Steve Hsu - Intelligence, Embryo Selection, & The Future of Humanity

Steve Hsu is a Professor of Theoretical Physics at Michigan State University and cofounder of the company Genomic Prediction https://www.lifeview.com/. We go deep into the weeds on how embryo selection can make babies healthier and smarter. Watch on YouTube https://youtu.be/80BhjRh-Q-s. Listen on Apple Podcasts https://apple.co/3wob9AK, Spotify https://open.spotify.com/episode/5RUU4dwbgn02WuAuqoLhWL?si=TFkt0ZL-QRa7fYGMGQRumA, or any other podcast platform. Read the full transcript here https://www.dwarkeshpatel.com/p/steve-hsu. Follow Steve on Twitter https://twitter.com/hsu_steve. Follow me on Twitter https://twitter.com/dwarkesh_sp for updates on future episodes. Timestamps (0:00:14) - Feynman’s advice on picking up women (0:11:46) - Embryo selection (0:24:19) - Why hasn't natural selection already optimized humans? (0:34:13) - Aging (0:43:18) - First Mover Advantage (0:53:38) - Genomics in dating (0:59:20) - Ancestral populations (1:07:07) - Is this eugenics? (1:15:08) - Tradeoffs to intelligence (1:24:25) - Consumer preferences (1:29:34) - Gwern (1:33:55) - Will parents matter? (1:44:45) - Wordcels and shape rotators (1:56:45) - Bezos and brilliant physicists (2:09:35) - Elite education This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.dwarkesh.com https://www.dwarkesh.com?utm_medium=podcast&utm_campaign=CTA_1

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

• 0:00Dwarkesh Patel?

  Today, I have the pleasure of speaking with Steve Shu. Steve, thanks for coming on the podcast. I'm excited about this.

• 0:05Speaker 01

  Hey, it's my pleasure. I'm excited too. And I just want to say I've listened to some of your earlier interviews and thought you were very insightful, which is why I was really excited to have a conversation with you.

• 0:14Dwarkesh Patel?

  That means a lot for me to hear you say because I'm a big fan of your podcast. My first question is, what advice did Richard Feynman give you about picking up girls?

• 0:24Speaker 01

  Wow.

• 0:27Speaker 01

  So, one day in the spring of my senior year, I was walking across campus. And I see Feynman coming toward me, and we knew each other from various things. And it's a small campus, and I was a physics major and he was my hero. So I guess I had known him since my freshman year.

• 0:46Speaker 01

  So he sees me and, you know, he's got this, I don't know if it's, I guess it's a Long Island or it's some kind of New York borough accent. And he says, "Hey, Shu." This is how he says my name, "Hey, Shu." And I'm like, "Hi, Professor Feynman." And so we start talking. And he says to me, "Wow, you're kind of a big guy." And I was a lot bigger then because I played on the, I was a linebacker on the Celtic football team. So I was about almost 200 pounds. I'm just over six feet tall. And so I was pretty like a gym rat at that time. And so he was like, I was much bigger than him, obviously. He was like, "Wow, you're a big guy, Steve. I gotta ask you something."

• 1:31Speaker 01

  And Feynman was born in like 1918. So he's not really like from the modern era. Like he was, he was, I guess he was going through graduate school when the Second World War started. And so to him, the whole concept of a health club, a gym, was like totally, you know, couldn't understand it. And that was the era, this was the 80s. So that was the era when Gold's Gym was like becoming a world, a national franchise. And so there were gyms all over the place, 24 Hour Fitness and stuff like this. So he didn't know what it was. And he's a very interesting guy. So he, his suspicion, he says to me, What do you guys do there? Is it just a thing to meet chicks, to meet girls? Or is it really for training?

• 2:14Speaker 01

  Do you guys really go there to get buff, to get big? And so I started explaining to him, I said, yes, you know, people are there to get big, but people are also checking out the girls and there is a lot of stuff happening at the health club or in the weight room. And so, you know, he grills me on this for a long time. And one of the famous things about Feynman is that he has this laser-like focus. So if there's something he really doesn't understand and he wants to get to the bottom of it, he will just focus in on you and just start questioning you and get to the bottom of it. That's the way his brain works. So he did that to me for like, I don't know how long. We were talking about lifting weights and everything because he didn't know anything about it. And at the end, he says to me, wow, Steve, I really appreciate that.

• 2:59Speaker 01

  Let me, you know, let me give you some good advice. And so then he starts telling me about how to pick up girls. And which I guess he, you know, he's a kind of an expert on. And he says to me, he goes, he goes, one of the things he says to me, he says like, I don't know how much girls really like guys that are as big as you. Like, I don't, I'm not, he thought like it might be a turn off actually. And he said, but you know what? You have a nice smile. So that was, that was the one compliment, you know, he gives me, you have a nice smile. And then he starts telling me, he says, you know, the main thing is it's a numbers game. Okay, you have to divorce your, you have to be totally rational about it. You're never going to see that girl again.

• 3:45Speaker 01

  Right? You're, you're in an airport lounge or you're, you're, you're at a bar. It's Saturday night in Pasadena or Westwood. And you're talking to some girl. And he says, you're never going to see her again. This is your one interaction with her. Five-minute interaction. Do what you have to do. And if she, for some reason doesn't like you, just go to the next one. And that's what he says. So, uh, you know, and he gives some colorful details and stuff. Uh, but the point is, he's like, you should not care what they think of you. You're, you're trying to do your thing. And, you know, he's a pretty,

• 4:23Speaker 01

  he had a kind of a reputation at Caltech as a womanizer. I could go into that too, but, uh, I heard this from the secretaries and stuff, but, um.

• 4:32Dwarkesh Patel?

  With the students or with like?

• 4:33Speaker 01

  No, no, with secretaries, mostly secretaries. We're almost all female at that time. He, he, he had thought about this a lot and he was just like, look, it's a numbers game. Just, I guess the, the PUA type, are you familiar with PUA culture?

• 4:47Dwarkesh Patel?

  Yeah.

• 4:48Speaker 01

  So the PUA guys would say like, yeah, don't, you know, it's like a operation, like you're, you're just doing something. You follow the algorithm. And you, whatever happens, it's not a reflection on your self-esteem. Or your internal self-image. It's just, that's what happened. And you just go on to the next one. And that was basically the advice he was giving me. Um, you know, and he said other things which were pretty standard, like, you know, be funny. You're a funny guy. You know, girls like that. Be confident, you know? Just basic stuff. But the main thing I remember was the oper- operationalization of it as an algorithm and that you should just not internalize whatever happens if you get rejected. Cause that's what really hurts. So you know, you're a guy, right? When you go across the bar to talk to that girl, maybe that doesn't happen in your generation. Maybe you just like swipe.

• 5:32Speaker 01

  But, but we had to go, it was terrifying. We had to go across the bar and talk to some lady and it's loud and you, you've got like a few minutes to make your case, basically. And, um, nothing hurts more and nothing is more scary than walking across up to the girl. Maybe she and her friends or something, right? So he was just saying like, you gotta train yourself out of that. Like you're never going to see them again. The space, space of humanity is so big, you'll never encounter them again. And it just doesn't matter. So just do your best.

• 6:07Dwarkesh Patel?

  Yeah, that's interesting because I wonder when he would, when, I mean, in the 40s, when he was at that age, was he doing this? I don't know what the cultural conventions were at the time, but I don't know, during the, were there bars in the 40s where you could just go hit on, hit on girls or?

• 6:19Speaker 01

  Oh, yeah, absolutely, absolutely. I mean, if you read literature from that time, or even a little bit earlier, like Hemingway, or John O'Hara, or, you know, they talk about, you know, how men and women interacted in bars and stuff like this, and in New York City, and, you know, um, yeah, so that was a thing. That was much more of a thing than I think for your generation. That, that's what I can't figure out with my kids. Like, what is going on? Like, how do, how do boys and girls meet these days? But, uh, back in the day, it was like the guy had to do all the work. And it was like you're the most terrifying thing you could do. And, um, you know, and you just have to train yourself out of that.

• 6:57Dwarkesh Patel?

  Right. Uh, but by the way, when, uh, for, for context for the audience, when Feynman says you were a big guy, like you, you were a football player at Caltech, right? And then there's a picture of you actually on, uh, your website where I, maybe this was after, after college or something, but yeah, you, you look like, uh, I, pretty ripped and it's kind of, um, I mean, I, today it seems more common because of gym culture and stuff, but I don't know, back, back then, I don't know how common that kind of, uh, that kind of body physique was.

• 7:24Speaker 01

  It's, it's amazing that you asked this question. I'll tell you a funny story because I was, one of the reasons Feynman found this so weird was because the way bodybuilding entered the United States or became widespread was a very interesting story because at first they were regarded as freaks and homosexuals and all kinds of stuff. And I remember growing up, our high school football coach, um, swimming was different. Swimming, because it's international, swimming picked up a lot of advanced training techniques from the Russians and from East Germans and stuff, but football was more, um, you know, uh, more kind of just American and not very international. And so

• 8:05Speaker 01

  our football coach used to tell us not to lift weights when we were, maybe when I was in junior high school. And they said it makes you slow. You're being bulky is no good. You gotta be, you know, you gotta be, uh, fast in football. And then something changed around the time I was in high school where the coaches figured out, because the swimmers, as a swimmer, I'd been lifting weights since I was an age group swimmer, like maybe at age 12 or 14, I started lifting weights. So, but then the football coaches got into it, and mainly because the University of Nebraska, and the University of Nebraska had a very famous strength program that really popularized it. And so, at the time, though, there just weren't a lot of big guys. And the people who knew how to train, the way everybody like you probably go to the gym and train using what would

• 8:49Speaker 01

  be considered kind of advanced knowledge back in the 80s, okay, like how to do a split routine, or squat on one day and do your upper body on the other day, next day, that was like considered advanced knowledge at that time. And so I remember once I had an injury, and it was in the trainer's room at the Caltech Athletic Facility. And the lady was looking, it was a female trainer, and she was looking at my quadriceps. And, because I'd pulled a muscle, and she was looking at the, if you know your anatomy, like right above your kneecap, your quadriceps kind of insert right above your kneecap. And if you have well-developed quads, you have, actually have a bulge, a bump, right above your kneecap. And she was looking at it from this angle where she was in front of me, and she was looking at my leg from the front, and she's like, wow, it's really

• 9:32Speaker 01

  swollen. And I was like, that's not the injury, that's my quadricep muscle. And she was a trainer. So, you know, and at that time, like I could probably squat, I could maybe squat 400 pounds at that time. So I was pretty strong, right? And I had big legs. And so anyway, the fact that the trainer didn't really understand like what well-developed anatomy was supposed to look like was just blew my mind. I was like, no, that's my, that's my quadricep, we built that up. And she's like, oh, I thought that was the injury. I was like, what, what are you talking about? So, anyway, we've come a long way. This is one of these things where you gotta be old to have any kind of understanding of how this stuff evolved over the last, you know, 30, 40 years.

• 10:13Dwarkesh Patel?

  But you know, I wonder if that was a phenomenon of that particular time? Or if like, if throughout human history, people have just not been that muscular?

• 10:21Speaker 01

  Never.

• 10:22Dwarkesh Patel?

  Because you hear stories of like Roman soldiers who are carrying like 80 pounds for 10 or 20 miles a day. And I mean, there's like a lot of sculptures in the ancient world where, I mean, not that ancient, but like the people look like they have well-developed musculature.

• 10:34Speaker 01

  So the Greeks were very special because they were the first to really think about the word gymnasium and there was a thing called the palestra, which where they would train like wrestling and boxing and stuff like this. They were the first people who were really seriously into physical culture and training, specific training for athletic competition. But if you look at like even in the 70s, so when I was a little kid and I remember in the 70s and now when I look back at old photos from the 70s, it's very apparent, guys are skinny, guys are so skinny. You know, the guys who went off and fought World War II, whether they were on the German side or the American side, they were like 5'8", 5'9" and they weighed like 130 pounds, 140 pounds.

• 11:17Speaker 01

  They were totally different than what modern US Marines would, you would think of look like, right? So yeah, physical culture was a new thing. Of course, the Romans and the Greeks had it to some degree, but it was kind of lost for a long time. And it was just coming back in the US when I was growing up. And so, yeah, if you, if you were, you know, 200 pounds of fairly lean 200 pounds and you could bench over 300, that was pretty rare back in those days.

• 11:46Dwarkesh Patel?

  Yeah, yeah, yeah. Okay, so let's talk about your company, Genomic Prediction. Yeah, do you want to talk about what this company does? Do you want to give an intro into what this is?

• 11:56Speaker 01

  Yeah, so if you don't mind what I should say, there are two ways to introduce it. One is the scientific view and then the other is the IVF view. And I can kind of do a little of both. So scientifically, the issue is

• 12:11Speaker 01

  We have more and more genomic data. If you give me a bunch, the genomes of a bunch of people and then you give me some information about each person like, do they or do they not have diabetes or how tall are they? Or, you know, what's their IQ score or something? Then any, all of your listeners will be familiar with AI and machine learning. It's a natural AI, machine learning problem to figure out which features in the DNA variation between people are predictive of whatever variable you're trying to predict, whatever the biological term is phenotype. So this is an ancient scientific question of how do you relate the genotype of the organism, the specific DNA pattern to the phenotype, the actual expressed characteristics of the organism?

• 12:56Speaker 01

  And if you think about it, this is what biology is. Like, once we had the molecular revolution and people figured out that DNA is the thing which stores the information, which is passed along and evolution selects on the variation in the DNA as it's expressed as phenotype and as that phenotype affects fitness, okay? Or reproductive success. All, that's the whole ballgame for biology. And I'm lucky that as a physicist who's trained in kind of mathematics and computation, I arrived on the scene at a time when we're going to solve this basic fundamental problem of biology through brute force AI and machine learning. So that, that's how I kind of got into this, right?

• 13:38Speaker 01

  Now you ask as an entrepreneur, like, okay, fine, you, Steve, you're doing this in your office with your postdocs and collaborators on your computers and stuff, but what use is it? Right? What, what, what use is all this stuff? The most direct application of this is in the following setting. Every year around the world, there are millions of families that go through IVF, typically because they're having some fertility issues and also mainly typically because the mother is older. Like typically in her 30s or maybe 40s. And in the process of IVF, because they use hormone stimulation, they generally produce more eggs.

• 14:20Speaker 01

  Instead of one per cycle, they might produce, depending on the age of the woman, anywhere between five, or 10, or 20, or even, I recently learned, for young women who are hormonally stimulated, if they're egg donors, they could produce 60 or 100 eggs in one retrieval cycle. And then, it's trivial, as you know, men produce sperm all the time, or just producing it. You can fertilize those eggs pretty easily, in a little dish, and you get a bunch of embryos, which, they grow, they just start growing once they're fertilized. Now, the problem is, if you're a family, and you produce more embryos than you're going to use, you have what we call the embryo choice problem.

• 15:03Speaker 01

  You have to figure out, like, okay, I have these 20 viable embryos, which one am I going to use? And so, the most direct application of the science that I described is, well, we can now genotype those embryos, from a small biopsy. And I can tell you things about the embryos. I could tell you, hey, number four is an outlier for breast cancer risk. I would think carefully about using number four. Number 10 is an outlier for cardiovascular disease risk. You might want to think about not using that one. The other ones are okay. And so, that is what genomic prediction does, and I think we work with 200 or 300 different IVF clinics on six continents now.

