Steve Hsu: So again, just asking, just trying to get it to the point yeah where the average listener can understand. Yeah. These Yamnaya guys, drinking a lot of milk. They have wagons. They have wagons. They may be maybe they're warlike.

Razib Khan: Maybe.

Steve Hsu: Okay. They invade, they invade a new territory, and yeah this is my guess. Tell me if this is not, you know, the ancient DNA science doesn't support this. They kill all the men. They take the women.

Razib Khan: Yeah.

Steve Hsu: And one Yamnaya guy's maybe got a lot of women.

Razib Khan: Yeah, yeah, yeah

Steve Hsu: and he fathers a lot of kids. Yeah

Razib Khan: yeah

Steve Hsu: And the, the elite Yamnaya families are reproducing yes in this way, by crazy, and their Y chromosome is spreading like crazy. Yes, that's what we see. And when you and I saw this on the posters, you knew exactly what it meant. But the people, the people there didn't, they didn't think, they didn't translate the, the s the data into this bloody.

Razib Khan: I, I was just like, like, "What the do you think is happening?"

Steve Hsu: Yeah.

Razib Khan: You know? It's like, okay, so the Yamnaya, also the well, the Corded Ware, which is the daughter population of the Yamnaya you know, like, you guys know like Thor's hammer, you know, they have that the you know, the Mjollnir, like little, little pendants.

Well, they the, the Corded Ware, like, their pendants were maces, okay? And so what do you do with a mace? You smash skulls, okay? So these are not gentle people, you know?

Steve Hsu: Welcome to Manifold and to, what's the name of your show?

Razib Khan: Oh, Unsupervised Learning.

Steve Hsu: Unsupervised Learning.

Razib Khan: Yeah

Steve Hsu: Yeah, yeah. We're here at Manifest. It's Saturday Night Live at Manifest. Yeah. Woo! We have
Razib Khan, Steve Hsu, and a cast of dozens.

Hundreds. Rationalists. Rationalists, rats. It's about quality, not quantity. What do we have here? We have rats, we have tech bros. Post rats. Post rats. Yeah. Doomers, dreamers. We got a continental philosopher. We got Amish

Razib Khan: guy

Steve Hsu: Amish guy. Super genius looking for eggs. Yamnaya girl.

Razib Khan: I mean

Steve Hsu: All right.

Razib Khan: Were there, were there Yamnaya girls?

Steve Hsu: So we promised a Q&A. It's gonna be totally unfiltered. We

Razib Khan: got space in here. People are looking, people are like yeah, peeking through. Ben, come in, come

Steve Hsu: in. Come on in. Come on in. It's gonna be totally unfiltered. All the way back there. Yeah.

Razib will answer all your questions with 100% truthfulness.

Razib Khan: Well, yeah, if sometimes I might have to plead the fifth, okay? But in general, I don't do that very often.

Steve Hsu: Okay. I don't do that.

Steve Hsu: And I thought we would start with ancient DNA deep history. Yamnaya girl is gonna ask the first question Speak up. Speak up. Okay. Okay, can

Audience 1: everybody hear me? Can the podcast hear me?

Steve Hsu: Well, we may have to repeat your question for the podcast.

Audience 1: Okay. Appreciate all your writing on ancient DNA over the past few years,

Razib Khan: given

Audience 1: where we stand.

Steve Hsu: Okay, let me repeat the question. Most pressing ancient DNA questions. Ancient DNA, yeah,

Razib Khan: genomics

Steve Hsu: yeah. Ancient genomics questions. Yeah, yeah, yeah that need to be answered.

Razib Khan: So I think a lot of the questions related to phylogeny have been answered. The main question so phylogeny is just obviously, like, the tree of life. Although really it's the graph of life.

So there's a lot of admixture, you know. So it's a graph. It's not just like a bifur it's not like series of bifurcations. I think one of them, which David Reich talked about on you know, my friend Dor Kesh Patel's podcast is that we're not totally sure about what's going on with the admixture with Neanderthals, Denisovans, population structure in Africa.

So there was a model of Out of Africa that was very popular and very simple and easy to explain about twenty, twenty-five years ago, and it's been complicated by these various admixture events. And what David has been saying, David Reich has been saying is, okay, like you have enough admixture events, it starts to feel like epicycles, in a geocentric model.

Like, there's something wrong with our assumptions, and a lot of people think that. That doesn't necessarily mean that Out of Africa is not correct. It just means we might not be thinking about things correctly in terms of just like from first principles. And, you know, Steve's a physicist. Obviously, you know, Lord Kelvin, you know, a genius guy but, you know, his assumptions were wrong, right?

And so there is a little bit of that going on I think with you know, understanding human prehistory before 100,000 years ago, right? More recently, most of the the big questions have been answered about the Holocene in terms of who is descended from whom, who fought whom. You know I've been talking about Roman genetics a little bit I think tomorrow at twelve.

