent when it comes to knowledge, that everything is the current best hypothesis and the current best model that best fits the current data. And the willingness to be wrong, the expectation that we're going to be wrong, and the celebration of success based on, how long was I not proven wrong for? Rather than, wow, I was exactly right. Because no one is going to be exactly right with partial knowledge. But the arc towards perfection, I think, is so much more important than how far you are on your first step. And I think that's what sort of the current pandemic has taught us. The fact that, yeah, no, of course we're going to make mistakes, but at least we're going to learn from those mistakes and become better and learn better and spread information better. So if I were to answer the question of where would you go to learn about coronavirus? First textbook. It all starts with a textbook. Just open up a chapter on virology and how coronaviruses work, then some basic epidemiology and sort of how pandemics have worked in the past. What are the basic principles surrounding these first wave, second wave, why do they even exist? Then understanding about growth, understanding about the R0 and RT at various time points, and then understanding the means of spread, how it spreads from person to person. Then how does it get into your cells from when it gets into the cells? What are the paths that it takes? What are the cell types that express the particular ACE2 receptor? How is your immune system interacting with the virus? And once your immune system launches its defense, how is that helping or actually hurting your health? What about the cytokine storm? What are most people dying from? Why are the comorbidities and these risk factors, even applying what makes obese people respond more or elderly people respond more to the virus, while kids are completely, you know, very often not even aware that they're spreading it. So the, you know, I think there's some basic questions that you would start from and then, I'm sorry to say, but Wikipedia is pretty awesome. Yeah, Google is pretty awesome.

It used to be a time. Used to be a time maybe five years ago. I forget, I forget when. But people kind of made fun of Wikipedia for being an unreliable source. I never quite understood it. I thought from the early days it was pretty reliable or better than a lot of the alternatives. But at this point, it's kind of like a solid, accessible survey paper on every subject ever.

There's an ascertainment bias and a writing bias. So I think this is related to people saying, oh, so many nature papers are wrong. And they're like, why would you publish in Nature? So many nature papers are wrong. And my answer is no, no, no. So many nature papers are scrutinized and just because more of them are being proven wrong than in other articles is actually evidence that they're actually better papers overall because they're being scrutinized at a rate much higher than any other journal. So if you basically judge Wikipedia by not the initial content, but by the number of revisions. Yeah. Then of course it's going to be the best source of knowledge eventually. It's still very superficial. You then have to go into the review papers, et cetera, et cetera, et cetera. But I mean, for most scientific project topics, it's extremely superficial, but it is quite authoritative because it is the place that everybody likes to criticize as being wrong.

You say that it's superficial on a lot of topics that I've studied a lot of. I find it. I don't know if superficial is the right word because superficial kind of implies that it's not correct.

No, no, I don't mean any implication of it not being correct. It's just superficial. It's basically only scratching the surface for depth. You don't go to Wikipedia, you go to the review articles.

But it can be profound in the way that articles rarely. One of the frustrating things to me about like certain computer science, like in the machine learning world articles, they don't as often take the bigger picture view. You know, there's a kind of data set and you show that it works and you kind of show that here's an architecture thing that creates an improvement and so on and so forth. But you don't say, well, like, what does this mean for the nature of intelligence for future data sets we haven't even thought about? Or, or if you were trying to implement this, like, if we took this data set of a hundred thousand examples and scaled it to a hundred billion examples, would this method, like, look at the bigger picture, which is what a Wikipedia article would actually try to do, which is like, what does this mean in the context of computer. The broad field of computer vision or something like that?

Yeah, no, I agree with you completely, but it depends on the topic. I mean, for some topics there's been a huge amount of work. For other topics, it's just a stub. So, you know, I got it.

Yeah, well, yeah, actually the well, which we'll talk on genomics was not.

Yeah, it's great. Very shallow. Yeah, yeah, it's not wrong. It's just shallow.

It's shallow, yeah.