• 15:46Dwarkesh Patel?

  Yeah, yeah. So, the super fascinating thing about this is that the diseases you talked about, or at least their risk profiles, they're, they're polygenic. So, you can have thousands of SNPs, single nucleotide polymorphisms, that determine whether you're going to get this disease or not. And, so, I'm really curious to learn, like, how you were able to transition to this space, and like, how your knowledge of mathematics and physics was able to help you figure out how to make sense of all this data?

• 16:15Speaker 01

  Yeah, that's a great question. So, you know... First of all, again, like I was kind of stressing like the fundamental scientific importance of all this stuff. If you go into a slightly higher level of detail, which you were getting at with the individual SNPs or polymorphisms, those are individual locations in the genome where I might differ from you and you might differ from another person. And typically, if you just take pairs of individuals, each human, each pair of individuals will differ at a few million places in the genome. Okay, and that's what's controlling. That's why I look a little different than you and, you know, so. Just a little. Just, yeah, a little bit, but I mean, you look better than me, but you know.

• 16:57Speaker 01

  The, uh, the question is the following. So a lot of times what theoretical physicists do is they, they have a little spare energy. They have some spare cycles and they, they get tired of thinking about quarks or something. And they wanna like maybe dabble in biology or they wanna dabble in computer science or some other field. And the thing that we always have to do as theoretical physicists, we always feel like, oh, I, I have a lot of horsepower. I can figure a lot out, a lot of stuff out. Like, for example, Feynman helped to design the first parallel processors at Thinking Machines. I gotta figure out which problems I can actually make an impact on, because other, I can waste a lot of time. Some people spend their whole lives studying one problem, like one molecule or something, or one, you know, biological system, and I, I don't have time for that. I'm just gonna jump in and jump out. I'm a physicist, right?

• 17:41Speaker 01

  That, that's a typical attitude among theoretical physicists. So the, the thing that I had to confront about 10 years ago was I knew the rate at which sequencing costs were going down. So I could anticipate we would get to the day, today, when there are millions of genomes with good phenotype data available for analysis. So that a typical run for us, a training run, might involve almost a million genomes or half a million genomes or something. So the, the mathematical question is, what is the most effective algorithm, given a set of genomes and phenotype information, to build the best predictor, right? So it's, it's, it can be boiled down to a very well-defined machine learning problem.

• 18:25Speaker 01

  And it turns out for some subset of algorithms, there are theorems. There are actually performance guarantees that tell you, that give you a bound on how much data you need to capture almost all of the variation in the in the in the in the features. And so I spent actually a fair amount of time, like probably a year or two studying these results, very famous results. Some of them were proved by a guy called Terrence Tao, who's a field medalist. And these are results on something called compressed sensing, which is a penalized form of high dimensional regression, which tries to build sparse predictors.

• 19:08Speaker 01

  Machine learning people might know it as L1 penalized optimization. And anyway, so the point is, we, the early, the very first paper we wrote on this was to prove that using real genomic genomic data, that these theorems that were very abstract could be applied in order to predict how much data you would need to quote, solve individual human traits. So we showed that you would need at least a few, around a few hundred thousand individuals and their heights, their genomes and their heights to solve height as a phenotype. And we proved that in a paper, using all this fancy math in 2012, I want to say the paper came out around 2012.

• 19:51Speaker 01

  And then around 2017, when we got a hold of half a million genomes, we were able to implement it in practical terms and show that our mathematical result from some years ago was correct. And we, the the transition from low performance of the predictor to high performance, there's a kind of what we call a phase transition boundary between those two domains occurred just where we said it was going to occur. So some of these technical details are really just not understood even by practitioners in computational genomics who are not Quite that mathematical. They don't understand actually these results that in our earlier papers, they don't really know why we can do stuff that other people can't do, or why we can predict how much data we're gonna need to do stuff. It's not well appreciated even in the field.

• 20:35Speaker 01

  But if you look carefully, when the the the big future AI in the in in in our future, in the singularity, looks back and says, hey, who gets the most credit for this genomics revolution that happened in the early 21st century? They're gonna find that AI is gonna find these papers on the archive in which we proved this was possible. And then five years later, we did it, and et cetera, et cetera. Right now, it's underappreciated, but the the future AI that Rocco's basilisk AI, when he looks back, is gonna give me a little bit of credit for.

• 21:04Dwarkesh Patel?

  Yeah, yeah. So I I was kind of a little interested in this a few years ago. And then at that time, I looked into like how these polygenic risk scores are calculated. And it was basically you just find the correlation between the phenotype and the the the the alleles that correlate with it. And you just add up how many copies of this allele do you have? What is the correlation? So it seemed like, and you just do a weighted sum of that. So that seemed like a very, um, it just seemed super simple, especially in an era where we have all this machine learning. And but it seemed like they were getting good predictive results out of that. So what is the delta between how good you can go with all this fancy mathematics versus just like a very simple, like, sum of correlations?

• 21:43Speaker 01

  Yeah, so you you you're absolutely right that the ultimate models that are used when you've done all the training and your the dust settles. The models are very simple. They have an additive structure. So it's basically like, I either assign a non-zero weight to this particular region in the genome, or I don't. And then I need to know what is the weighting. But then the the function is a linear function of it's an additive function of the state of your genome at some subset of positions. So the ultimate model that you get is very simple. Now, if you go back 10 years when we were doing this, there were lots of claims that it was gonna be super nonlinear, that it wasn't gonna be additive, the way I just described it. It was gonna there were gonna be lots of interaction terms between regions.

• 22:27Speaker 01

  People, some biologists are still convinced that's true, even though like we already know, like, we have predictors that don't have interactions. Okay, the other question which is more technical is that. There in any small region of your genome, the state of the individual variance is highly correlated because you inherit them in chunks. And so you need to figure out which one of those you want to use. You don't want to activate all of them because you might be overcounting. So that's where this L1 penalization sparse methods, they force the predictor to be sparse. And that is a key step. Otherwise, you might overcount. You might have 10, 10 different variants close by that have roughly the same statistical significance if you just do some simple regression math.

• 23:12Speaker 01

  But then you don't know which one of those tend to use and you might be overcounting effects or undercounting effects. So, so this, what you end up doing is a super high dimensional optimization where you, you only activate, you grudgingly activate a SNP when the signal is strong enough. And you, once you activate that one, the algorithm has to be smart enough to penalize the other ones nearby and not activate them because you're overcounting effects if you do that. So there's a little bit of subtlety into it, in it, but the, the main point which you made, which is that the ultimate predictors, which are very simple and additive, just sums over effect sizes times states, actually works really well. And that is related to a deep statement about the additive structure of the genetic architecture of individual differences.

• 23:56Speaker 01

  So in other words, it's kind of weird that the ways that I differ from you are merely just because I have more of something and you have less of something. And it's not like, oh, these things are interacting in some super, you know, incredibly un-understandable way. And so that's a very deep thing, which again is not appreciated that much by biologists yet, but over time, I think they're going to figure out that there's something interesting here.

• 24:19Dwarkesh Patel?

  Right, no, I thought that was, that was super fascinating and I, I commented on that about that on Twitter. What is really interesting about that is, I guess, two things. One is you have this really interesting evolutionary argument about why that would be the case. You, you might want to explain. And the second is, it makes you wonder if it's just, if just becoming more intelligent is just a matter of like turning on certain SNPs. It's not a matter of like all this incredible optimization that, you know, it's, it's like solving a Sudoku puzzle or anything. If, if that's the case, then why aren't we already, why hasn't the human population already been selected to be maxed out on all these traits if it's just a matter of a bit flip?

• 25:01Speaker 01

  Yeah, so, okay, so the first issue, which is how, why, you know, why is this, why is this genetic architecture so simple, surprisingly simple? And again, 10 years ago, we didn't know it was going to be simple. So when we were, when I was checking to see whether this is a field that I should go into, because either we are capable or not capable of making progress. We had to study the more general problem of the nonlinear possibilities as well. But eventually, we realized that probably most of the variance was going to be captured in an additive way. So, you know, we could narrow down the problem quite a bit. There are evolutionary reasons for this. There's a famous theorem by Fisher, who's the father of population genetics and also of really what you call frequentist statistics. And so Fisher proved something called the fundamental, Fisher's fundamental theorem

• 25:46Speaker 01

  of natural selection, which says that if you impose some selection pressure on a population, the rate at which that population responds to this selection pressure, like say, like it's the bigger rats that outcompete the smaller rats. At what rate does the rat population then start getting bigger? He showed that it's dominated by the additive variance, that that dominates the rate of evolution. And it's easy to understand why. If it's a, if it's a nonlinear mechanism that you need to make the rat bigger, when you sexually reproduce and that gets chopped apart, you might break the mechanism. Whereas if each little allele has its own independent effect, you can just, you can

• 26:27Speaker 01

  just inherit them without worrying about breaking the mechanisms. So it was well known for, at least among a tiny population of theoretical population biologists, that that additive variance was the dominant way that populations would respond to selection. So that was already known. And the other thing is that humans have been through a pretty tight bottleneck and were not that different from each other. So it's very plausible to me that if I wanted to edit an embryo, a human embryo, and make it into a frog, then there's all kinds of nonlinear subtle things I have to do. But all those very nonlinear, complicated subsystems are fixed in humans. You have the same system as I do.

• 27:10Speaker 01

  You have the non, the human not frog, or ape not frog version of that region of DNA, and so do I. But the small ways in which we differ are just these little additive switches, mostly little additive switches. And so that's the this deep scientific discovery over the from the last, say, five, 10 years of work in this area. Now, you were asking about why evolution hasn't like completely quote optimized all traits in humans already. Now, I don't know if you ever do deep learning or very high dimensional optimization, but you realize like in that high dimensional space, you're often moving on a surface which is slightly tilted, so you're getting gains, but it's also kind of flat.

• 27:53Speaker 01

  So even though you like scale up your compute or data size by an order of magnitude, you don't move that much farther. You get some gains, but you're never really at the global max of anything in this in these high dimensional spaces. I don't know if that makes sense to you, but but it's quite plausible to me that two things are important here. One is evolution has not had that much time to optimize humans. And what do you mean by optimization? Because the environment that humans live in has changed radically in the last 10,000 years, like for a while we didn't have agriculture, now we have agriculture, now we have swipe left if you want to have sex tonight. You know, we the the environment didn't stay fixed. And so when you say like fully optimized for the environment, what what do you mean?

• 28:37Speaker 01

  The ability to diagonalize matrices might not have been very adaptive 10,000 years ago. It might not even be adaptive now. But but anyway, so so it's a complicated question. One can't reason that naively about, oh, well, if if God wanted us to be 10 feet tall, we'd be 10 feet tall, or if if it's better to be smart, my brain would be like this this big or something. So you can't reason that naively about stuff like that.

• 29:04Dwarkesh Patel?

  I see. Yeah, OK, so I guess it could make sense, for example, with certain health risks, like the thing that makes you more likely to get diabetes or heart disease today might be, I don't know what the the plaitroch affect of that could be, but maybe that's not that important when you're not that obese.

• 29:17Speaker 01

  Let me just point out that that that most of the diseases that we care about now, most of them, not the rare ones, but the common ones, they manifest when you're like 50, 60, 70 years old. And there was never any evolutionary big advantage, I think, of being super long-lived, right? So there's even a debate about whether like, OK, if the grandparents are around to help raise the kids, that that raises the fitness a little bit of the family unit. But but. Most of the time in the past, and most of our evolutionary past, humans just died fairly early. And so a lot of these diseases would never have been optimized against evolutionarily. But we see them now because we live under such good conditions. We can, you know, people regularly approach 80 or 90 years.

• 29:58Dwarkesh Patel?

  Regarding the linearity and additivity point, I was gonna make the analogy, and I'm curious if this is valid. But when you're programming, one thing that's good practice is to have all the implementation details in separate function calls or separate programs or something, and then have your main, you know, loop of operations just be call different functions, like do this, do that, so that you can easily comment stuff away or change arguments. And this seemed very similar to that, where you have, just by turning these snips on and off, you can change what the next offering is gonna be. And you don't have to worry about, like, actually implementing the, whatever the underlying mechanism is.

• 30:41Speaker 01

  Well, what you said is related to what Fisher proved in his theorems, which is that, you know, if suddenly it becomes advantageous to have X, like white fur instead of black fur or something, it would be best if there were little levers that you could move somebody from black fur to white fur continuously by just modifying those switches in an additive way. It just turns out with, for sexually reproducing species where the DNA gets scrambled up in every generation, it's better to have switches of that kind. And so the other point related to your software analogy is that there are, there seem to be modular,

• 31:23Speaker 01

  fairly modular things going on in the genome. So when we looked at, we were the first group to, I think we had, like initially, like say 20 major disease conditions we had decent predictors for. And we just started looking carefully at the Just something as trivial as the overlap of, you know, my sparsely trained predictor turns on, uses these features for diabetes, but it uses these features for schizophrenia. And how much overlap? Just the stupidest metric is like how much overlap or variance accounted for overlap is there between pairs of disease conditions? And it's very modest. It's actually the opposite of what naive biologists would say when they talk about pleiotropy, or they're just disjointed.

• 32:08Speaker 01

  They're just disjointed regions of your genome that are governing certain things. And so why not? You have 3 billion base pairs. There's a lot you can do in there. There's a lot of information in there. So you can have, if you need 1,000 to control diabetes risk, I can have, I think I estimated you can easily have 1,000 roughly independent traits that are just disjoint in their genetic dependencies. And so if you think about like D and D, like your strength and your dex and your wisdom and your intelligence and charisma, those are all disjointed. They're all just independent variables. So it's like a seven dimensional space that your character lives in. Well, there's enough information in the few million differences between me and you.

• 32:51Speaker 01

  There's enough for a thousand dimensional space of variation. Like, oh, how big is your spleen? My spleen's a little bit smaller, yours is a little bit bigger. That can vary independently of your IQ. Oh, it's a big surprise. The size of your spleen can vary independently of the size of your big toe. Oh, yeah, yeah, there's about 1,000. If you just do information theory, there's about 1,000 different parameters I can vary independently with the number of variants that I have between me and you. So, and this thing, because you understand some information theory, is kind of trivial to explain, but try to explain to a biologist, you won't get very far.

• 33:27Dwarkesh Patel?

  Yeah, yeah. Do the log two of the number of, is that basically how you do it? Yeah, okay.

• 33:33Speaker 01

  That's all it is. I mean, well, I mean, well, it's in our, it's in our paper. Like we basically look at, okay, how many, how many variants are typically accounting for most of the variation for any of these major traits? And then imagine that they're mostly disjointed. Well, just how much length of DN, how many variants do you need then to independently vary 1,000 traits? Well, it's a few million. Differences between me and you are enough, right?

• 33:59Dwarkesh Patel?

  Right.

• 34:01Speaker 01

  It's trivial, it's very trivial math. Once you understand the base, how to reason about information theory, then it's very trivial. But it ain't trivial for theoretical biologists, as far as I can tell.

• 34:13Dwarkesh Patel?