I think that there's a lot of stuff to explore about the dynamics of more recent historical processes. A lot of that stuff hasn't been published until very recently because ancient DNA is very expensive and laborious, and so they initially were focusing on things like Neanderthals because there's like no other way you could really understand this stuff.

So if you're gonna spend that much money and time and labor, you're gonna focus on Neanderthals. Now they're doing things like looking at, oh, post-Roman Britain. All right? There's other ways we can figure that out through archaeology, through some, like, textual references. But now, now ancient DNA is industrialized and scaled enough that they're using, using it to answer those historical questions, right?

The second thing I wanna say is, like, aside from phylogeny there's the bigger question of population genomics and, you know, evolutionary dynamics, and one of the major forces, major parameters is selection, okay? And obviously, there was a recent, recent paper Akbari et al. This was this paper took about eight years to publish actually from beginning to end, from what I've heard from people.

Anyway so, you know, they had a novel method using ancient DNA to detect natural selection. I think there's gonna be a lot more work in that particular vein about natural selection over the last ten, twenty, thirty thousand years. And obviously

Razib Khan: you know, obviously some of it could be random, it could be just like drift, but really probably a lot of it is adaptation and selection.

And like these are really hard questions to answer, answer statistically because you have a lot of forces affecting the genome, and teasing them apart independently, it's not a trivial It's not trivial, right? So I think that's gonna be a decades long process to really understand that how, how we came to be the way we are Going further back into the past you know, there are some people like Svante Paabo, you know, he's obsessed with why are how are we different than Neanderthals, and are we, like, special compared to Neanderthals, all these things.

I'm not necessarily sure we are, but you know, that's what they're gonna be focusing on. Can I quick follow up? Yeah, again, follow up, Phoebe.

Audience 1: So when Wright was on, Mark, how she was talking about this phenomenon that Neanderthals, they have, what, I think it's 5% modern human DNA Yes but their Y and X chrome their Y chromosome and their mitochondrial DNA are 100% modern human.

Razib Khan: Yeah. Do

Audience 1: you have any thoughts

Steve Hsu: on how Repeat, repeat the question. Repeat

Razib Khan: the question Yeah, yeah. So Neanderthals have about 5%. So Neanderthals 50,000 years ago, which is on the eve of their extinction, about 5% of their ancestry could be derived from a basal modern lineage, modern-ish lineage. So that means it split off like say two to 300,000 years ago.

So it's actually a further split probably than any of the splits of the modern lineage, of the extant modern lineages, right? So yes, it's modern, but it's only modern because it's closer to the stem modern lineage than it is to Neanderthals, first of all and then the Y and the mtDNA thing, that's been known for a while.
And so if you do Y and mtDNA phylogenies, Denisovans are outgroups to Neanderthals and modern humans. That was really confusing until they realized, oh, well, they coalesced back to a certain time. There's all sorts of ad hoc explanations about the Y and the mtDNA could be like that. Like one explanation is that, okay, there's like natural selection that's favoring the Y and the mtDNA somehow, okay?

That, that it's not a neutral process. I don't, I don't know if I believe that. Also, there's other things related to the effective population size of the Y and the mtDNA. They're uniparental so basically they're, you know, it's transmitted only paternally and maternally. And so if the effective population size is smaller, the coalescence back to the common ancestor is actually going to be more shallow.

Like it could be like some sort of stochastic process related to that. Basically, we don't know, okay? And I think David cops to the fact that we don't know. Someone asked me like point estimate. David like has like a, a short commentary preprint about like, you know, Neanderthal gene flow, Neanderthals eventually becoming like overwhelmed by modern genes and all this stuff, the model to explain this.

And I, you know, I give his model like a 30% probability. We don't, you know, we it's really unclear what's going on, and that's really difficult because people want to know very clear answers to these sorts of questions about the origins of modern, the, the modern, you know, lineage. Also, and like I gave a talk at Avalon in San Francisco to a peaks.

I think it's recorded somewhere. But basically, there's a lot of circumstantial evidence now, not just from, you know, the Cousins et al paper I think that some of you have seen, that the modern lineage is actually a fusion of two very, very deeply branched lineages, like further than Neanderthals.

And in fact, Neanderthals and Denisovans might be a branch of a group of humans that is most of our ancestry, but we have a minor ancestry component about like 30% that's from something that split off 1.5 million years ago

Audience 3: And this is the African group?

Razib Khan: Yeah, it could be. We don't necessarily know where they are.

So it's population A, population B. There's like no terms for that. And basically population B, so if you do a gene ontology GO analysis of like where the genes are, population A, which is more like Neanderthals and Denisovans, like split like 700, 600,000 years ago, all of our metabolism genes, et cetera, et cetera, all of our like, you know, cellular function genes, a lot of our morphological genes are from that.