Every time I criticize something, I should feel partly responsible. Basically, if more people from my community went there and edited, it would not be shallow. It's just that there's different modes of communication in different fields. And in some fields the experts have embraced Wikipedia, in other fields, it's relegated. And perhaps the reason is that if it was any better to start with, people would invest more time. But if it's not great to start with, then you need a few initial pioneers who will basically go in and say, ah, enough, we're just going to fix that. And then I think it'll catch on much more.

So if it's okay, before we go on to genomics, can we linger a little bit longer on the beauty of the human genome? You've given me a few notes. What else do you find beautiful about the human genome?

So the last aspect of what makes the human genome unique, in addition to the similarity and the differences and the individuality, is that so very early on people would basically say, oh, you don't do that experiment in human, you have to learn about that in fly, or you have to learn about that in yeast first, or in mouse first, or in a primate first. And the human genome was in fact relegated to sort of, oh, the last place that you're going to go to learn something new that has dramatically changed. And the reason that changed is human genetics. We are these species in the planet that's the most studied right now. It's embarrassing to say that, but this was not the case a few years ago. It used to be, you know, first viruses, then bacteria, then yeast, then the fruit fly and the worm, then the mouse. And eventually human was very far last.

So it's embarrassing that it took us this long to focus on it or the.

It's embarrassing that the model organisms have been taken over because of the power of human genetics, that right now it's actually simpler to figure out the phenotype of something by mining this massive amount of human data than by going back to any of the other species. And the reason for that is that if you look at the natural variation that happens in a population of 7 billion, you basically have a mutation in almost every nucleotide. So every nucleotide you want to perturb, you can go find a living, breathing human being and go test the function of that nucleotide by sort of searching the database and finding that person.

Wait, why is that embarrassing? It's a beautiful data set.

It's embarrassing for the, for the model organism.

For the flies.

Yeah, exactly.

I mean, do you, do you feel, on a small tangent, is there something of value in the genome of a fly and other these model organisms that you miss that we wish we would have. Would be looking at deeper?

So directed perturbation? Of course. So I think the place where, the place where humans are still lagging is the fact that in an animal model you can go and say, well, let me knock out this gene completely and let me knock out these three genes completely. And the moment you get into combinatorics, it's something you can't do in the human because there just simply aren't enough humans on the planet. And again, let me be honest, we haven't sequenced all 7 billion people. It's not like we have every mutation, but we know that there's a carrier out there. So if you look at the trend and the speed with which human genetics has progressed, we can now find thousands of genes involved in human cognition, in human psychology, in the emotions and the feelings that we used to think are uniquely learned. Turns out there's a genetic basis to a lot of that. So the human genome has continued to elucidate through these studies of genetic variation so many different processes that we previously thought were something that, like free will. Free will is this beautiful concept that humans have had for a long time. In the end, it's just a bunch of chemical reactions happening in your brain. And the particular abundance of receptors that you have this day based on what you ate yesterday day, or that you have been wired with based on, you know, your parents and your upbringing, etcetera, determines a lot of that quote, unquote, free will component to, you know, sort of narrower and narrower, you know, sort of slices.

So how much on that point, how much freedom do you think we have to escape the, the constraints of our genome? You're making it sound like more and more we're discovering that our genome is actually has the, a lot of the story already encoded into it. How much freedom do we have?

I. So let me describe what that freedom would look like. That freedom would be my saying, ooh, I'm going to resist the urge to eat that apple because I choose not to. But there are chemical receptors that made me not resist the urge to prove my individuality and my free will by resisting the apple. So then the next question is, well, maybe now I'll resist the urge to resist the apple and I'll go for the chocolate instead to prove my individuality. But then what about those other receptors.

That, you know, that might be all encoded in there?

So it's kicking the bucket down the road and basically saying, well, your choice may have actually been driven by other things that you actually are not choosing. So that's why it's very hard to answer that question.