  But the result is so interesting because I remember reading in the selfish gene that like he hypothesizes the reason we have aging, or one of the possible reasons we have aging, is that there's antagonistic cryotropy. There's, there's something that makes you healthier when you're young and fertile, that makes you unhealthy when you're old. And evolution would have selected for such a trade off because when you're young and fertile is when evolution and your genes care about you. And so, but if there's enough space in the genome before you, where these trade offs are not necessary, then this may be like a bad explanation for aging? Or do you think I'm straining the analogy?

• 34:49Speaker 01

  No, no, you're, it's, I love your interviews because the point you're making here is really good. So Dawkins, who is a kind of evolutionary theorist, but from the old school, when they had almost no data, okay, to deal, you know, you can imagine how much data they had compared to today. He would like to tell you a story about a particular gene that maybe it has this positive effect when you're young, but it makes you age faster. So there's a trade off. And, you know, we know about things like sickle cell anemia and, you know, we know stories like that. And no doubt there are stories like that which are true about specific variants in your genome. But that's not the general story. The general story, which we only discovered in the last five years, is that almost every

• 35:34Speaker 01

  trait is controlled by thousands of variants, and those variants tend to be disjoint from the ones that control the other trait. So they weren't wrong, but they didn't have the big picture.

• 35:45Dwarkesh Patel?

  Yeah, I see. And, and then, yeah, so you had this paper, I think it was polygenic health and extra general health and disease risk. And then you showed that with 10 embryos, you could increase disability adjusted life years by four, which is like a huge increase of like, if you think about like, if you could just live four years longer in a healthy state.

• 36:04Speaker 01

  Yeah, what's the value of that? What would you pay to buy that for your kid?

• 36:08Dwarkesh Patel?

  Right, yeah. But I don't know, this seems like, going back to that earlier question about the trade-offs, or like about why this hasn't already been selected for. If you're right, and there's no like trade-off to do this, just living four years older, even if that's on your fertility, just like being a grandpa or something, that seems like an unmitigated good. So why, it's kind of mysterious that that hasn't already been, you know, selected for.

• 36:33Speaker 01

  So, no, I'm glad you're really asking about these questions, because these are things that people are very confused about, you know, even in the field. So, first of all, let me say, if you have a trait that's controlled by 10,000 variants, okay, like height is controlled by order 10,000 variants, and probably cognitive ability a little bit more. The square root of 10,000 is 100, okay? So if I could come to this little embryo, and I said, "I want to give it one extra standard deviation of height, plus one standard deviation of height," I only need to edit 100. I only need to flip 100 minus variants to plus variants. These are very rough numbers, but you know, one standard deviation is like

• 37:18Speaker 01

  the square root of n, right? If I flip a coin n times, and I want a better outcome in terms of number ratio of heads to tails, and I want to increase it by one standard deviation, I only need to flip square root of n heads, because if you flip a lot, you're going to get a very narrow distribution peaked around half, and the width of that distribution is square root of n, okay? So, once I tell you, "Hey, your, you know, height is controlled by 10,000 variants, and I only need to flip 100 genetic variants to make you one standard deviation for a male, that would be three inches tall, two and a half for three inches taller," suddenly you realize, "Wait a minute, there's a lot of variants up for grabs there." I mean, if I could flip 500 variants in your genome, I would make

• 38:03Speaker 01

  you five standard deviations taller. You'd be like seven feet tall, and I didn't have to do that much work. And there's a lot more variation where that came from, okay? Because I only flipped 500 out of 10,000, I could have flipped even more, right? So, there's this kind of quasi-infinite well of variation.

• 38:22Speaker 01

  Which evolution or genetic engineers could act on? And again, the early population geneticists who breed corn, who breed animals, they know this. This is actually something they explicitly know about because they've done calculations. Now, interestingly, the human geneticists who are mainly concerned with diseases and stuff are often not familiar with what the math that the animal breeders already know. And you might be interested to know that like the milk you drink is comes from heavily genetically optimized cows who are who are actually bred artificially using and they're and they're using almost exactly the same technologies that we use at genomic prediction, but they're doing it to optimize like milk production and stuff like this. So there is a big well of variance.

• 39:06Speaker 01

  It's a consequence of this super multi polygenicity of the trait. And it does look like people could, coming back to your question about longevity, it does look like people could, quote, be engineered to live much longer than they currently do.

• 39:28Speaker 01

  By just say flipping the variants that make that reduce risk for individual diseases that tend to shorten your life. And then the question is back to, well, why didn't evolution give us lifespans of a thousand years? Like back in the Bible, people in the Bible used to live for a thousand years. Why don't we? I mean, that probably didn't really happen. But the question is, you have this very high dimensional space and you have a fitness function. And how big is the slope in a particular direction of that fitness function? Like, like, how much more successful reproductively would Joe have been? Joe Caveman have been if he lived to be 150 instead of only, you know, 100 or something?

• 40:11Speaker 01

  And there just hasn't been enough time to, you know, explore this super high dimensional space. That's the that's the actual answer. But now we have the technology. We're going to fucking explore it fast now. That's the that's the point that people, you know, the big light bulbs should go off. Like, no, we we're mapping this space out now.

• 40:34Speaker 01

  Pretty confident in 10 years or so that CRISPR gene editing technologies will be ready for massively multiplexed edits. And we're going to start navigating in this high dimensional space as we like. So that's the, that's the more long-term consequence of these scientific insights.

• 40:53Dwarkesh Patel?

  Yeah, that's super interesting. And what do you think will be the plateau for a trait like, you know, how long you live? Well, is it, because four is, I guess, with the current data and techniques. You think it could be significantly greater than that or?

• 41:05Speaker 01

  Well, we did a very simple calculation, which, which amazing that it gives the kind of the right result. We said like, given this polygenic predictor that we built, which isn't perfect. I mean, we're going to, it's going to improve a lot as we get more data. Given this polygenic predictor for overall health, uh, which is used in selecting embryos today. If you just say like, well, out of a billion people, what's the best person typically? What would their score be on this index? And then how long would they be predicted to live? It's about 120 years. So it's actually spot on. It's basically that's, that's one in a billion type person lives to be like 120 years old or so roughly.

• 41:47Speaker 01

  How much better can you do? Probably a lot better. Probably, I mean, I don't want to speculate, but because other effects, nonlinear effects, things that we're not taking into account will start to play a role at some point. So it's a little bit hard to estimate what the limiting, what the true limiting factors will be. But the one statement which is super robust and I'll stand by it. I'll debate any Nobel laureate in biology or whatever wants to talk about it. Um, there's clearly a lot of variants available to be selected on or edited. There's, that's just, there's no question about that. And that's been, that's been established in animal breeding and plant breeding for, you know, a long time now. So we can, you, if you want a chicken that grows to be this big instead of this big, you can do it.

• 42:29Speaker 01

  If you want a cow that produces literally 10 times or a hundred times more milk than a regular cow, you can do it. The egg you ate for breakfast this morning? Those bioengineered chickens, they lay almost an egg a day. A chicken in the wild lays like an egg a month. How the hell did we do that? By genetic engineering, that's how we did it. So.

• 42:52Dwarkesh Patel?

  Yeah, yeah. And that was just brute, uh, brute, our artificial selection and the, the no fancy machine learning there.

• 42:58Speaker 01

  Uh, last 10 years, last 10 years, it's gotten sophisticated machine learning, genotyping of chickens. Um, artificial insemination, um, modeling of the traits using ML. Last 10 years has, it's basically for cow breeding now, it's totally, totally done by ML now.

• 43:19Dwarkesh Patel?

  I, I had no idea. That's super interesting. Um, what is, uh, so you mentioned that you're accumulating, um, data and improving your techniques over time. Is there a first mover advantage to, uh, you know, genomic prediction company like this? Or is it, is it just whoever has the newest, best, uh, algorithm for going through the biobank data?

• 43:40Speaker 01

  Yeah, I, it's, that's another, it's, it's another super question. So for your entrepreneurs in your audience, I would say in the short run, if you ask like, oh, what, how's, you know, what's, what the, what should the valuation of GP be? You know, that's, that's how the venture guys would want me to answer the question. Um, there is a huge first mover advantage because very important is the channel relationships between us and the clinics. And nobody's going to be able to get in there very easily when they come later because we're, we're developing trust and a big track record with clinics all over the world. And we're well known. Could 23andme or some company that has a huge amount of data?

• 44:22Speaker 01

  And if they were to actually get better AI ML people working on this, um, could they kind of blow us away a little bit and build better predictors? Cause they have just much more data than we do. Possibly, yes. Um, now there's a core expertise that we have in doing this kind of work for years and years and years that we're just really good at it. And so even though we don't have as much data as 23andme, we might still, our predictors are better than theirs right now. Um, and I'm out there all the time, working with biobanks all around the world, like in countries like, I don't want to say all the names, but other countries trying to get my hands on as much data as I can. So, um, but there, there may not be a lasting.

• 45:04Speaker 01

  You know, advantage beyond the actual business channel connections to that particular market. It may not be a defensible, purely scientific moat around the company. We do have patents on specific technologies about how to do the genotyping or how to do error correction on embryo DNA and stuff like this. We do have patents on stuff like that, but this general idea of like, who's going to be the best at predicting human traits from DNA? It's unclear who's going to be the winner in that race. Maybe it'll be the Chinese government in 50 years. Who knows?

• 45:37Dwarkesh Patel?

  Yeah, that's interesting. I mean, if you think about like a company like Google, like theoretically, it's possible you could come up with a better algorithm than PageRank and then beat them. But it just seems like probably the engineer at Google is going to be the one that comes up with whatever edge case or whatever improvement is possible.

• 45:52Speaker 01

  That's exactly, see, it's exactly what I would say. I would say like, yeah, maybe, I mean, PageRank actually by now is totally deprecated. But, but even if somebody else comes up with a somewhat better algorithm or somewhat better, maybe they have a little bit more data. If you have a team that's been doing this for a long time and you're really focused and good, it's still tough to beat you, especially if you have a lead in the market.

• 46:14Dwarkesh Patel?

  Yeah, yeah. And then, so what layer of the stack do you, so you guys are, are you guys doing the actual biopsy or is it just that they upload the genome and you're the one processing and just giving recommendations? Is it, is it just like an API call basically or?

• 46:27Speaker 01

  So it's great. I love your question. So it is totally standard. Every good IVF clinic in the world regularly takes embryo biopsies. So that's totally standard. It's like a lab tech doing that. Okay. And then what happens is they take the little sample and they put it on ice and they just ship it. And the DNA as a molecule is extremely robust and stable. In fact, my other startup solves crimes that are 100 years old from DNA that we get from some semen stain on some rape victim, you know, victim, serial killer victim's bra strap. We've done stuff like that.

• 47:05Dwarkesh Patel?

  Jack the Ripper, when are we going to solve that mystery?

• 47:08Speaker 01

  If they can give me samples, we can get into that.

• 47:13Speaker 01

  For example, we just learned that you can recover DNA pretty well from the like if someone licks a stamp and puts it on their correspondence. So that, I mean, if you can do Neanderthals, you can do a lot for solving crimes. So in the IVF workflow, our lab, which is in New Jersey, can service every clinic in the world because they just, they just take the biopsy, they put it in a standard shipping container, and they just send it to us. And then we, we actually genotype the DNA in our lab, but we've actually trained a few of the bigger clinics to actually do the genotyping on their site. And at that point, it's just like they upload some data into the cloud and then they get back some

• 47:54Speaker 01

  stuff from our platform. So at that point, where it's going to be the whole world, man, every human, every human who wants their kid to, you know, be healthy and, you know, get the best they can, it's gonna, that data's gonna come up to us and it's gonna, the report is gonna come back down to their IVF physician.

• 48:10Dwarkesh Patel?

  Right, yeah. Which, which is great if you think, well, like, let's say you think there's a potential that this technology might get regulated in some way. You could just have, like, just go to Mexico or something, have them upload the, have them upload the genome, you know, you don't care where they uploaded from, and then, then you can just get the recommendations there.

• 48:29Speaker 01

  Yeah, I think we're gonna evolve to a point where, because the, the genotyping technology is getting better and better, eventually, we are gonna be out of the wet part of this business and only in the bit and cloud part of this business. Because especially the clinic, no matter where it is, they're gonna have a little sequencer, which is this big, and their tech is gonna do it, and then they're just gonna hit upload, and then they get the report back like three seconds later from us, for the physician to look at, and the parents can look at it on their phone or whatever. Actually, we're basically there, actually, with some clinics. So, yeah, it's gonna be tough to regulate because it's just bits, right? So you have the bits, and you're in some repressive, terrible

• 49:12Speaker 01

  country, you know, that doesn't allow you to select for some special traits that people are nervous about, but you just upload it to some vendor who's in Singapore or in, you know, some free country, and, and they, they give you the report back. Doesn't have to be us. I mean, we're not, we don't do those, we don't do the edgy stuff. We only do the health-related stuff right now. But if you want to know how tall this embryo is gonna be, I'll tell you a mind blower.

• 49:40Speaker 01

  When you do face recognition in AI, you're basically mapping someone's face into a parameter space of like, I think it's on the order of hundreds of parameters, right?

• 49:52Speaker 01

  Each of those parameters is super heritable. So in other words, if I take two twins and I I measure their, I photograph them and the the algorithm gives me the value of that parameter for twin one and twin two. They're very close, obviously. That's why I can't tell the two twins apart. The the face recognition can ultimately tell the twins apart. They're really good face recognition. But you can you can just conclude that almost all these parameters are the same for those twins. So it's highly heritable.

• 50:17Speaker 01

  So we're gonna get to a point soon where I can do the inverse problem, where I have your DNA and I predict each of those parameters in the face recognition algorithm, and then from that I reconstruct the face. So I say like this embryo, when she's 16, this is what she's gonna look like. When she's 32, this is what she's gonna look like. And I I'll be able to do that for sure. It's just a data, you know, it's just an A I M L problem right now. But but the the basic biology is clearly gonna work. So um, so then you're gonna be able to say like, oh, look, here's a report. Look, let's embryo four is so cute. Why don't we, you know?

• 50:57Dwarkesh Patel?

  But before we get married.

• 50:58Speaker 01

  Now, we we don't do that, but it's gonna be possible.

• 51:02Dwarkesh Patel?

  All right. Before you get married, you'll wanna see what they what their genotype implies about their faces longevity. Yeah, it's interesting. You hear stories about these cartel leaders who will, you know, get plastic surgery or something to evade evade the law. You could just have a check where you're like lick this lick this uh lick this lab. And then, yeah, does that match the face that you would have had five years ago when we caught you on tape?

• 51:26Speaker 01

  Yeah, well, and also, you don't, it's this little bit back to old school Gattaca, but you don't even need the face. You can just say like, I'm gonna take a few molecules of, you know, skin cells from you. I mean, take a few skin cells off of you and just unotype you. I know exactly who you are. So I've had conversations with the spooky intel folks.

• 51:48Speaker 01

  About, you know, they're very interested in like, oh, if some Russian diplomat comes in, and we think he's actually a spy, but he's, you know, with the embassy over there, and he has a, he has a coffee with me, and I save the cup and send it to my brother, my buddy at Langley. Can we figure out who this guy is and that he has a daughter who's going to Choat? You know, can do all that now.

• 52:12Dwarkesh Patel?

  Oh, that's, I, I, I, it seems to me, if that's true, then like in the future, people will be so concerned, like world leaders or something, when they're visiting a foreign country, they're not gonna wanna eat anything, drink it, like, they'll be wearing like a hazmat suit to make sure they don't like lose a hair follicle.