That's our dominant. But population B has an enormous amount of neurological genes that are enriched for it. So that's like our brains are from B and ourz our body is from A. That's the one way I would explain it. So I think that that's another thing that a lot of people are, are try to figure out, like, how do we come to grips with that, you know?

Like, it could be we are, we are the hopeful monster that actually worked out. Like, that's one way I could explain it, right? So, and it's not just, like, what like, you know, there's some scientific, like Steve knows this, he's sci there's, like, sometimes, like, one scientific group quote, "discovers something," and, like, nobody else, like, really sees it.

It's like whatever. It's just, like that's not, it's, it's not like this. There are other groups that are looking at the genomics and using really, really sophisticated computational models. They're like, "Yeah, it's just, it looks like deep structure. It looks like we have deep structure." So that means that we are, modern humans are probably a hybrid lineage, which, you know.

I don't know how to explain that to people. Well, I mean, I do, but I know how to explain to people here, but try to explain to a normal person on the street, you know? So

Steve Hsu: Yeah. Let me just add a comment about the selection, polygenic selection that the Wright group published on. This is probably the most controversial issue because it, it, it's related to group differences.
And in simple models that people have been talking about for many, many years, including people like Greg Cochran it's very easy to convince yourself that selection effects could lead to one standard deviation differences in traits over timescales of as little as one or a few thousand years, and the right group found many examples of this.

So I think that unless you're really ideologically opposed to that possibility, I think you have to accept that that is very possible and very plausible that something like that could have happened in, in human lineage. And, and so differential selection could lead to differences in, in group averages.

So if you look at the northern part of Europe and the southern part of Europe, there's about a one standard deviation difference in height, and that's largely genetic. Maybe not completely genetic, but at least partially genetic. And that probably arose over a relatively short timescale, like maybe ten thousand years or less.

So isolated human populations can differ from each other. I'm not saying that they do differ from each other, but there is no principled reason for why they couldn't, and the Wright paper is delicate. The reason it took eight years for them to publish this paper Yeah, yeah, yeah. That

Razib Khan: that's what I'm saying

Steve Hsu: is because everybody wanted to shoot holes in it. Everybody wanted to, you know, everyone looked for every reason that it could be wrong, and, but it's finally published.

Razib Khan: Well, I think what I would say is, like, in principle, you know yeah, both Steve and I have, like, talked to many geneticists over the last 20 years.

You know, first of all, there's things people will say privately that are totally different. We can't really talk about well, it's irrelevant 'cause it's not the public record, right? So there's, like, kind of like a, a secret knowledge. But even in the outward in the outward-facing communications, they will say things like, "Well, in principle, this could be, but we don't know."

And they really want it to remain that way. Let's be honest. They really want it to remain that way. And I'll give you a concrete example. Okay, I'm not gonna say who because

Audience 4: It's Alex, but you know.

Razib Khan: So I have a friend now, I have a friend who was like figured out he was like, "You know, I've been working on a, a way to figure out, like, between group heritability differences and stuff like that," right?

And I was telling a population genetics friend of mine, I was like, "Yeah, like, you know, he's been working on this." Then my friend, like, blurt out like, "Why would he do that?" You know? And it's 'cause, like, it's really stressful because, you know, they get accused of, like, studying racial differences, right? And they're like, when someone does it, they're like, "Oh. You know? It's like, you know. Or like, I mean, I'll tell you, this is a real story, and I don't really give a 'cause, like, I don't give, give a about Marcus Feldman. But my, my ex So Marcus Feldman, if you guys know, a big population geneticist. He's done a lot of great work. You know, he, pioneered cultural evolution theory, which you know, Pete Richardson and Robert Boyd extended over many decades, and Feldman is really mad about that because he should only do it.

But anyway, he told my former mentor Spencer Wells, when Spencer started, like, go studying human genetics, he was like, "Okay, I" He's like, "You have to say one thing. You have to say we are all Africans over and over and over again." So Marcus Feldman is very, very ideologically like, focused on not having his field, quote, "misrepresented," right?

But I mean, you know, more recently, he's been taking, you know, potshots at Paige Harden you know, co-authoring with a anthropologist at Stanford, and you know, I mean, we all know where he's coming from, and he's a very powerful person with a lot of influence, and you know, he's been able, you know, like, like Knut, he's been able to prevent the water from rising for a long time, right?

I think, I think the we're near the end of that. That's what I would say. We're near the end of that. The water is gonna rise, and people are gonna have to figure out how they deal with it.

Steve Hsu: All right, let's take another question. Anybody have a question?

Razib Khan: Yeah, someone not Yamnaya girl.