It's hard to know what to do with that. I mean, if, if the genome has, if, if there's not much freedom, it's.

It'S the butterfly effect. It's basically that in the short term, you can predict something extremely well by knowing the current state of the system. But a few steps down, it's very hard to predict based on the current knowledge. Is that because the system is truly free? When I look at weather patterns, I can predict the next 10 days. Is it because the weather has a lot of freedom and after 10 days it chooses to do something else? Or is it because in fact the system is fully deterministic and there's just a slightly different magnetic field of the earth, slightly more energy arriving from the sun, a slightly different spin of the gravitational pull of Jupiter that is now causing, you know, all kinds of tides and slight deviation of the moon, etc. Maybe all of that can be fully modeled. Maybe the fact that China is emitting a little more carbon today is actually going to affect the weather in Egypt in three weeks. And all of that could be fully modeled in the same way. If you take a complete view of a human being now, I model everything about you. The question is, can I predict your next step? Probably. But at how far? And if it's a little further, is that because of stochasticity and sort of chaos properties of unpredictability of beyond a certain level, or was that actually True free will.

Yeah. So the number of variables might be. So you might need to build an entire universe to be able to simulate.

A human, and then maybe that human will be fully simulatable, but maybe aspects of free will will exist. And where's that free will coming from? It's still coming from the same neurons or maybe from a spirit inhabiting these neurons. But again, it's very difficult empirically to sort of evaluate where does free will begin and sort of chemical reactions and electric signals end.

So on that topic, let me ask the most absurd question that most MIT faculty roll their eyes on. But what do you think about the simulation hypothesis and the idea that we live in a simulation?

I think so. Complete bs okay. There's no empirical evidence. No, it's not. Absolutely not. Not.

Not in terms of empirical evidence to not. But in terms of thought experiment, does it help you think about the universe? I mean, so if you look at the genome, it's encoding a lot of the information that is required to create some of the beautiful human complexity that we see around us. It's an interesting thought experiment. How much, you know, parameters do we need to have in order to model some, you know, this full human experience? Like if we were to build a video game.

Yeah.

How hard it would be to build a video game that's like, convincing enough and fun enough and, you know, has consistent laws of physics, all that stuff. It's not interesting to use a thought experiment.

I mean, it's cute, but, you know, it's. It's Occam's razor. I mean, what's. What's more realistic? The fact that you're actually a machine or that you're, you know, a person? What's. What's, you know, the fact that all of my experiences exist inside the chemical molecules that I have, or that somebody's actually, you know, simulating all that?

Well, you did refer to humans as a digital computer earlier, so.

Of course, of course. But that does not.

Kind of a machine, right?

I know, I know, but I think the probability of all that is nil. And let the machines wake me up and just terminate me now, if it's not. I challenge your machines.

They're going to wait a little bit to see what you're going to do next. It's fun. It's fun to watch, especially the clever humans. What's the difference to you between the way a computer stores information and the human genome stores information? So you also have roots and your work. Would you say you're. When you introduce yourself at a bar.

It depends who I'm talking to.

Would you say it's computational biology? Do you. Do you reveal your expertise in computers?

It depends who I'm talking to, truly. I mean, basically, if I meet someone who's in computers, I'll say, oh, I mean, professor in computer science. If I meet someone who's in engineering, I say computer science and electrical engineering. If I meet someone in biology, I'll say, hey, I work in genomics. If I meet someone in medicine, I'm like, hey, I work on, you know, genetics.

You're a fun person to meet at a bar. I got you. But, so.

No, no, but what I'm trying to say is that I don't. I mean, there's no single attribute that I will define myself as. You know, there's a few things I know, there's a few things I study, there's a few things I have degrees on, and there's a few things that it grants degrees in. And I publish papers across the whole gamut, the whole spectrum of computation to biology, et cetera. The complete answer is that I use computer science to understand biology. I develop methods in AI and machine learning, statistics in algorithms, et cetera. But the ultimate goal of my career is to really understand biology. If these things don't advance our understanding of biology, I'm not as fascinated by them. Although there are some beautiful computational problems by themselves. I've sort of made it my mission to apply the power of computer science to truly understand the human genome, health, disease, you know, and the whole gamut of how our brain works, our body works, and all of that, which is so fascinating.