• 52:28Speaker 01

  Yeah, the new, the next time Pelosi goes, she's gonna be in like a space suit. If she cares, or, or the other thing is they're just gonna give in. They're just gonna be like, yeah, my DNA is everywhere. If I'm a public figure, my DNA, I can't track it. It's all over.

• 52:42Dwarkesh Patel?

  Yeah, but the thing is, like, there's so much, as you, I'm sure you know, there's like speculation that Putin might have cancer or something. If we have his DNA, we can just see like, oh, actually, like, his probability of having cancer at age 70, or whatever he is, is, you know, 85%. So, yeah, that, that's like a verified rumor, you know, that would be interesting.

• 52:57Speaker 01

  I don't think that would be very definitive. Like, I don't think we'll reach that point where you could say, yeah, Putin definitely has cancer because of his DNA, which I could have known when he was an embryo. That, I don't think it's gonna reach that level, but we could say, yeah, he, he is high risk for this kind of cancer. You know, you could say that.

• 53:13Dwarkesh Patel?

  Yeah, yeah, yeah. So, if, if this like reaches, if, if like, let's say in 50 years or 100 years, all the majority of the population is doing this, and if that means that the diseases that are highly heritable get pruned out of the population, does that mean we'll only be left with the diseases that are like lifestyle diseases? So, you won't get like breast cancer anymore, but you will still get, you know, you'll still get like fat, or, I don't know, whatever, lung cancer from smoking or something.

• 53:42Speaker 01

  I think it's hard to discuss the asymptotic limit of what's going to happen here. I'm not, I'm not very confident about making predictions like that. You know, it, it could be that we'll get to the point where everybody is, well, everybody who's rich or has been through this stuff for a while, especially if we get the editing working, is super low risk for all the, like, the top 20 killers, the diseases which, you know, have most life expectancy impact. And yeah, maybe those people live to be 300 years old naturally. I don't think that's excluded at all. So I think that's within the realm of possibility.

• 54:24Speaker 01

  But, you know, it's going to happen for a few lucky, you know, Elon Musk-like people before it happens for schlubs like you and me. So, you know, there's going to be very angry inequality protesters about, you know, the Trump grandchildren who models predict will live to be 200 years old.

• 54:43Dwarkesh Patel?

  Right.

• 54:44Speaker 01

  You know?

• 54:45Dwarkesh Patel?

  Yeah, yeah, yeah.

• 54:46Speaker 01

  People are not going to be happy about that.

• 54:47Dwarkesh Patel?

  Yeah, that's so interesting. And, okay, so one way to think about these different embryos is, like, if you're going to produce multiple embryos, you can get to select from one of them. Each of them is like, each of them is like a call option, right? And therefore, you probably want to optimize for volatility as much, or if not more than just the expected value of the trait. And so I'm wondering if there's mechanisms where you can, I don't know, like, increase the volatility in myosis or in some other process. So you, you just got a higher variance, you can just select from the tail better.

• 55:20Speaker 01

  Well, I'll, I'll tell you something related to that, which is quite amusing. So I had conversations with some pretty senior people at the company that owns all the dating apps. So you can look up, you can figure out what company this is, but they own Tinder and Match and stuff like this. And they're kind of interesting, interested in, wow, what if we have a special feature where instead of Tinder Gold or Platinum, you upload your genome and you match, we talk about how well you match the other person based on your genome. Actually, one person told me something which was really shocking is that apparently guys lie about their height on these apps.

• 56:04Speaker 01

  And if you could have a DNA verified-

• 56:06Dwarkesh Patel?

  Truly shocked, truly shocked.

• 56:08Speaker 01

  And if you could have a, if you could have a DNA verified height on, because our accuracy is like an inch or something.

• 56:14Dwarkesh Patel?

  Right.

• 56:15Speaker 01

  So it would prevent like really gross distortions, like someone claims they're 6'2" and they're actually 5'9". Probably the DNA could say that's unlikely, actually. But, but no, the, the, the application to what you were discussing is more like,

• 56:32Speaker 01

  let's suppose that we're selecting on intelligence or something. And let's suppose that the regions where your girlfriend has all the plus stuff

• 56:44Dwarkesh Patel?

  Ooh.

• 56:45Speaker 01

  is complementary

• 56:46Dwarkesh Patel?

  Yes, yes.

• 56:47Speaker 01

  to the regions where you have your plus stuff.

• 56:49Dwarkesh Patel?

  Ah.

• 56:50Speaker 01

  So we, we could model that and said like, your kids, just because of that, you know, the, the, the complementarity of the structure of your genome in the regions that affect intelligence, you're very likely to have some super smart kids way above your, the mean of your, you and your girlfriend's values. So you could actually say things like, yeah, it's better for you to marry that girl than that girl. You know, if you're gonna go, as long as you're gonna go through embryo selection, we can throw out the outlier, the bad outliers.

• 57:20Dwarkesh Patel?

  That is so fascinating.

• 57:21Speaker 01

  All that's technically feasible. And I think actually it's true that one of the earliest patent applications, they'll, they'll all deny it now. Uh, what's her name? Uh, gosh, I can't remember. The, the CEO of 23andMe, um, Wojcicki.

• 57:37Dwarkesh Patel?

  Yeah.

• 57:38Speaker 01

  Uh, she'll deny it now, but I think if you look in the patent database, the very, one of the very earliest patents that 23andMe filed when they were like still a small startup was about exactly this, is like advising- Parents about mating and how their kids would turn out and stuff like this. So, you know, we don't even go that far in GP. We don't even talk about stuff like that. But they were thinking about it when they founded 23andMe.

• 58:03Dwarkesh Patel?

  That is unbelievably interesting. And by the way, speaking of height, this just, this just occurred to me, but you know, it's like supposed to be highly heritable, but especially people in like Asian countries, who we have the experience of like having grandparents that are much shorter than us and then parents that are shorter than us, which suggests that like the environment has like a big part to play in it, just like malnutrition or something. Yeah, so how do you square that, the fact that like often our parents are shorter than us with the idea that like the height is supposed to be super heritable?

• 58:34Speaker 01

  Another great observation. So the correct, the real correct scientific statement is we can predict height for people who are born, who will be born and raised in a favorable environment. So in other words, if you live close to a McDonald's and you're not, you know, you can afford all the food you want, then the height prediction, the height phenotype becomes super heritable because the environmental variation now doesn't matter very much. But you and I both know, if we go back to where our ancestors came from, people are a lot smaller. And also if you look at how much food, how many calories and how much protein and calcium they eat, it's totally different than what I ate and what you ate growing up.

• 59:17Speaker 01

  So we're not saying, we're never saying the environmental effects are zero. We're saying for people raised in a certain very favorable environment, maybe the genes are a cap on what can be achieved and we can estimate, you know, we can predict that.

• 59:35Dwarkesh Patel?

  Yeah.

• 59:35Speaker 01

  So in fact, in our data, actually, we have, like I have data from Asia where yeah, you can see there clearly are big, much bigger environmental effects, age effects, actually, just older people for fixed polygenic score on the trait are much shorter than younger people.

• 59:55Dwarkesh Patel?

  Oh, okay, interesting. Yeah, that actually raises the next question I was about to ask, which was, how applicable are these scores across, you know, different ancestral populations?

• 1:00:08Speaker 01

  Huge, huge problem right now because most of the data is from Europeans. And what happens is that as you, if you train a predictor in this ancestry group, and you go to a more distant ancestry group, there's a falloff in the quality of prediction. And this is, again, this is like frontier question, so we don't know the answer for sure. But most people believe, or many people believe that what happens is that there's a certain correlational structure in each population where if I know the state of this SNP, I can predict the state of these neighboring SNPs. And that, that is a product of the mating patterns and the ancestry, you know, of that group.

• 1:00:49Speaker 01

  And sometimes the predictor, which is just using statistical power to figure things out, will grab one of these SNPs as a tag for the truly causal SNP that's in there. It doesn't really know which one is truly causal, it's just grabbing a tag. But the tagging quality falls off if you then go to another population. Like, this was a very good tag for the truly causal SNP in the British population, but it's not so good a tag in the South Asian population for the truly causal SNP, which we hypothesize is the same. It's the same underlying genetic architecture in these different ancestry groups. We don't know, that's a hypothesis, but even so, the tagging quality falls off. So, my group, you know, we've spent a lot of our time looking at

• 1:01:34Speaker 01

  performance of predictor trained in population A on distant population B, and doing all this stuff, modeling it, trying to figure out, trying to test hypotheses as to whether it's just the tagging decay which is responsible for most of the falloff. So, all of this is an area of very active investigation. I think, you know, it'll probably be solved in five years. The first really big biobanks that are non-European are coming online. And so, we're, I think we're going to solve it, you know, in some number of years.

• 1:02:02Dwarkesh Patel?

  Oh, what does the solution look like, I guess? Because if you don't know, unless you can identify like the causal mechanism by which each SNP is having an effect, how can you know that something is a tag or whether it's the actual underlying, you know, switch?

• 1:02:18Speaker 01

  The resolution will be, again, the nature of reality determines how this is gonna go. So, and we don't know the underlying biology. If it's true, and this is the amazing thing, like, people argue about, like, human biodiversity and all this stuff, and we don't even know whether the specific mechanisms that, say, predispose you to being tall or to having heart disease are the same in these different ancestry groups. We assume that it is, but we don't know that. And, you know, like, as we get further away, like, to Neanderthals or Homo erectus, you might be like, "Yeah, they have a slightly different architecture there than we do."

• 1:02:59Speaker 01

  But let's assume that the causal structure is the same for South Asians and for British people, okay? And then it's a matter of improving the tags. And you might say, "Wait, wait, wait a minute, Steve. How do I know? How do I know? If I don't know which one is causal, what do you mean by improving the tags?" This is a machine learning problem. So, the question is, if there's a SNP, which, when I use it across multiple ancestry groups, is always coming up as very significant, maybe that one's truly causal. As I vary the tagging correlations in the neighborhood of that SNP, I always find that that one is in the intersection, the intersection of all these different sets.

• 1:03:42Speaker 01

  That makes me think that one's gonna actually be causal. So, that's a process we're engaged in now, is to try to basically do that. It's basically just a machine learning problem, but we need data. That's the main issue.

• 1:03:56Dwarkesh Patel?

  Yeah, I was kind of hoping that wouldn't be possible, because one worry you might have about this research is that, you know, like, it'll in itself become taboo or cause other sorts of bad social consequences if you can, like, definitively show that on certain traits, there's differences between ancestral populations, right? And so, I was kind of hoping that maybe there's, like, just an evasion button where, like, "Yeah, we can't say, because they're just tags, and the tags might be different between different ancestral populations." But I guess what better machine learning will know?

• 1:04:24Speaker 01

  That's the situation we're in now, where you have to do some fancy analysis. If you want to claim, like... Italians literally have lower height potential than Nordics, which is possible. And there's been a ton of research about this because there's signals of selection. It looks like the the the alleles which are activated in height predictors. Looks like they looks like they've been under some some selection between North and South Europe over the last 5,000 years. For whatever reason, we don't know the reason, but there are some. But this is a this is a thing which is debated by people who study molecular evolution. But suppose it's true, OK? And then what that would mean is that when we finally get to the bottom of it and we find

• 1:05:09Speaker 01

  all the causal low side for height, literally, the average value for the Italians is lower than the average value for the people living in Stockholm. And that might be true. People don't get that excited. They get a little bit excited about height, but they would get really excited if this were true for some other traits, right? Suppose like your extraversion, you know, the causal variance affecting your level of extraversion is systematic, the average value of those weighted, the weighted average of those states is different in, you know, Japan versus Sicily. Mm hmm. Right? Yeah. People might freak out over that. Right.

• 1:05:53Speaker 01

  I'm supposed to just say that's obviously not true. It's obviously not true. It can't be true. How could it possibly be true? Because there hasn't been enough evolutionary time for those differences to arise. After all, it's not possible that despite what we what looks to be the case for height over the last 5,000 years in Europe, no other traits could possibly have been differentially selected for over the last 5,000 years. That's the really dangerous thing. So there there are few people who understand this field well enough.

• 1:06:24Speaker 01

  To understand what you and I just discussed and who are so alarmed by it that they're just trying to suppress everything. There are people like that. But most of them actually don't really follow it at this, the technical level that you and I are discussing it. So they're just like, kind of instinctively negative about it, but they don't, they don't really understand it very well.

• 1:06:43Dwarkesh Patel?

  That's good to hear, because I, that's, yeah, in a lot of other spaces, you see this pattern that by the time that somebody might want to regulate, or in some way interfere with some technology or some information, it already has achieved wide adoption. You could argue that that's the case with crypto today. But if it's true that like, a bunch of IVF clinics across the world are using this, using, you know, this, these scores to do selection and other things. Yeah, by the time that people realize the implications of this data for other kinds of social questions, by that time, this will already be like an actual consumer technology, hopefully.

• 1:07:22Speaker 01

  I think that, I think that's true. I think the main outcry will be if it turns out that there are really big gains to be had, and only the billionaires are getting them. But, but that might have the consequence of causing countries to make this free part of their national healthcare system. So Denmark, Israel, they pay for IVF for infertile couples. So it's, it's part of their national healthcare system. And they're pretty aggressive about genetic testing. In Denmark, one in 10 babies born now is born through IVF.

• 1:08:00Dwarkesh Patel?

  Right.

• 1:08:01Speaker 01

  So, yeah, so it's not clear how it's going to go. But, yeah, I mean, we're in for some fun times. There's no doubt about it.

• 1:08:10Dwarkesh Patel?

  Yeah, well, I guess one way it could go is, some countries decided to ban it altogether. And another way it could go is countries decide to give everybody free access to it.

• 1:08:19Speaker 01

  Nationalize.

• 1:08:20Dwarkesh Patel?

  Yeah, exactly. If you had to choose between the two, I guess you would want to go for the second one, which I, which I guess would be the hope. And maybe only those two are compatible with people's, I don't know, their moral intuitions about this kind of stuff.

• 1:08:33Speaker 01

  It it's very funny because most wokest people today hate this stuff. But most progressives like Margaret Sanger or you know, anybody who was progressive, the intellectual, well, in some sense, the forebears of today's wokest. In the early 20th century, they were all what we would call today eugenicists because they were like, oh, shoot. We've thanks to Darwin, we now know how this all works and we should take steps to keep society healthy and not in a negative way where we kill people we don't like, but we should just help society do healthy things and when they reproduce and you know, have healthy kids.

• 1:09:09Dwarkesh Patel?

  Right.

• 1:09:09Speaker 01

  And so now there's this whole thing has just been flipped over among progressives. So.

• 1:09:14Dwarkesh Patel?

  Yeah.

• 1:09:15Speaker 01

  It's sad.

• 1:09:16Dwarkesh Patel?

  Yeah, even in India, like that was that was like very recently, less than 50 years ago or when Indira Gandhi, you know, she's like the left side of India's political spectrum. And yeah, she obviously she was infamous for putting on these like forced sterilization programs and yeah, so you know, I don't want to credit the person, but somebody made an interesting comment. They wouldn't want their name associated with this maybe, but somebody made an interesting comment about this where they said, they were asked like, oh, is it true that progressives in history, the history always tilts towards progressives and if so, aren't aren't isn't everybody else doomed? Aren't their views doomed? And the person made a really interesting point, which is that, yes, whatever we consider left at the time tends to be winning.