Steve Hsu: While you think of your question, I'll tell you a little anecdote. Many years ago, Razib and I were at the American Society for Human Genetics meeting, which is a big annual meeting, I think 10, 20,000 yeah, yeah people come all, from all over the place. It's about

Razib Khan: 10 to

Steve Hsu: 20. Yeah. Yeah. For this meeting. And it was the year of the star-shaped phylogeny. And I'll let it, I'll let Razib explain what that means, but we were going through the poster session, I don't know if you remember this story.

Mmmm. But you and I were walking through the poster session. Oh. And there, and there are like hundreds of posters. Yeah. Like, a whole, like, big meeting hall room is just full of posters. So, so if you don't get a big, like, a big parallel session talk, then your research is on a poster. Yeah. And we were looking at the posters, and he ran from poster to poster because he kept seeing more and more star-shaped phylogenies.
And what is what is it, what is a star-shaped phylogeny? Yeah. What does it mean?

Razib Khan: So basically, like, obviously when you do a phylogenetic tree, you see this, like, branching tree. It's got this, like, regular pattern, right? And the regular pattern is generated through mutations that accumulate through regular process of, you know, well, mutation and reproduction, like generation by generation.

And so you just, like, look at the number of mutations, and that like, you know, defines back to the common ancestor. But a star-shaped phylogeny is so when you look at the pattern of mutations, there's a point where there's an explosion. And so it's like you have, like, a common ancestor and then you have, like, one step away, all of these people.

And that means that there was, like, a massive population spike in the lineage. And so a bunch of the Y chromosomes in particular, human Y chromosomes, are star-shaped phylogenies. About, like half of the males in this room are, are, are part of a star-shaped phylogeny lineage, right? And that basically means that it's like a super male lineage that exploded, at, at some point in history.

I mean, the most famous one is the Genghis Khan. I think that's, I think it's 03. It's a branch of 03. And it's basically exploded, like, 1,000 years ago. I think we know why. But there are

Steve Hsu: no, no, I want you to say why. Sure. So, so, like, I think everybody here knows, but, but just for the average listener

Razib Khan: Well, I mean, what, what

Steve Hsu: What is going on? Well, it's like, you know How can one male

Razib Khan: Well, what is good in life, right?

Steve Hsu: Yeah. But how can one male but I mean explode into a circle of descendants? What, what's going on there?

Razib Khan: Yeah, I mean, you know, so, like, is

Steve Hsu: it I mean And, and how's your star-shape plan going?

Razib Khan: You know what? Like plead the Fifth.

Steve Hsu: No, but, but the, the science part. Tell us the science part. Yeah, the science part.

Razib Khan: Over generations, obviously, like, small differences. You guys understand compound I think everyone in here kind of understands compounding. So, like, you know, moderate differences, like, like a, like a, like a selection coefficient of 5 to 10%, for example, like at a locus, is really strong, right?

And so when you have, like, even moderate reproductive skew, it's gonna amplify so that these elite lineages that have social cachet, social status, are going to, like, sweep through the population very fast, and so that's what's happening with the star-shaped phylogeny. So with the Genghis Khan, you had, like, the golden family, which are the descendants of Genghis Khan, like the Borji the Borjigin clan of the Mongols, and it seems like all of the, all of the people that are direct paternal, like, all the males that are direct paternal descendants had extremely high social status for many and that is actually true. They had really high social status for many centuries, to the point where even having a maternal lineage, I think, like, Timur had the maternal lineage even having the maternal lineage became prestigious, right? And so this is probably what happened 4,000 or 5,000 years ago where we see a lot of star-shaped phylogenies, actually.

So a lot of the coalescence, a lot of the common ancestors of star-shaped phylogenies go, go back to about, like, 4,000 to 5,000 years ago, going back to Yamnaya girl. Like, several of them are Indo-European. They're clearly Indo-European and which, yeah.

Steve Hsu: So again, just asking, just trying to get it to the point yeah where the average listener can understand. Yeah. These Yamnaya guys, drinking a lot of milk. They have wagons. They have wagons. They may be maybe they're warlike.

Razib Khan: Maybe.

Steve Hsu: Okay. They invade, they invade a new territory, and yeah this is my guess. Tell me if this is not, you know, the ancient DNA science doesn't support this. They kill all the men. They take the women.

Razib Khan: Yeah.

Steve Hsu: And one Yamnaya guy's maybe got a lot of women.

Razib Khan: Yeah, yeah, yeah

Steve Hsu: and he fathers a lot of kids. Yeah

Razib Khan: Yeah.

Steve Hsu: And the, the elite Yamnaya families are reproducing yes in this way, by crazy, and their Y chromosome is spreading like crazy. Yes, that's what we see. And when you and I saw this on the posters, you knew exactly what it meant. But the people, the people there didn't, they didn't think, they didn't translate the, the s the data into this bloody

Razib Khan: I, I was just like, like, "What the fuck do you think is happening?"