So the dream, there's not an equivalent sort of complementary dream of understanding human biology in order to create an artificial life, an artificial brain, artificial intelligence that supersedes the intelligence and the capabilities of us humans.

It's an interesting question. It's a fascinating question. So understanding the human brain is undoubtedly coupled to how do we make better AI? Because so much of AI has, in fact been inspired by the brain. It may have taken 50 years since the early days of neural networks, till we have all of these amazing progress that we've seen with deep belief networks and all of these advances in go, in chess, in image synthesis, in deepfakes, in, you name it, but the underlying architecture is very much inspired by the human brain, which actually posits a very, very interesting question. Why are neural networks performing so well? And they perform amazingly well. Is it because they can simulate any possible function? And the answer is no. No, they simulate a very small number of functions. Is it because they can simulate every partial function in the universe? And that's where it gets interesting. The answer is actually, yeah, a little closer to that. And here's where it gets really fun. If you look at human brain and human cognition, it didn't evolve in a vacuum. It evolved in a world with physical constraints, like the world that inhabits us. It is the world that we inhabit. And if you look at our senses, what do they perceive? They perceive different, you know, parts of the electromagnetic spectrum. You know, the hearing is just different movements in air, the. The touch, etc. I mean, all of these things, we've built intuitions for the physical world that we inhabit, and our brains, and the brains of all animals evolved for that world. And the AI systems that we have built happen to work well with images of the type that we encounter in the physical world that we inhabit. Whereas if you just take noise and you add random signal that doesn't match anything in our world, neural networks will not do as well. And that actually basically has this whole loop around this, which is, this was designed by studying our own brain, which was evolved for our own world, and they happen to do well in our own world, and they happen to make the same types of mistakes that humans make many, many times. And of course, you can engineer images by adding just the right amount of, you know, sort of pixel deviations to make a zebra look like a bamboo and stuff like that, or like a table. But ultimately, the undoctored images, at least, are very often mistaken. I don't know, between muffins and dogs, for example, in the same way that humans make those mistakes. So there's no doubt, in my view, that the more we understand about the tricks that our human brain has evolved to understand the physical world around us, the more we will be able to bring new computational primitives in our AI systems to again, better understand not just the world around us, but maybe even the world inside us, and maybe even the computational problems that arise from new types of data that we haven't been exposed to, but are yet inhabiting the same universe that we live in, with a very tiny little subset of functions from all possible mathematical functions.

Yeah. And that small subset of functions is all that matters to us humans, really. That's what makes.

It's all that has mattered so far. And even within our scientific realm, it's all that seems to continue to matter. But I mean, I always like to think about our senses and how much of the physical world around us we perceive. And if you look at the LIGO experiment over the last year and a half has been all over the news. What did LIGO do? It created a new sense for human beings, a sense that has never been sensed in the history of our planet. Gravitational waves have been traversing the Earth since its creation a few billion years ago. Life has evolved senses to sense things that were never before sensed. Light was not perceived by early life, no one cared. And eventually photoreceptors evolved and the ability to sense colors by sort of catching different parts of that electromagnetic spectrum, and hearing evolved, and touch evolved, etc. But no organism evolved a way to sense neutrinos floating through Earth or gravitational waves flowing through Earth, et cetera. And I find it so beautiful in the history of not just humanity, but life on the planet that we are now able to capture additional signals from the physical world than we ever knew before. And axions, for example, have been all over the news in the last few weeks. The concept that we can capture and perceive more of that physical world is as exciting as the fact that we were blind to it is traumatizing before, because that also takes.