• 1:09:58Dwarkesh Patel?

  But what is left changes a lot over time. Right? So if in the early 20th century, prohibition was a left cause, right? It was a progressive cause and you know, that changed. Now that's no law. I mean, the opposite is a left cause.

• 1:10:10Speaker 01

  Yeah, now legalizing pot is progressive.

• 1:10:13Dwarkesh Patel?

  Exactly. So the way if the, you know, if conquest second law is true and everything just tilts left over time, just change what left is, right? That's a solution.

• 1:10:21Speaker 01

  Yeah, absolutely. I mean, the I, I, I, of course, one can't demand that any of these woke guys be like intellectually self consistent or or even like say the same things from one year to another. But if one could, you know, you wonder what they think about these literally communist Chinese. I mean, these are literally communists. They're recycling huge parts of their GDP to help the poor and do all this other stuff that, you know, medicine is free, everything, you know, education is free, right? They're literally socialists. They're literally communists. But in Chinese, the Chinese characters for eugenics is a totally positive thing. It's just like healthy production. It means healthy re- well, that's actually what it means in Greek too, but more or less.

• 1:11:06Speaker 01

  But but the the whole viewpoint on all this stuff is like one hundred and eighty degrees off in in East Asia compared to here. And even even in the among the literal communists, you know, so.

• 1:11:18Dwarkesh Patel?

  Um, so let's talk about one of the traits that people might be interested in potentially selecting for, which is intelligence. Um, do are do we, what is the potential that we'll be able to actually acquire the data to be able to um correlate the genotype with intelligence?

• 1:11:37Speaker 01

  Well, that that's the most personally frustrating aspect of all this stuff. Like, if you ask me, like, ten years ago, when I started doing this stuff, what did I think we were gonna get? I think everything is gone, kind of on the optimistic side of what I would have predicted. So everything's good, you know? Didn't turn out to be interactively nonlinear. It didn't turn out to be interactively poly uh pleiotropic. You know, all these good things, which nobody could have known a priori how they would work, turned out to be good for gene engineers of the twenty first century.

• 1:12:11Speaker 01

  The one thing that's frustrating is because of crazy wokeism and fear of crazy wokeists, the most interesting, what I consider the most interesting phenotype of all, is lagging because everybody's afraid, even though there are very good reasons for medical researchers to want to know the cognitive ability of people in their studies. For example, when you want to study aging or decline of cognitive function, memory in older people, it's you want to have baseline measurements of how good their cognitive function was when they were younger, right? So very good reasons for why you want to have all this data. But researchers are afraid because it's also linked to all these controversial social issues.

• 1:12:56Speaker 01

  And so the amount, there's just a ginormous amount of genomic data where there's actually no cognitive measurement attached as a field to that data, which would have been very cheap to measure. Again, wokeists hate this, but I can measure your IQ on like a 12-minute test, no problem. Right? I mean, not with perfect accuracy, but I can get a pretty, I can get a very useful measurement. If I just take, like, the NFL has this thing called the Wonderlic, which every player that's being considered for the draft is asked to take this Wonderlic. You can go back and look at the Wonderlic scores of every NFL player. It's a short test, it's like 12 minutes long or something. And it's, it's pretty highly correlated, it's like probably correlates point eight

• 1:13:38Speaker 01

  or point nine, point eight maybe with a, with a more fulsome IQ measurement. So it would be trivial and inexpensive to gather this data, and then once we have, my prediction from this earlier math that I was talking about, is that when you get to have order a million, it could be one million, it could be two million, well phenotyped people and genomes, we would be able to build a pretty decent IQ predictor that might have a standard error of maybe 10 points or something. So that would be incredibly, for science, just, you know, unlimited interesting stuff in there, but not getting done.

• 1:14:18Dwarkesh Patel?

  Yeah, and if there are differences between, I mean, differences in how things are tagged between different ancestral groups, I'm not talking about like average differences or anything, just how the genotype is tagged. And if the Chinese do this first, then that's like, they have an advantage that can't be transferred over, I guess, right? Because it's only applicable, or advantageously applicable to their population.

• 1:14:46Speaker 01

  That's a great point. You can easily imagine, even in a small country like Singapore or Taiwan, has enough data to do this, no problem. Estonia. And they could do it and have this thing working and just not share it with anybody. So it's certainly possible. Now, that's a little, it's a little bit too science fictiony because the leaders who run these countries are not transhumanist rationalists, people who read, you know, your blog, my blog posts on the internet, they're not.

• 1:15:16Dwarkesh Patel?

  They're not dominant coming.

• 1:15:17Speaker 01

  So I don't think anything that exciting is gonna happen, but maybe it will.

• 1:15:21Dwarkesh Patel?

  Yeah. And do you think the potential for pleiotropy is higher with intelligence? I mean, with certain populations, oh, of course, by the way, disclaimer, 5,000 years is not enough, blah, blah, blah. But given that.

• 1:15:34Speaker 01

  Obviously.

• 1:15:35Dwarkesh Patel?

  Obviously.

• 1:15:36Speaker 01

  Obviously.

• 1:15:37Dwarkesh Patel?

  But given that you see with certain populations like Ashkenazi Jews, you have a higher incidence of, is it nervous system disorders, you know, like taste acts and other things. And that seems potentially to be the trade off of, you know, the higher average intelligence. You think that maybe the pleiotropy has a higher chance of occurring with intelligence?

• 1:15:59Speaker 01

  It can only be speculation, okay, at this stage. Now, with the history of the Ashkenazi Jews, they also went through some very narrow population bottlenecks. So there's some special aspects of their genetics. And whether it's related to cognitive function or not, you know, we don't really know for sure. But there are lots of reasons why they have fairly high proportion of inherited diseases and things like that that they're dealing with. This is one of the reasons why Israel is so progressive when it comes to genetic screening and IVF and things like this. One thing people talk a lot about is schizophrenia. So they say like, oh, schizophrenia could be correlated with creativity. So if your brother's schizophrenic, maybe you're more

• 1:16:42Speaker 01

  likely to be creative. And he's super creative, but we don't know what he's talking about.

• 1:16:48Speaker 01

  And so people say like, oh, if you start screening against schizophrenia, maybe we'll, we won't get creative geniuses. So there's all kinds of polytropic things that are possibly true. But the thing I keep wanting going, I want to go back to this is that if it's 10 or 20,000 different genetic variants, locations in your genome that are more or less determining your genetic cognitive potential, I can go around.

• 1:17:15Dwarkesh Patel?

  Ah, yes, yes, yes.

• 1:17:16Speaker 01

  It's a high dimensional space. If I find out this little cluster, okay, you can make someone smart in this little, using this stuff in this cluster, but it makes them dull, or it makes them autistic, or it makes them, they don't have big muscles. Like, okay, I'll just go around. I don't need to use those. I have plenty more. Look over here. I, those 500 I don't need to use. I will use these 500. And this is why it's important to look at historical geniuses who were pretty normal. And maybe they were even good athletes, and maybe they even were good with the ladies. Okay, these people existed. So you have these existence proofs that I can, if I need to, if I'm a really good genetic engineer and I

• 1:18:01Speaker 01

  can operate in this 10,000 dimensional space, whatever obstacle you put for me, I will just drive around it. And I just need some good data. I need lots of data. I need lots of AI, ML, and I'll do it. And that's the answer, which, again, most people don't really get this, but it's true.

• 1:18:18Dwarkesh Patel?

  Right, yeah. So I mean, there's a thing where if two traits are correlated at the ends, you know, the person who is like, for example, the smartest will not necessarily be the person who is, who is a strong, I guess these aren't necessarily correlated. But like, the person who has like the highest mathematical ability will not be the person who has the highest verbal ability, even though the two are correlated. And at some point, it'll be interesting, because parents will have to make that trade off, even if two things are extraordinarily correlated. And it'll be interesting to see how they make that.

• 1:18:43Speaker 01

  Eventually, you're really gonna have to trust your friendly neighborhood genetic engineer to advise you.

• 1:18:49Speaker 01

  You know, it's gonna be like a lot of modeling going on in the background.

• 1:18:52Dwarkesh Patel?

  Right. Now, I guess for the time being, we're stuck with educational attainment as a correlate. And that concerns me because educational attainment also probably correlates with other things that somebody might want or they might not want, which are conscientiousness and conformity, which is, you know, if you're Brian Kaplan, in the case against education, he says that the three things education signals are conscientiousness, conformity, and intelligence. You want the intelligence. Probably most parents actually do want conscientiousness and conformity, but some might not, right? So, yeah, is, could you, I guess, hopefully, we can get the direct intelligence data itself. But if we, if we don't, is there some way to segment out the conformity part of that educational attainment data?

• 1:19:34Speaker 01

  Well, well, here's the thing, like, in my dream world, like, if I were the CEO of 23andMe or something, what would I do? Oh, warning, they're actually secretly doing this, but you didn't hear that from me. I would have little surveys on the site that's like, oh, can you do a personality survey? And one of the categories will be conscientiousness and one will be extroversion. Right? And one will be like, conformity is not a traditional a big five thing, but you could have questions that kind of measure like how conformist someone is. And of course, we know how to do a little math, so we can, we can diagonalize the matrix of correlated measurements of all these different things.

• 1:20:14Speaker 01

  So I might be able to remove the chunk within EA, which is due to conformism, remove the chunk which is due to conscientiousness and leave behind the chunk which, oh, wow, that, and that correlates really highly with my separate IQ predictor, G predictor that I built separately using a different method. All these things are very, like, at our level, these are understood. The solutions, these problems are understood. It's just a data problem.

• 1:20:42Dwarkesh Patel?

  Yeah.

• 1:20:43Speaker 01

  Okay.

• 1:20:45Speaker 01

  I'll tell you an interesting thing. So we, my group was the first to do, there are 20,000 sibling pairs in the UK Biobank. So we were the first to say, you know, this is like three years ago or more. You know, some people don't really understand these polygenic scores and they're very skeptical. And they think, oh, we're capturing, we're not really capturing the real stuff, et cetera, et cetera. Well, you know what, I will just look to see how well we can predict which of the two brothers who experienced the same environment is gonna be taller. How well does my predictor do that?

• 1:21:20Speaker 01

  I'm gonna predict which of these two brothers has diabetes. Does a diabetes predictor really do that? And you're modding out all the environmental shit because they grew up in the same family, right? So, and we showed that the predictive power falloff, if you're trying to do this trick with unrelated pairs of people versus brothers who grew up in the same house or sisters, is minor. It's a small falloff in predictive power. So, so basically, we are getting the true genetic stuff, okay? One of the interesting things is when you look at EA,

• 1:21:54Speaker 01

  if you ask, I built an EA predictor, does it work better or worse when I try to predict which of the two brothers got more education? It turns out it works much worse because part of what that predictor is capturing is some maybe property of the parents who beat them and made them go to school. But both brothers got beaten and had diabetes. So, the reduction in quality of EA prediction for brothers is quite a bit higher than if you're just trying to predict G. So we have predictors we built that just predict G. And those have a much smaller reduction in quality

• 1:22:34Speaker 01

  when you apply them to brothers of SIBs than in unrelated pairs. And so I did, I went through that a little fast so people can go look up the paper. But the point is, we can see EA is a weird, is a very different trait than G from these kinds of results. That's super fascinating. And, and again, like people who criticize us have no idea how sophisticated the work is. They just don't, they don't read our papers. If they try to read our papers, they can't understand them. But we've done all this stuff. So it's now a guy who comes from a physics background or from an AI ML background. If I just start explaining to them, they're like, oh, yeah, OK, cool. You guys are doing that? Yeah, that's how it works out. Good, you can absorb it. But a lot of our critics just can't absorb it. It's literally a G. It's literally a G thing.

• 1:23:19Speaker 01

  They can't absorb it. So, um, but they just want to keep criticizing us forever. So, you know?

• 1:23:25Dwarkesh Patel?

  Yeah, yeah. The funny thing is, when I read your papers, I have a much easier time. Uh, like the prose part and the explanation and the organization is, uh, I don't know if it's your physics background or whatever. But I noticed with Scott Aaron's papers as well. It's like they're written like essays. They're so easy to, uh, as long as you understand the underlying ideas, they're so easy to absorb. Whereas if I just read like a random, uh, thing on bio archive, it just, I, I, I don't even know where to get started with this. It just written so turgidly.

• 1:23:52Speaker 01

  Uh, I'm totally with you. I mean, of course, there, there are multiple reasons for this, but one, one is that, yeah, maybe I'm a outsider. So I'm trying to write it very clearly and conceptually, maybe like a theoretical physicist would write it. But also it's like, it's a slightly selected population. Like Scott has a enormously populated blog, popular blog, and he writes these huge posts all the time. And I have a blog too. So, so we are a little bit better at expressing ourselves or clarifying ideas than the average scientist who's just trying to get, get the thing out and get it published in nature.

• 1:24:21Dwarkesh Patel?

  Yeah. Awesome. Okay, so let's talk a little bit about what consumers will actually want. Um, Goren has this really detailed, uh, post about embryo selection. Okay. And, um, he, he, he writes in it, uh, my belief is that the total uptake will be fairly modest as a fraction of the population. And he's talking about embryo selection here. A large fraction of the population expresses hostility towards any new fertility related technology whatsoever. And the people open to the possibility will be deterred by the necessity of advanced family planning, the large financial cost of IVF, and the fact that the IVF process is lengthy and painful. So, um, yeah, he seems like very pessimistic about the possibility that this is something that, you know, millions of people are using. Um, what, what do you think?

• 1:25:06Dwarkesh Patel?

  What's your reaction to his, uh, to his take here?

• 1:25:11Speaker 01

  There are two perspectives that you could adopt in looking at this. One is a perspective of a venture capitalist where you say, how big is this market? What's it worth to dominate this market? You know, what valuation should I accept from these pirates at GP?

• 1:25:32Speaker 01

  The other perspective is, hey, I'm really worried that humans are all gonna engineer themselves to be blonde and 6'4" and we're gonna be suddenly susceptible to all kinds of diseases and one single cold virus will kill all of us. So there's like two different perspectives on like what level of penetration this technology will have, right? There are two different perspectives. So from the venture guy's perspective, I will just say this, one out of 10 babies in Denmark is born this way. Would you like to capture a market that, you know, interfaces with one out of 10 families? And that's gonna grow, of course, right? One out of 10 families in all developed countries, maybe including China.

• 1:26:16Speaker 01

  You have the genome of mom and dad and the kid. And maybe you can sell them some health services later on. Maybe you can, maybe it's sticky, your relationship with these people. Okay, so that's for the venture guys, okay? You know how to get in touch with me.

• 1:26:34Speaker 01

  From the, oh, I'm really worried about human evolution. Or when are we gonna get another von Neumann? That's a different question. And it may be that it'll never be more than 10 or 20% of the population that's using IVF. And then through IVF, embryo selection and maybe potentially editing someday.

• 1:26:57Speaker 01

  So in that sense, why worry? There's always gonna be this natural reservoir of the wild type, you know, that have much more genetic diversity, et cetera, et cetera. So I think there's a very, maybe this is like the Goldilocks world. But imagine the Goldilocks world where... You know, there's plenty of wild type people and then there's plenty of people using these advanced technologies and everybody's happy, including our investors.

• 1:27:23Dwarkesh Patel?