Steve Hsu: Yeah.

Razib Khan: You know? It's like, okay, so the Yamnaya, also the well, the Corded Ware, which is the daughter population of the Yamnaya you know, like, you guys know like Thor's hammer, you know, they have that the you know, the Mjollnir, like little, little pendants.

Well, they the, the Corded Ware, like, their pendants were maces, okay? And so what do you do with a mace? You smash skulls, okay? So these are not gentle people, you know? Yeah, well, so what you see is like, so for example, in Poland, the dominant Y chromosome was I2, and I2 goes from, like, 80 to 90% to, like, 5% within two centuries.

Like, I2 males disappear, right? But what you see with the Y chromo with the mtDNA, mitochondrial lineages, they don't change as much. So it seems pretty clear that the Corded Ware males assimilated a lot of, it was actually Globular Amphora, which is a particular last Neolithic culture, it was a climax Neolithic culture of Poland, Eastern Europe.

And so the, the fusion of these two you know, populations. So Corded Ware the pots, it's, it's named after the pots, and the pots have like, these patterns and the hypothesis is the women, the women, the Neolithic women were the potters. And these men, they had baskets that they traditionally had when they were nomads, and they liked to see that pattern on the pots, so that's why they did the corded patterns, right?
And so you see these cultural syntheses happening. One other thing is, like you know, there's the Indo, there's the pre-Indo, per- pre-Indo-European substrate hypothesis for Northern Europe, where there's, there are all these words that are not Indo-European that are in these languages. So apparently, the word lady does not have an Indo-European etymology.

Steve Hsu: Hmm

Razib Khan: right? And so I think you have this pattern, and, like, there are there are sites in Germany I think from a single grave culture, where you see that the men have all the same Y chromosome, and the women have all different mitochondrial lineages, and also the isotope analysis indicates that a lot of the women are from very different locations, so they're patrilocal.

And also some of the women, when you look at, like, the isotopes, it does seem like they had a predominantly vegetarian diet. And so what you're seeing is the archeologists are capturing the assimilation of neolithic women into these agropastoralist bands, right? So we we're seeing through archeology and ancient DNA you know, a historical process you know, in terms of, like, you know, Northern Europe was born in blood, like a lot of the world, right? So

Steve Hsu: Okay.

Steve Hsu: Next question does not have to be about ancient DNA. Sure. It can be about Yeah. So you know.
Audience 5: If we assume that demand for preimplantation genetic testing is going to increase can IVF services

Steve Hsu: I so I'm not sure I understand the full question, but I think the first thing you said is if demand for polygenic screening of embryos yeah goes through the roof, can IVF actually keep pace with that?

Audience 5: Yeah. Can, can IVF provide

Steve Hsu: Yeah. I think that's not a problem because actually IVF is actually very scalable and fundamentally pretty inexpensive. It's basically some hormone shots and then an egg retrieval and then some lab work to you know create the embryos and then maybe biopsy the embryos for the embryo selection.

It's fundamentally very inexpensive. The cost in Asia, like, if you go to places like pretty advanced countries like South Korea and Taiwan, IVF is way less expensive, maybe a factor of two or more less expensive than here in the United States, and their success rates are comparable. So it, it is very, very scalable.
It, it just has to get to the point, like you could imagine the government of China saying, "Hey, we wanna have IVF at scale or egg freezing at scale," and believe me, they will make it very economical and, and and do it at scale. And train

Audience 5: staff and go, these things can 10X in

Steve Hsu: I five

Audience 5: years or

Steve Hsu: two years or I, well, I, yeah, I think they can. However, I question the premise that we're gonna see such rapid adoption of polygenic screening. I think it's gonna take longer. I mean, I wish it would go your way for, you know, even for pecuniary reasons, but, but I don't think it's gonna go that fast.
Audience 5: And I suppose the, the average customer or patient might change as well. They might be younger, more fertile, and you might see larger effect sizes in terms.

Steve Hsu: Of, yeah. So we, we are seeing more young people who elect to do IVF for the purpose of selecting their embryos, so that's a new thing. Younger women who go through an IVF cycle will generally produce many more eggs.

We've had cases where a younger woman would produce like 30 eggs in one cycle. We have billionaire-type people selecting from 100 embryos for their child. So that, that's the extreme limit of, of what people are doing right now, but that could become normative eventually. I don't think it'll become normative in 10 years, but it could become normative. Yes.

You can have this as, you know, open-ended as you want. But you guys know a lot of very informed people. What, what do you guys kinda think where AI is going You know question is: Where is AI going? On this one, I don't think the inside knowledge is that different from, you know.

If you follow X or you follow you know rationalist people writing about AI like SV, I think they're pretty close to what you would hear from top researchers or people in the labs. Now, having said that, there's a wide variety of projections for what we'll see. Some people really believe in fast takeoff.