  Yeah, something tells me that that will not be satisfying enough to the people who are concerned about. I have a sense that this whole argument about like the, oh, we're not going to have the evolutionary diversity or whatever, that's just a front for just like a moral reservation about this technology.

• 1:27:39Speaker 01

  Exactly. It's a front for people who just hate it. But what is Gorn saying? Like, is he saying that like, well, you know, these 10% of babies born in Denmark, they're already mostly screened for chromosomal abnormalities. And if I take that same data and I can generate this other report, are you really not going to look at that report? Are you, are you going to say like, well, you know, one of my, one of these kids is going to be super high risk for, you know, macular degeneration or some, you know, something, but I'm not gonna, I'm not gonna look at which one, but I'm already screening them for chromosomal abnormalities. Is that really going to happen? I don't think so. I think 10%, that 10% of the population that's using IVF is going to look at the report,

• 1:28:22Speaker 01

  which can be generated by the, you know, the, at the cost of running some bits through AWS server, right? So I'm not sure what he means by that. Like, I mean, Gorn, I admire him a lot, but what, what does he mean by that? It's not very many people are going to adopt it. Does he mean like the percentage adoption within IVF families? Or the fraction of the population that's already doing IVF? Because those are, those are already big numbers. So I don't know what he means, but.

• 1:28:50Dwarkesh Patel?

  Yeah, yeah. You know, it's interesting. Like, I guess one way to think about genetic prediction, given your earlier statement that, you know, these Scandinavian countries, a lot, there's a huge amounts of IVF happening there. And part of that is because of how old people are when they're having babies. A venture capitalist can think of your company as a way to get exposure to demographic collapse, right?

• 1:29:09Speaker 01

  Yes, it's been, it's been mentioned. He, um, by the way, it's like three to five percent in the US. So it ain't, it ain't small. Like if you go to a kindergarten, there are some IVF babies running around in the playground. So, um, it's not small. Uh, so I, I don't know whether the perspective is, is this a big enough market for you to make money in it? Or is this like going to change the future of the human species that, you know, you can have different perspectives?

• 1:29:37Dwarkesh Patel?

  Yeah. By the way, Gorn is such an interesting character. Um, I've been reading from him for a long time, but I obviously like his, his persona is very mysterious. Um, I, I don't know if you have like something, obviously nothing that isn't already public, but like what is going on here? Like, how did this person get into like, it's a really interesting and detailed report that he published in every selection? And it's, it's super interesting. What is going on here?

• 1:30:02Speaker 01

  Well, Gorn is a super smart guy and he, you know, I, I, I know a lot of scholars and serious scientists and intellectuals in the academy and outside. And I will pay, even though I didn't quite agree with his take that you just

• 1:30:20Speaker 01

  mentioned. I mean, it might not be technically wrong because he used words there that I'm not sure what he means by those words, but, and I'm not sure he would disagree with the, the quantitative things I just mentioned to you. So, but I just want to say some positive things about Gorn because I, I like to read his stuff. And so in the early days, he was following a lot of this stuff about genoid prediction and embryo selection. And, you know, he's written stuff on that. He's written stuff on GPT three and alignment risk. He's written lots and lots of insightful things and I think he's quite impressive. Even if you compare him to like the most, you know, famous academic scholars, like whether it's a Steve Pinker or,

• 1:31:02Speaker 01

  you know, somebody who just has written a lot of stuff that people read and has obviously been thinking deeply about a lot of different things during the course of a very serious life, reading and thinking and writing. I think Gorn is super awesome. I think he's right up there with those guys. So I think it's awesome that we live in this internet age that some totally anonymous dude. can produce really good thinking about a wide variety of things and he's not wrong. Most of the stuff he writes about embryo selection is pretty much right. So, um, yeah, I, I, I have very high opinion of Gorn.

• 1:31:35Dwarkesh Patel?

  And yeah, so it's interesting with people like Gorn, you can, i- it's almost in the model. You can think of, of early 20th century or late 19th century, these gentlemen scholars, who would just pontificate about a lot of different subjects.

• 1:31:51Dwarkesh Patel?

  I, I wonder if we're going to see a return of this sort of generalist thinker. And maybe we've, um, we've over indexed on specialists, but now it's like, now it's the time for like somebody like you, right? Like theoretical physics, bringing all that computational and mathematical knowledge to genomics. Um, is that the new trend in, uh, science, at least at the upper levels?

• 1:32:11Speaker 01

  I don't think it's a trend. So, so in terms of Gorn having a platform. So, so first of all, he's there, he's thinking, really, you can tell he's thinking, he's reading a lot, he's thinking, and then he's writing very insightful stuff. And he has an audience thanks to the internet, right? So people can read it.

• 1:32:30Speaker 01

  That is an amazing positive trend, which I think will continue. So I think we're in a kind of, in a way, we're kind of in a golden age for intellectual, you know, exchange, even the, this conversation that you and I are having is an example of that.

• 1:32:45Speaker 01

  The, the thing I'm afraid is not going to happen, just because science is so specialized now. And it takes so much, you know, money and resources and institutional support within a university or lab or something to, to get stuff done. I don't, I think it's getting less and less common to find polymathic people who are actually able to do things at the frontier where they really make a significant contribution and it's recognized by the natives in that subspecialty. That's becoming rare, rarer and rarer. It was much less rare in the time of like Feynman and von Neumann and people like that, just because the science was smaller.

• 1:33:28Speaker 01

  You know Feynman played around with some molecular biology when molecular biology would become a big thing. He was friends with Francis Crick, who was down in San Diego. And so he would do stuff like that. And now it's almost impossible. And people would tell me, Steve, like, serious theoretical physicists would be like, Steve, why are you fucking around with this stuff? You're wasting your talent, you know? They would literally say stuff like that to me. So I don't think the trends are good for that. But for general intellectual exchange, I think the trend is good.

• 1:33:57Dwarkesh Patel?

  Yeah, that's interesting. Going back to IVF, do you think the gains will be greater in any given trait you could think about? For parents who are already high in that trait? Or for parents who are lower in that trait compared to the average of the population?

• 1:34:14Speaker 01

  I don't think that the base level of mom and dad is a very big factor, actually. The big factor is how good are your predictors? And how many embryos are you looking at? Or how good are your editing tools? By the way, I just want to reinforce something I recently learned. It was so amazing that it freaked me out. Because I thought, oh, I'm kind of in this field, so, in this industry, so I kind of know about it. But we were, our company was having some conversations with a company that handles egg donation. So it's in the IVF space. And the egg donors are typically young women, like 22, 23. They could even be college-age women who are paid, you know, a fair sum of money

• 1:34:59Speaker 01

  to go through an IVF cycle and just donate the eggs to some, you know, billionaire family or whoever wants, you know, whoever needs the eggs. And I was told that 60 to 100 eggs per cycle is not unknown. So it's totally shocking because usually it's an older woman who's like in her 30s or 40s who's going through it and they're struggling just to get, you know, some viable embryos. And then, so then when you run that same process with a 19-year-old, what do you get? And I was kind of shocked at how high these numbers were. So in principle, let's just imagine you're a billionaire oligarch,

• 1:35:43Speaker 01

  but very tech savvy and you want to have some, you want to have a large family and you want to have really, you know, high-quality kids, maybe very long-lived, healthy kids. You might be selecting the best out of hundreds. Like there are 100 parallel universes I could live in. I get to peek into each one and then choose, I'm going to step through door number 742, because that's the outcome I like. Not that expensive, actually.

• 1:36:13Speaker 01

  But amazing that people can now do this.

• 1:36:15Dwarkesh Patel?

  I guess that'll imply that the returns of being young when you have kids are gonna incre-- because you IVF is like theoretically supposed to be, oh, you can have kids when you're old as well now, right? So it's evening the playing field. The addition of this with the additional embryos where somebody was young is like, no, it's actually we're tilting it way in favor of young now. At least if you care about the, those kinds of traits that IVF could, sorry, genetics screening could help you figure out. So I, let me just ask you what you think about some of the possibilities that Warren talks about in that post. One is that we might be able to turn induced pluripotent stem cells into embryos and then

• 1:36:56Dwarkesh Patel?

  we'll be able to select across hundreds of embryos without having to harvest eggs.

• 1:37:03Speaker 01

  Yeah, so eggs are the limiting factor, sperm is cheap.

• 1:37:09Speaker 01

  And the technology, the stem cell technology to take a skin cell and revert it to the pluripotent state so that it can become some other kind of cell, not a skin cell, but maybe an egg cell.

• 1:37:28Speaker 01

  That technology has been more or less mastered for mice and rats, I believe. At least maybe rat is the most common

• 1:37:38Speaker 01

  model system. So there's a few labs, like in Japan, where they seem to have fully mastered this and they've done multiple generations of rat doing, using induced pluripotency to make the eggs. And so my guess would be to get it working in humans is not that hard. It's a matter of some years of just slaving away in the lab to get it working. And I know of startups that are actually working on this. And now there's going to be some trepidation. Initially, like, why would you do that? If you can just pay some nineteen year old to be your egg donor? Or something? Uh, for example, some gay couples really want to do it because that's maybe they think they can also, uh,

• 1:38:21Dwarkesh Patel?

  Oh.

• 1:38:21Speaker 01

  Well, maybe they could, you know, their partner's skin make it into an egg. Okay. So.

• 1:38:29Speaker 01

  There are reasons why you do it, but for a lot of people, I think they would say like, that's, that, that's a, that egg was made through a new and untested process. I'd rather have an egg where I don't have that additional risk in this whole. Thing. So I don't know, adoption wise, what's gonna happen there. But I do think that. It's just a technological prediction. It will be possible. It won't, it isn't, we're not that far from being able to do it. I mean, the fact that we can do it in rat means I think we're not too far. And yeah, it could have huge implications for natural selection. If you really wanted to. Be able to select from best of one thousand embryos. There's no technical, I mean, eventually there's no technical barrier. Now, I would say that.

• 1:39:13Speaker 01

  On roughly the same time scale for. The pluripotent. Production of eggs. To get mature, to be tested, so that people are confident in it. Um. I think on that same time scale, multiplex, very accurate. Sort of CRISPR based editing. Will also. Uh, arrive. And so at that point, it's like, why are you fooling around with this? Just, I'll just go in and did it, make the changes I need to make.

• 1:39:45Speaker 01

  And over that same time scale, I think. It's roughly the time scale over which we're going to figure out where the real causal.

• 1:39:52Dwarkesh Patel?

  Yeah, yeah.

• 1:39:52Speaker 01

  Close they are. Which is what really needs.

• 1:39:54Dwarkesh Patel?

  I was just about to ask, cause.

• 1:39:55Speaker 01

  Yeah.

• 1:39:56Dwarkesh Patel?

  Otherwise you're just changing the tag. Yeah.

• 1:39:58Speaker 01

  Yeah.

• 1:40:00Speaker 01

  So all of this is stuff that you're younger than me, so I'm fully confident you're gonna see it all. I may not see all of it, but I'll see it, you know, the technology perfected. I won't necessarily see this impact on society. But you'll probably see it all.

• 1:40:16Dwarkesh Patel?

  I'm hoping it's ready by the time I'm ready to have kids, which is still a while away, so. Another possibility that Gorn discusses is iterated embryo selection, where you just, you can keep. I'll let you describe how it actually works, but what do you think about this possibility?

• 1:40:30Speaker 01

  Yeah, so there it's like you make the embryo, you make a bunch of embryos, and then you decide which ones you like. And then before you actually make it into a person, so that then that person grows up and reproduces, you actually reproduce just using iteration of embryos. That's also plausible too, so I think all of these, you know, very molecular technologies have a have a chance of working. I don't know anybody who's working on that, actually, really like spending all their time working on that, but yeah, that could work as well. Well, I just want to say that, you know, like I made these jokes about the wokes and progressives and people like that who hate us. And I actually just feel it's kind of wrong-headed of them. I think actually the goals, like,

• 1:41:14Speaker 01

  I actually consider myself a progressive. I don't consider myself woke, but the goals of having healthy people, maybe healthy, beautiful people who live to be 200 years old. Who's against that? You know, like, I'm also against inequality in society. I think, you know, consistent with growth and advancement in science and technology, we should try to have a fairly egalitarian society. I'm for all those things. So I think if you're a wokester who's watching this interview to just like hate Steve Shue or something, think about it. Think about why you're angry at me. Like, I'm actually exploring how the world actually is, and don't you want to know? How the world actually is? If we have an inequality problem

• 1:41:59Speaker 01

  because some people don't do well in school. Don't you want to give those families these resources so they can fix it for the next generation? Isn't that the ultimate goal of what you want? I mean, just think about it.

• 1:42:11Dwarkesh Patel?

  Yeah, to steal, I guess to steal, I guess to steel man them a little bit, somebody might say, listen, one of the things that prevents just runaway divergence between families over time in the model of like Piketty or something is, is just reversion to the mean. And I, you know, I listened to your conversation with Gregory Clark where he says this is kind of already the case. But to the extent that it doesn't get like magnified over time, the reason is, yeah, it's hard to like maintain a legion in genetics because of reversion to the mean. If you can keep that up and if there's like increasing returns to having good genes because you can then afford these kinds of treatments, then the possibility of like, you know, instead

• 1:42:56Dwarkesh Patel?

  of like a normal distribution for society, you can just have a bimodal distribution that keeps getting further and further apart. That is a potential possibility.

• 1:43:05Speaker 01

  The Morlocks and the Eloi. Yeah, I mean, I think that is a fair concern that this could lead to grotesque, huge inequality. And that is a risk of the technology in it. A lot of that depends on society, too. I mean, like, when someone confronts me with that, I will acknowledge it as a legitimate concern, but then I'll say like, you know, we live in a country which is the rich, in some sense, the richest country in the world, and there are plenty of people who don't even have healthcare. Are you worried about that inequality? Like, we have a lot of inequality. There's a lot of things for you to worry about when it comes to inequality, and this is some technology which could contribute to it, but doesn't have to.

• 1:43:47Dwarkesh Patel?

  Actually, maybe this might not be globally beneficial, but for at least this particular debate, it might be beneficial if the case was, when I asked you like, oh, do people who are lower on some trait have a greater potential for increasing that trait than somebody who's higher up on it? If that was the case, then you could just say like, listen, the, the, the smart people are just going to ask them Todd at some point, whereas the dumb people can just catch up over time, right?

• 1:44:12Speaker 01

  Well, I, I think, again, like, if you're more of a left guy and you like government intervention, and so this becomes part of the government healthcare system and it's free. And you say, "We will allow more aggressive edits or more embryos to be produced for below-average families." There's a very natural way you can redistribute, just like you're gonna forcibly take a bunch of money from me when I die that I would rather pass on to my kids. You're gonna forcibly take it from me. Well, you can forcibly give more genomic prediction resources to people who need them. It's easy.

• 1:44:46Dwarkesh Patel?

  Okay, so in your, just to shift topics quite a bit here, in your, you had an interesting post on that recent Twitter viral meme about the word cells and shape rotators, about how actually the concept of a shape rotator is combining two separate abilities, math and spatial ability, that are, yeah, when you do like principal component analysis and psychometrics, they turn out to be different but correlated. I, as a programmer, I'm really curious about which of those is the one that is required more for that particular skillset? Because I'm the kind of person, you know, when we're talking about like abstractions and data structures and the flow of a program,

• 1:45:30Dwarkesh Patel?