Some people literally believe in 2027, which I think is insane. Other people real researchers who have to actually build the capabilities that scare everyone, those people don't think it's gonna happen that fast. One, I think, piece of information I don't think is widely known, so I just came back from Beijing, and they're very advanced there in robotics.

The problem of how to construct a world model where the, the robot itself by looking at some 2D image constructs a three-dimensional physical representation of the world, that's a completely unsolved problem, and that's not something LLMs are good at. Even vision models are, are not good at that. So there's an unsolved problem.

It's sort of like, oh, big, big breakthrough has to happen before robotics really can, you know, reach science fiction levels. Having a robot which is glued to position and just does a repetitive set of tasks, that's quite different from having a robot that sees the world and knows what's going on and does stuff, and I think there's actually quite a big gap right now between those two things.

Razib Khan: Yeah. I haven't, like, given it so in terms of, like, AI my P doom has been dropping, I will say, over the last couple of years. I work at a tech company you know, and everyone uses AI. Actually, one of my, one of my employees, you know, colleague, Justin over there you know, maybe he can offer his opinion.

But I think AI is a tool. It's been great. Obviously, like, vibe coding and Claude code and all these things are like, you know, whatever. You know, they've changed, they've changed the game a lot of ways. They've made a lot of people more productive. They've also made other people less productive.

I think we're really, really trying to figure it out right now. The main thing that is you know, my intuition is just like, I'm like, "Oh, okay, they don't really have agency." Okay? So I'm not honestly, I'm just not as scared as I was there are people in group chats who are saying in the fall of 2022 that they think that like humanity, I mean, people that are I see around here, that humanity will be over in six months. Obviously, it's not over in six months, right? Some of these people do not I mean, I remember what they said, right?

But they're still dooming right now. So I'm just saying it's like maybe your model of the world isn't, isn't totally accurate, right? Because I did listen and I did remember because I was like, "Yo, six months is not that far away." And they were like, "Yeah, I'm pretty sure that the humanity will be over," you know?

Steve Hsu: I do want to say I take the doomer argument seriously. The timescale is hard to predict, but I think in my lifetime, for sure, 100%, we will have superhuman level intelligence. And if we want to give them some kind of agency or, or goal orientation or long-term planning, we can already kind of do that through kludgy, kludgy methods.

But I think through some improvements in the, in the architecture of these neural nets, making them a bit more maybe like our brains, we can do it at a more fundamental level, and I think that's definitely gonna happen in the coming decades. So, so the, the long run risk from this is that we will create an alien intelligence which is superior to ours, and I think that absolutely will happen.

So I'm not dismissive of these doomer you know, concerns. I just maybe don't have my timelines aren't quite as short as some people's Yeah, I

Razib Khan: I, I agree with that. Yeah.

Steve Hsu: Question.

Audience 6: Can you give some examples of personality traits that you think are heritable, and how heritable they would be?

Steve Hsu: Oh, yeah.

Okay. We that, great question. So when you measure

Razib Khan: some, right. Personality, personality traits that are heritable and how heritable
Steve Hsu: they would be Thank you. Thank you. Yes. Yeah. That was the question. So big five personality traits, things like extroversion versus introversion, agreeableness versus disagreeableness conscientiousness neuroticism, neuroticism. Okay. There's, those are the

Razib Khan: Extroversion, introversion yeah

Steve Hsu: those are the, those are

Razib Khan: Openness

Steve Hsu: Yeah. So those are the big five traits, and a problem for psychometricians is how do you measure them?

So, like, could I just give you a personality inventory, and maybe if you answer honestly, then I can extract your big five traits? Maybe I can back that up by interviewing your friends, and having your friends evaluate you on the same kind of survey, and I can look at the correlations, right? So the test-retest reliability or the, the ability to characterize your personality is not as strong through these instruments as it is for your just general cognitive ability.

So test-retest reliability for cognitive ability is very high. It's, like, over 90%, right? But for these other things, it's much lower. When you estimate the heritability of something, you're asking, say, like, "Oh, how similar are these two twins on extroversion even though they were raised in different families?"

Well, because your instrument is noisy, it doesn't seem like they're as similar. They might be actually very similar, but the instrument is noisy, so you often get somewhat different scores for them. Once you correct for that unreliability of the instrument, the implied heritability for these behavioral for these big five traits is almost as high, I think, if you look at the data, as cognitive ability.

So I, number one, that means that if you did have large data sets, you could probably find polygenic predictors for all of these traits. We don't currently have that data, but we, we could have it in the future, and then we could select for it in embryos

Razib Khan: Yeah, I think the measurement is the big thing that I've, you know, our friend James Leigh, our mutual friend James Leigh, who's a psych a psychometrician at Minnesota, you know, he always said that, you know, the heritability of personality is he said, you know, probably higher because intuitively we see, like, for example, you know, I have a daughter, and you know, her personality is quite similar to mine.
So, you know, when you see like a 0.3 heritability or 0.5 heritability for the big five, like what's going on here, you know? So I think, you know, a lot of it is measurement. I think personality is quite heritable. A lot of like even the way people speak, the tics.

Steve Hsu: Yep.

Razib Khan: You know, like affect is quite heritable. You know, I don't know. It's just

Steve Hsu: Yeah. One, one thing that I didn't know this when I first started in doing stuff in genomics, I kind of learned it by accident by looking at a lot of embryo genotypes and stuff, is that the number of recombination breaks that happen in meiosis is actually quite limited.

So when you actually look at an entire chromosome, there might be only one or two breaks on that chromosome. So you're the child is inheriting huge chunks of DNA from each parent, and that's why you say people would say things like, "Oh, Steve, your daughter has your nose." It's 'cause like she got a whole chunk of polygenic variation that affects the nose shape, and she, you know, it, she's got my nose.
So similarly for personality, you can imagine like, like big chunks of personality could like be transmitted this way, um, to the, to the children. And so yes, eventually we'll be able to we'll be able to select for extroverted kids if you want them.

Razib Khan: So I will follow up on the recombination thing. The number of recombinations from males is lower than females. And so the chunks are gonna be larger. So my My daughter I've already said this, so it's sometimes my ex-wife gets annoyed when I talk about my kids. But my daughter is 30% my dad, 20% my mom. Okay? And that's because the recombinate, there's a, there's high variance in what gets transmitted, especially from the sperm.

The, I think the recombination units for women are, like, expected value is, like, 35, and I think it's in the low 20s for males. Okay? And so it's those low 20s, recombination means that the more recombination, the more it shuffles, and so that'll actually reduce the variance. And so lower number of recombination means higher variance in terms of from the paternal and maternal units on each side, right?

And so you know, if my, if my daughter is 30% my dad and 20% my mom, like, from a polygenic perspective, from a polygenic characteristic perspective, you would assume she's much more like my father in many ways, which actually she is, et cetera, et cetera, right?

Steve Hsu: Yeah. The, the chunkiness, though, refers to, it's not, like, poly like, you would think from polygenic variation that there's some blending going on.

Like you have an, your nose or your, the length of your, the shape of your forearm is, like, a blend of your mother and father. But actually, it doesn't work that way. You can often get, like, a big chunk which looks just like Dad's. Your ankle looks just like Dad's ankle, right? And that's because of this chunkiness. So it's a little

Razib Khan: genetics is discrete.

Steve Hsu: It's a discrete process. It's a little surprise was surprising to me because I thought, oh, recombination is, like, very more of a fine thing, but it's actually not. It's, like, huge chunks of chromosome that, that recombine. So question?

Audience 7: I think I heard you on your Polygenic screening is bottlenecked by data. Would you say a little bit more about what types of data?

Steve Hsu: So question's about what bottlenecks advances in polygenic research in polygenic prediction, and I've said in the past that it's mainly data limited, and I'll explain in what sense it's data limited.

The algorithms are good enough that the real problem is the availability of data. When we do when we build predictors with varying amounts of data, training data, we can actually extract the rate of improvement. It's a, it's scaling, what they now call scaling laws for AI, but we already were doing this in genomics.

We can see how much, excuse me, how much better the predictor would be if we had, say, an extra 100,000 cases of lung cancer. We could predict lung cancer that much better, okay? So we, we can actually make pretty precise statements about how the data limit works, right? The sense in which we have limited data is within a particular ancestry group, say, people of European ancestry, we still might not have enough genotyped people who had lung cancer, and we, we can see the slope is non-zero.

We haven't saturated yet on prediction power for that trait in the data that we currently have available. For height, it's kind of saturated. Like, we can't really we're not really getting big gains in height. We're down to, like, less than a few centimeters or less error bars in our predictions, which is similar to the difference in height between two identical twins who were raised apart.

Now, but the other bottleneck is most of the data is Europeans, European ancestry people. The next largest set is East Asians, so we now have East Asian predictors, which are comparable to the predictors for people of European ancestry. But other groups, South Asians, African Americans, Africans, Indonesians, you know, all those groups, we do not have enough data. Those countries have not built massive national, you know, funded by national governments biobanks, and therefore we, we just don't have the data to build the accurate predictors.

Audience 7: Is the commercial incentive, what is the commercial incentive for datasets and making decisions

Steve Hsu: It's, it's not really a I mean, so far it has not been a commercial incentive. It's just governments funding research actually. And so, you know, the, the, the biggest thing you could do if you, if you want to bring this technology to South Asia is get the Indian government to fund like at least a one million person biobank with good medical records, good phenotyping. India might be harder than other places There's a lot of genetic variation between Yes.