  I'm the kind of person that intuitively likes to think about it from, I just like imagine what it looks like visually. Whereas I know friends who, I said like, "Okay, so clearly, programming is a visual, spatial ability." And they said that actually, they don't imagine it visually at all, that for them, it's much more of just like looking through the loop and like, "What's gonna happen next? What's gonna happen next?" So yeah, I'm curious, like, which of these is a better description of a programming as life?

• 1:45:54Speaker 01

  Yeah, I think it, I think your description captured the whole story, that people are very different in the way they attack, even though they're attacking the same problem, the way that their brain does it. I think that's one of the most fascinating things about this field of psychometrics and psychology is that, you know, we're really trying to get into that. One of the things that fascinated me when I was being educated and going through training as like, in theoretical physics and math, is like, looking at how my, you know, whatever classmates at Caltech or Richard Feynman or somebody approached a problem, which might be totally different than the way I would do it, or the way that we would communicate about the solution once we got it. And there clearly are people who are visual, like Feynman was a very visual thinker.

• 1:46:37Speaker 01

  Other people are more kind of logic go verbal. Where they're like stepping through things and it might even be like they hear the arguments as they're stepping through it or something so. Everybody's different and it I think those things are super fascinating. Something that's kind of gone out of. Fashion now, but was very in fact, very in, very standard when I was growing up is like when I took shop class. I don't know if you had to take shop class in junior high or high school, but we had to take like shop class. Which where you go and bend metal and literally they have machines that would. We I made like an ashtray or something out of steel or something. So yeah, in that class, which is very spatially loaded, like you could have guys like I had a friend who was.

• 1:47:22Speaker 01

  You know, had a very high SAT score and went to Princeton to study English. That guy could not.

• 1:47:29Speaker 01

  Spatial rotate at all. He was totally lost in figuring out how to like do the bends to make the ashtray or whatever, right? So you see that very clearly and in those old days when things were more based. When you went to shop class, sometimes they just give you a standardized test, which was a standardized test of spatial ability. So we're all like my generation is like. Uh, you don't have to lie to me about all these things. We saw how it works. We saw people take the standardized test for spatial. Ability spatial, you know, visualization, and then we saw people try to fucking work the. You know, the machine, the metal bending machine and some people just couldn't do it like they couldn't actually make the thing look the way the product looked the way it was just look. So in the real.

• 1:48:13Speaker 01

  Economy of atoms and grams of steel, kilograms of steel, which has all moved to China now or something. That all this stuff is super important, like you can't just like theorize about like, okay, then I have this module that does this and this function is going to have these types and. Well, that's nice. That's super valuable in this part of the economy, but somebody's got to get this plant working. And it's got to be efficient and we got to put the machines here and here so we don't have to carry the shit too far from here. You know, there's a lot of like, that's very spatially loaded stuff, which used to be part of the American economy and education system, and now I think it's all gone. But it's real, it's not fake, no, no, people are not making this up.

• 1:48:53Speaker 01

  And and psychometricians of the nineteen fifty's and sixty's would have been like, yeah, here's here's my ten volume treatise on spatial, you know, measuring spatial visualization ability or something, so.

• 1:49:04Dwarkesh Patel?

  Yeah, yeah, even if you read the biography of somebody like Einstein, I mean, he was especially known for being a spatial thinker of.

• 1:49:10Speaker 01

  Oh, he was incredibly visual, incredibly visual.

• 1:49:12Dwarkesh Patel?

  Right, just like thought experiments that are just basically, what does it look like or what does it feel like to be moving at this speed or whatever? Yeah, that's interesting.

• 1:49:22Dwarkesh Patel?

  Yeah, so I guess in the case of programmers, I I'm not sure I got your answer, but for that particular discipline, which do you think is the more pertinent skill?

• 1:49:30Speaker 01

  I was gonna say, people are gonna do it different ways.

• 1:49:34Dwarkesh Patel?

  Yeah.

• 1:49:34Speaker 01

  I do think that if you compare the category of engineers to the category of software developers, engineers generally, I think, have higher, on average, higher spatial ability and they're using it. Whereas you can be an awesome programmer with like zero, I think zero spatial ability, that's my guess.

• 1:49:53Dwarkesh Patel?

  Yeah, I wonder if the, you know, in when you're studying history or something, you notice that some people are really attracted to the military history aspect of it and seeing how the units are moving and stuff. And I wonder if that's because they have a higher spatial ability and they're just, they just need to be able to understand how the units are moving and so on.

• 1:50:12Speaker 01

  I was gonna say, this is a very weird thing for me to reveal, but like, sometimes when I'm having trouble falling asleep, I'll be visualizing, like recently I was thinking about, you know, how I would use a ballistic missile to target like an aircraft carrier, right? But like, have the Chinese actually solved this problem or, you know? And sometimes if I'm trying to go to sleep, I'll just like be visualizing, like, okay, when you're at about an altitude of, you know, five kilometers, what can your radar see and how much resolution do you need? And then how much time do you have for course correction to hit the ship? And, you know, like, I'll be thinking about stuff like that for relaxation.

• 1:50:51Speaker 01

  But I'm and it is highly visual and but you're and also quantitative because you have to make some estimates. But but like, I think that would be typical of like a lot of physicists. Because if we start talking about it, we'd be like, oh, yeah, right. And you've got you've only got about point of order a tenth of a second to do this. And you but you're thinking do it in milliseconds. So we're OK. And then anyway, that kind of thinking is very prevalent amongst certain types of.

• 1:51:13Dwarkesh Patel?

  Right, right.

• 1:51:16Dwarkesh Patel?

  Now, I'm curious why it's the case that people from physics so often transition to finance. I think that was something you were considering at one point. Is the underlying knowledge in mathematics is just the same? Or is it just such a credible signal of mathematical ability? And G that, you know, quant firms and whatever, they want to hire physics students?

• 1:51:42Speaker 01

  The answer is a little bit complicated. I think all the factors you mentioned are true. But one of the things was that in the early phase, like in the 80s and 90s, when a lot of people in my generation went into finance, a lot of them went to do to trade derivatives. And if you look at options pricing theory, it looks a lot like physics. It's kind of like the mathematics of random walks, basically. And so there was a very tight, not tight connection, but the concepts were were strongly related that were necessary. Now, if you brought it out a little bit more to say like, OK, but nowadays, if you go to really big quant funds and they're looking for signal and analyzing tons of data,

• 1:52:25Speaker 01

  and they're not trading derivatives, they're trading, you know, just actual names like stocks or whatever.

• 1:52:32Speaker 01

  A lot, I think there's more loading on machine learning and CS background now. And the physicists who go in, they're having to they're they're using that subset of their skills. But the funds would just as soon hire a CS or ML type guy. To do it. So it's a little bit of a complicated answer.

• 1:52:50Dwarkesh Patel?

  Yeah, that's super interesting because I mean, back in the 90s and early 2000s, I don't know, I was watching.

• 1:52:57Dwarkesh Patel?

  I read that book about the fall of long-term capital management, then obviously, I guess this is a cautionary tale, but still, it's kind of cool to.

• 1:53:03Speaker 01

  Actually, there are two books. There's one called When Genius Failed, and then there's another. There's actually three books, at least three books, but they're all good.

• 1:53:11Dwarkesh Patel?

  Yeah, and then you just hear about like, obviously, the people who created options, pricing theory are there, but the applying, you know, calculus or animal ox and stuff. And the stuff I don't understand, but just super cool that you have these mathematicians that are just coming in and applying these ideas to finance.

• 1:53:28Speaker 01

  I do want to say one thing about physicists, which is a little different from mathematicians and computer science guys. Maybe not so different from data science guys, but definitely different from most computer science guys and most math guys, is that we spend a lot of time looking at bad, noisy data. So even if you're a theorist, you had to go through these lab courses where, I mean, for me, those lab courses were among the hardest, like the worst. Because you had to go in and build some electronic equipment to take some data and, you know, it could be extremely noisy, like you're measuring muon cosmic rays coming through the roof and hitting your detector, and then you have to analyze the data. And when you're building this thing, you screw it up. And so like you get data that makes no sense, or you

• 1:54:13Speaker 01

  something about the amplifier wasn't right, or you're used to seeing data that sucks. And you have this theoretical view of what should be happening, like maybe you're visualizing it like the muon comes in and it does this. And interpolating between the theoretical view of what should be happening with the particles and the systems and what the actual data looks like and saying like, oh shit, we didn't do this, or we didn't shield this part, so that's why we're getting that. That's something physicists are very, very used to doing. And mathematicians are often shitty at it. They just accept, oh, I just accept this is the data. Now I will, now I'll reason with this data. And the same could be true for computer science people.

• 1:54:57Speaker 01

  But you need someone who's actually had to deal with shitty data and tried to connect it to a very elegant mathematical model. That that's something physicists kind of uniquely are used to.

• 1:55:07Dwarkesh Patel?

  No, but I think that's also true of CS people, which is that you have, obviously in debugging, you, there is many potential problems that could happen. One of them, obviously, is just you wrote the code wrong. But often you get it, the actual implementation just right. It just, there's so many layers of abstraction beneath you and above the actual hardware that you, you have to figure out, like, why is the correspondence between this idea I had and the actual program output not the same?

• 1:55:36Speaker 01

  I think that's fair because, yeah, debug, when you debug your code, there are many different ways it could have failed. And you have to actually, in a sense, step back and model. Like, well, maybe this module is feeding me something back wrong and that's what's causing the problem or it's this other layer. So that is very analogous to like when we have to deal with a physical experiment in the lab. The thing with physics, though, is that we were really, really geared toward getting toward the underlying reality. Like, if it's really late at night and my, my buddy, my lab partner and I just want to get out and go to sleep, we can't like tell ourselves that things are okay. We didn't actually screw up the shielding on that. It's okay, we'll just bring the data home and look at it. No, we got to actually decide.

• 1:56:21Speaker 01

  Do we have to spend three more hours ripping this thing apart and reshielding it? Or we have to get to the real underlying reality. We can't fake it. We can't just pretend like, oh, the, this admission scheme will work perfectly. You know, like, you know, we can't, we can't lie to ourselves about it. And that, that, I guess that's true for coders too. But, but anyway, it's just very different from like social scientists and stuff where they can just decide, ah, I don't like that reality. I'll just make up this model for how society behaves and then I'm done. We can't do that.

• 1:56:51Dwarkesh Patel?

  Yeah, so given the skill set that theoretical physicists have, as you just mentioned, is it potentially the case? I mean, obviously, the common criticism of like physics as a community is that they're absorbing too much talent. You know, like three or four standard deviations above average intelligence people are working on. A field that I guess in popular convention at least seems like isn't making as much progress. And then, so should more people in physics be making the step that you made? Which is just like, yeah, I learned all these skills in theoretical physics. I'm going to move out of it. Maybe finance is one way in which like we're getting these pro-social benefits from the skills that physics builds. But also, yeah, just like stepping into fields like genomics or things like that. Should more physicists be just using their skills elsewhere?

• 1:57:38Speaker 01

  Yeah, so number one, the attrition rate is super high. So for, even if you cut, you say like, take this set of kids that are plus three or four standard deviations in ability. And they enter, they enter a physics major at Princeton or MIT or something. What fraction of them actually end up as practicing physicists? It's pretty small, so they're bleeding off at all points. Like, you know, Bezos started in physics and toward the end of his Princeton career, switched to computer science. And, you know, Elon was in graduate school in physics, at applied physics or physics at Stanford, and he bled out. So it, it's already the case that for me, one way to say it is the education is phenomenal. You should try to get that education.

• 1:58:21Speaker 01

  It'll pay off for you later and probably you're going to bleed out. You're going to attrit away and do something else. Now, if you say like, okay, of the thousands of theoretical physicists or physicists who do fundamental research, including experimentalists around the world, there are tens of thousands. And maybe some of those guys should also be doing some more cancer research or doing financial modeling. Yeah, maybe so, maybe so. I mean, maybe even some of those guys should, you know, we should tear off, you know, we should remove even more of those guys and have them do more applied stuff. That, there's still some argument in favor of that. But we do need a core of people that are trying to do these really hard fundamental, answer these hard fundamental questions about nature.

• 1:59:00Dwarkesh Patel?

  By the way, the Bezos example is really interesting. And obviously, by the way, for the people in the audience who might not know is he was asked once why, I mean, his original plan, I think, was to become a theoretical physicist. And the reason he didn't pursue it is that he noticed one of his friends was just so much obviously more gifted than him at that skill that the story he tells is like they were, him and Bezos was working on a problem for many hours and making no progress. And one of his friends just looks at it in an instant and he's like, oh, the answer is, I don't know what it was, but like blah, blah, blah, cosine of something. And then he's just like, yeah, I just eliminated all the terms. I recognize a similar problem. And so Bezos was like, okay, this is not my competitive advantage.

• 1:59:42Speaker 01

  Can I just say one, I gotta add one anecdote. The guy, that guy, so I know a lot of the guys that were in Bezos's eating club and were also, because we're very similar in vintage and who took all these classes with him. And a lot of them were late, one was a friend of mine from high school, but another were guys, a whole another set were guys that I went to grad school at Berkeley, because there's a whole Princeton contingent that would go to Berkeley for grad school that I knew that were Bezos's classmates. So I know all these guys, I know all these Bezos stories. The funny thing is the guy you're talking about, whose name I believe is Yasanta, is an Indian guy, Sri Lankan guy. He went to grad school at Caltech. And so actually, he and I, I don't remember how, I looked him up at one point, I think

• 2:00:26Speaker 01

  we met up at Caltech when I was visiting at one point and he was in grad school. And so I actually met this guy and talked to him about Bezos, it's kind of, because we had friends and other, we weren't, we weren't focused on Bezos, we had other friends in common, but I actually met this guy that is in that anecdote that you just mentioned.

• 2:00:40Dwarkesh Patel?

  Oh, no, because actually, that's, that's really good to know, because that, that is very relevant to my question. I, my friend and I have this continual debate about the importance of intelligence at the peaks of entrepreneurial ability or engineering ability. And he tries to use that anecdote to say that, oh, clearly Bezos was not smart enough to be a theoretical physicist, so therefore, intelligence is not that important beyond like a certain. Not especially high point and afterwards, as such Bezos was creative or blah, blah, blah. He was hardworking. And I don't know, my perception of the story was like, okay, he's not smart enough to be a theoretical physicist.

• 2:01:21Dwarkesh Patel?

  He's like below five standard deviations or four standard deviations above the, but like clearly just studying Princeton at, sorry, physics at Princeton is itself a testament that he's probably at least like two or three standard deviations about, well, at least three, I mean. But, okay, so can you tell me more about like, what was the perception of those people you talked to at Princeton? About Jeff Bezos, what like, is it that he just like, he just was super high in other traits like hardworking or creative, or it's actually intelligence was super high, just not high enough to be a theoretical physicist?

• 2:01:52Speaker 01

  Yeah, this is a great topic that, you know, I think a lot of people are interested in this topic. And even among my close friends, including these friends who know Bezos or knew Bezos in school. We all talk about this kind of stuff. So first of all, you got to make a distinction between the very abstract kind of intelligence, which is useful in physics and math, or maybe computer science versus a more kind of generalist intelligence. And those are correlated, but they're not the same thing. And so, you know, I would say Bezos is probably very off scale for ability to work hard, take risk, function under pressure, be focused and generalist intelligence. So he's just probably off scale on, you know, if you're just, since these traits are at least somewhat uncorrelated,

• 2:02:37Speaker 01

  if you're top 10% in each of these five simultaneously, already a pretty rare individual, right? Cause plenty of the physics guys who did better than Bezos in the physics classes, they could not lead a company, they could not put together a presentation that would convince a venture capitalist to invest. So it's sort of, you know, different skillsets that we're talking about. I think the idea that there's a unidimensional measure of cognitive ability is just not that useful. You know, I'm probably guilty, people will say, wait, Steve Shu just said that, but he's the guy most responsible for promulgating this perspective. But it's only because it's, it's the simplest thing to talk about is if you compress it to one general factor, it's just easier to talk about. It doesn't mean that the other components are not meaningful.

• 2:03:20Speaker 01

  We just got done talking about verbal versus spatial versus some more generalized mathematical. Talent. So obviously, it's a much, it's a high dimensional, not that high dimensional, but it's at least a multi-dimensional space of abilities that we're talking about. Now, the point about Bezos, I think, which is non-trivial, though, which I think is directly relevant to like, the life experiences of like physicists who leave physics and do other stuff, is that very often in an engineering setting, or a startup setting, people will be like, "You don't know shit about that. What are you talking about?" Right? But the reality is people who do perform on a technical problem, okay, not about what's

• 2:04:00Speaker 01

  the right way to get a good, a warm intro to this VC, not something like that, but but some technical problem that the startup has to solve. Like in Bezos's case, it was often like optimization of some supply chain thing, or optimization of some sorting process, or reducing the error rate and some like, you know, address labeling, you know, it was a very well-defined thing, once you, operations problem. And the people in the company, you know formally say, like, when Bezos comes in the room, he will give us, he will give us very good feedback on the solution to this ops problem, that, you know, it could be out of the blue better than what we said, or at least he finds the problems with what we said, or he, if we did a good job on it, he gets it right away, which is

• 2:04:43Speaker 01

  some executives might not get it right away. So, my point is that people who have these super high, just raw, gee abilities, they generally can be useful in these technological environments, even if they don't have a lot of background, like, they can still come in and be helpful, and sometimes they can solve problems that the people who are well-trained in that area are having trouble with. I think that is fair. But it's not fair to say there's just some, you know, you know, unidimensional measure of intelligence, and this guy always beats this guy, this guy always beats, it doesn't work like that. But it's just that, it's just that some of these off-scale guys are just generally more useful than the critics would like to give them credit for.

• 2:05:23Dwarkesh Patel?

  Your life story is kind of an example of that. But, you know, I had another experience of this, which was, I recently interviewed Sam Beckenfree, who is the CEO of FTX, on my podcast. And one interesting, like, for that interview, and I guess generally for all interviews, I try to come up with questions that I think that The guest has probably not heard before. And in that one, in that case, I tried really hard to come up with questions that he might not have heard before. And that might have been like really interesting and challenging. You know, I listened to all the interviews ever done and then, yeah, I've prepped for a long time. And if you listen to that interview, the thing you'll notice is the way he answers these questions, it like, sounds like he was just, "Oh, I was just talking to somebody about that. Let me just say again what I was just thinking."

• 2:06:08Dwarkesh Patel?

  It's like, no matter how creative a question I could try to throw at him, it just, his ability to grok,

• 2:06:16Dwarkesh Patel?

  like, all the context explained in the most, in the way that an audience would understand, it was kind of exceptional.

• 2:06:25Speaker 01

  Being a super successful founder selects for the ability to figure out, "Okay, this guy's, this investor is from private equity. This is how he's going to think about the problem. And this is how I should explain it to him. This guy's from a very tech-heavy venture fund. This is how I got to talk to him about the problem. This is the due diligence guy that they sent me, and he's a computer science professor at Stanford. I got to talk to him in this language." So founders are very selected population for being very good multi-band communicators across different cultures and stuff like this. "This is a dumb investment banker from Goldman, and, I mean, they're not dumb, but he's not technical at all. And so I got to explain it to him this way. But this guy's a lawyer. I got to talk to him that way."

• 2:07:07Speaker 01

  So it's not surprising to me that, you know, this guy would have those capabilities. It's not, it's selected for in that population.

• 2:07:16Dwarkesh Patel?

  Right, yeah. Okay, so you are a practitioner of jiu-jitsu and other martial arts. And one of the things in, one, you know, obviously, one notable aspect of those disciplines is that you can really punch above your weight, right? And that, you know, Royce, Royce Gracie in the UFC is a great example of this.

• 2:07:40Dwarkesh Patel?

  Is, is that possible with a trait like intelligence? Is it possible that we have techniques or other ways of compensating for your, I guess, analogously, your just raw weight that, what jiu-jitsu is to fighting?

• 2:07:59Speaker 01

  That's a great question. So in a way, jujitsu is like applied physics. Because you're thinking about like, I have two arms, you have two arms. You know, is it easier for you to punch me and knock me out before I can close the distance and force you to grapple with me? It really is. The reason I like jujitsu so much is because it's very rational. It's basically scientific analysis of what two humans can do to each other. And so it's a technology. And in terms of what technologies people can use like to amplify their brain power, obviously, we're surrounded by it. So like, here's an interesting thing.

• 2:08:40Speaker 01

  Suppose you and your girlfriend are trying to like get the answer to some question, and you're both using Google. There's an enormous variance in who immediately puts the search term in that gets the right answer, like the top hit is the answer, direct answer to your question. And that's very G loaded. But you could use technologies to improve yourself. You know, if you train, if training is not the right word, but if you kind of get good at using certain technologies or certain information channels, you can amplify your ability beyond just what the raw, you know, capability.

• 2:09:16Dwarkesh Patel?

  Yeah.

• 2:09:17Speaker 01

  So I think my answer is that there are tools. But there's nobody who like, there's no dojo where you can go and Henzo Gracie just like search teaching you immediately like this, do this, this, this, and this. And then you're gonna, the guy is bigger than you, but you're gonna take him down and choke him out. There isn't something like that for cognition.

• 2:09:37Dwarkesh Patel?

  Right.

• 2:09:37Speaker 01

  But I can see like, people can amplify their capabilities in different, either more or less effectively.

• 2:09:45Dwarkesh Patel?

  Now, you had a blog post a long time ago about elite education. And in it, you, you talked about how even if you're controlled for SAT, at the very top jobs, the people from elite schools are overrepresented. And so I'm curious, do you think this is, this is because of a selection effect based on like Harvard selecting based on personality as well? And that, you know, that selects for a certain high achievers? Or is it something about being at Harvard that makes you a high achiever? What is going on?

• 2:10:19Speaker 01

  So first of all, I researched this question pretty aggressively when I was first, when I first became an entrepreneur. Because I was like, well, we can raise this much money, we can get these meetings with these funds. But how the hell did this guy get, raise a hundred million for this stupid idea? Like, what, what the hell? And then, so I would start looking into this guy's background. I'd be like, well, he went to Harvard and, oh, he was in Skull and Bones, you know, whatever. So I got intensely interested in like, okay, these super outlier guys. Like, how did this guy get a job writing for The Simpsons? You know, like, what, I would like to write, you know, this other guy would like to write for The Simpsons, but he went to Ohio State, so he's like ten layers of social networking away from The Simpsons. But the Harvard guy's not, actually.

• 2:11:03Speaker 01

  His buddies at the Crimson all write for the, you know, The Simpsons, right? So there are multiple factors why, take two kids, they both scored fifteen, eighty on the SAT. One goes to Ohio State on the Ohio Region Scholarship for Engineering. And the other one says, no, fuck no, I'm gonna go to Harvard, even though the engineering school there sucks, but I'm gonna go to Harvard instead. Okay, so what's the difference in their lives? One, somehow, maybe the guy went to Harvard because he kind of understands how the world works a little better than the other dude. Okay, two, when he gets to Harvard, he's gonna meet a lot of super ambitious, aggressive, smart kids. Some of those kids are children of super wealthy people, some of them are children of super influential people.

• 2:11:49Speaker 01

  And all of them are trying to get ahead, they're super ambitious. They know what it means to like, make managing director at Goldman, or become a partner at McKinsey. They know what those things are, okay? And you, if you didn't know them, because you grew up in Ohio, you learn them right away. Because you, you see what Joe, who was two years ahead of me, but had the room across the hall, he interviewed, now he's at McKinsey, now he's doing this. You just get a better view of what's possible in the elite sector of society from that exposure. So there, there are multiple factors, networking. Some of these Harvard kids come from super wealthy families. Some of them, their dad used to play golf with, you know, the head of the, you know, of, you know, the fund that he's trying to get a meeting with, right? So.

• 2:12:32Speaker 01

  It's all those things together. I'm not saying it's good, but I kind of want to understand how the world works. I kind of understand why this other dude can raise so much more money than I can raise or get meetings that I can't get, right? So that's how I was initially interested in this question.

• 2:12:48Dwarkesh Patel?

  Why are China and India underrepresented in, like massively underrepresented in Nobel Prizes per capita? And, you know, even in, in computer science, when I would like try to find papers on certain subjects, often those papers, like very, it was rare that they would come from China or something like that. And when they did, it was just, the quality was much worse than the ones that I could find from like a professor in the US. And I'm curious why you think that is. This is, so obviously, it's clear that it can't just be the population or anything like that, because when those researchers come to the US, they're producing stellar research. What is happening here that, like, why is, is this effect real? And if so, what is the explanation?

• 2:13:33Speaker 01

  Well, the, the easy answer to that question is, many of the things, or almost all the things you mentioned are lagging indicators. So, so they reflect the fact that the West was developed and had a strong scientific and engineering tradition when China and India were desperately poor and just didn't have any of that. And in my own life, I went, in the last 20 years, from when I would visit a university in China, or even like South Korea and Taiwan, I could see them go from, they had plenty of talented undergraduates, but the best of those undergraduates always want to come to the US for their PhD. They went from that to now some of the best undergraduates stay there, and the researchers

• 2:14:17Speaker 01

  who are professors there are becoming world-class. But, but that, that happened only in my adult lifetime. So you can see it's a heavily lagging indicator. Interestingly, like, in my physics career, I knew several, I think the Indian term is called toppers. I don't know if you know this term, toppers. So the, the people who take the IIT exams, they, they literally rank every kid in the country who takes the exam, right? So, so I knew guys who were number one, or number two, or number five on the IIT. Entrance exam, but they but they ended up going to Caltech or they ended up going to MIT or you know, so there's this there's this huge brain drain. I mean, it's super powerful elite

• 2:14:57Speaker 01

  brain drain and MIT recently has just been recruiting. If you win one of these Olympiads, you get a gold medal in the informatics Olympiad or the math Olympiad, MIT will try to get you to come to MIT. So so there's this huge sucking of talent into the United States, which is great, I think. But that's why when you go to IIT, even though the undergraduates are super smart, the professors are actually, no offense to my colleagues who teach there, but the professors there would generally, if they get a bid from UCLA, they'll move to UCLA on average. So that's the difference. But that's gradually evening out.

• 2:15:32Dwarkesh Patel?

  Are there any downsides to the fact that we can have, we can pay researchers or postdocs in the US less because we're partially paying foreign workers in visas? Is that just market arbitrage that has, you know, that's just like positive externalities for the economy? Or is there some downside to the fact that it's not competitive for native born workers?

• 2:15:56Speaker 01

  Good for the US overall, on average. Bad for developing countries because you're stealing their talent. It's bad for native born Americans who have to compete against the best brains from all over the world. So much harder for an American kid to, you know, get the job he deserves, you know, at these elite levels where he's strongly impacted by immigration. So, you know, you got winners and losers. Whether there's a long term problem for America. So now, now, like there, there's some guys who are super obsessed who like comment on my blog every now and then who study like, where are all the IMO, you know, International Math Olympiad winners

• 2:16:40Speaker 01

  going? Where are they? What, you know? And they claim they're seeing this huge drop off in like, Kids who grew up in America who are not like first like children of immigrants, but they've been here a while, they just never win these competitions anymore. So, so ultimately you might be kind of discouraging the native talent pool by just letting the door, opening the door and bringing in like all these super talented people from outside. So there, there could be some second order effects that aren't so good.

• 2:17:11Dwarkesh Patel?

  Although it's interesting when you look at an industry like tech, where there's a similar aspect of foreign competition being allowed in because of Beijing visas, but the compensation has remained really competitive. Is it just because tech is a super, it's like super inelastic demand for the talent?

• 2:17:32Speaker 01

  Yeah, because, because you're maybe a little more focused on things like software development and ML and stuff. But if you look at like more kind of traditional engineering fields, which aren't as hot, probably those guys like, you know, like an engineer at Boeing or those guys would probably say like, no, my, my fucking salary is heavily suppressed by the existence of hungry engineers from India and China and stuff like that. So, you know, software, because it's been so hot for so long, doesn't feel this effect so much. It's got plenty of elasticity.

• 2:18:04Dwarkesh Patel?

  Awesome. Okay. Steve, this is so much fun. This, I, I really, really enjoyed this conversation. And in preparing for it and in talking to you, I like really got to learn a lot more about this subject that I was interested in for a long amount of time. Is there anything else that we should touch upon on any of the subjects we covered today or have failed to cover today?

• 2:18:22Speaker 01

  Wow, we, we've covered so much and I, I just, I really think you're a great interviewer. Because your, your questions are like always getting at a key thing that I think a lot of people are confused about. And there's a lot of depth there. So I, I thought it was great. There's plenty more that we could talk about. We should just get together and do this some other time. But I don't think you left anything out.

• 2:18:43Dwarkesh Patel?

  If you're willing, I would love to do a version two of this where we talk some, some about your physics work. And yeah, the, the other subjects we might've missed this time around.

• 2:18:50Speaker 01

  Yeah, we got to talk about many worlds and quantum computers.

• 2:18:54Dwarkesh Patel?

  This will be fun. I, in the meantime, do you want to give people, you know, the, your website, your podcast and your Twitter? So they know where to find you?

• 2:19:06Speaker 01

  Yeah, so my, well, my last name is H S U. That's like the hardest thing for people because it's kind of anti phonetic. H then S then U. And just search for me. I'm on Twitter. I have a blog and I have a podcast called Manifold, which doesn't have a huge listenership, but I try to keep the quality level really high where I try to get really best in class kind of guests and then we're willing to go into some depth. So it's got a very kind of niche audience. But if you like, if you like the conversation that we just had here, you'll probably like Manifold. So you can look for that in all the usual places you get your podcasts and also on YouTube.

• 2:19:45Dwarkesh Patel?

  The podcast is like, it's similar. It's exactly what I'm trying to do here in the sense that it's like really in-depth conversations. But you just know so much about so many different fields. So it's so fun to listen to where you're having expert level conversations in everything from social science to foreign policy to, yeah, obviously your fields. But like, that is, that is a podcast I really recommend. So yeah, Manifold Podcast is one you should check out.

• 2:20:16Speaker 01

  Yeah, my pleasure.

• 2:20:18Dwarkesh Patel?

  Thanks for watching. I hope you enjoyed that episode. If you did and you want to support the podcast, the most helpful thing you can do is share it on social media and with your friends. Other than that, please like and subscribe on YouTube and leave good reviews on podcast platforms. Cheers. I'll see you next time.