India might be harder than other places because of caste and just the genetic structure could be quite complex, and Razib knows a lot of more about it than I do. But basically there, those are the two different types of, of data bottleneck that we face.

Razib Khan: So I was, like, banned from tweeting it, but I could probably, like, say this. There's a med archive paper about you know, like, bullshit that geneticists have to deal with about like you know, some Indian genome consortium, and, like, the labels for the Indian populations were, like, retarded, and, like, I couldn't figure out. I was like, "Okay, there's three population of Kerala.

They're, like, one, two, three. What the are these?" So apparently one of the members of the consortium was like, "We cannot put any ethnolinguistic and caste labels in these. So they were all redacted 'cause of one group, right? It's like, anyways. Yeah. I couldn't tweet that out, but, like, I can say it.

Audience 7: So for the last follow-up, so is it true that the hardest things to make progress on are traits that have a low base rate and are polygenic, so the, the hardest thing to connect to

Steve Hsu: Yeah. So, so rare diseases. If the disease only affects 1% of the population. Yeah and you have a million people, you only have 10,000 cases. And so often the bio banks will recruit in an enriched way to get people who have certain conditions to increase our power. So that, that definitely is an issue.

Razib Khan: Okay. Is LD a, an issue in Africa, or is it just like, are we doing whole genome now?

Steve Hsu: I think we're still at a pretty early stage in actually understanding the population structure within, like say within sub-Saharan Africa. Yeah, yeah. I, so yeah. I, I think it's

Razib Khan: Well, so I guess what I'm alluding to here is, like, there, there, there's the issue obviously is just, like, your, your, your training set of variation in the population. But then one of the things within Africa in particular is because all non-Africans descend from a bottleneck population 60,000 years ago, much more homogenous, whereas within Africa there's a lot more variation, which means that you need, like, a higher density of markers to get the same amount of prediction.

So it's not just, like, the population size of Africa. In Africa you need a certain marker density. Yeah. And that's not always

Steve Hsu: Yeah, absolutely. Yeah. So, so if you look at PCA maps of the world, like within Africa, there are orthogonal dimensions that you can go into that are like with huge amounts of variation.
So you can like, "Oh, all the, all the other people who left Africa are kinda like this little blob here," and then, like in this other direction is the variation within Africa, and it's like, wow, I it's gonna take a lot of data to deal with that. Next question.
Audience 7: Outside the chromosomes, how much is there information which is heritable in people?

Steve Hsu: Well, there's a mitochondrial DNA. And then I don't know what you mean by outside, but there's, there's methylation, there's, there's epigenetic effects. I think that, you know People are especially anti-hereditarians are very invested in epigenetics, and epigenetics is a real thing, but I think the actual variance accounted for by epigenetic stuff is still dominated by just what's in the standard DNA. That's, that's my view. I think that the data sort of supports that. I don't know if you have yeah

Razib Khan: yeah. So I wrote a, a Substack piece that Luca is it Mangioni? Like the guy who killed us. Yes. He really liked it.

Steve Hsu: Really? Yeah. Wow. I didn't know you talked with him. Yeah, he

Razib Khan: really liked it. It's in the emails. So he really liked it.

Steve Hsu: Wow. He reads your He didn't tell him to say that he reads your, he reads your stuff. Nice.

Razib Khan: I don't know.

Steve Hsu: Nice.

Razib Khan: But you can look at it. Like you can look it up. But anyway, so basically epigenetics is really important 'cause epigenetics basically means like how is it, how is it that, that, you know, your liver tissue has the same DNA as, I don't know, like the lining of your stomach, but like the tissue is totally different?
Well, it's like, you know, differential gene expression, like the genes are turned on and, you know, whatever.

I don't wanna talk about alternative splicing. There's all sorts of details that are going on how the genes are expressed. That's epigenetics. What people are really obsessed about though with epigenetics is Neo-Lamarckism, which basically means that like, you know, like I punch Amish guy in the face, that causes like some cellular tissue, like insults, and like that makes an epigenetic mark, and somehow that's transmitted to his offspring, right?

In humans, I don't think that happens. Okay? So just Google Razib Khan epigenetics, like, you know, this was a couple year. I mean, maybe. Also, there is a lab that has in okay, this is. Again, like since this is like personal communication, quasi personal communication, I can say things like this. This is not in the like published record, but there was a, there's a lab in like Washington State, I think it's in Wazoo and they always publish these like fish studies about epigenetic inheritance, and like, you know I'll tell you, a friend of mine who did a lot of epigenetics works in plants, which that it's a thing in terms of heritable transmission, and my friend was like, "Yeah, nobody has been able to like replicate that work in any other lab." So basically he's like, "I don't think it's true."

Steve Hsu: If you like, I

Razib Khan: think it's fraud.

Steve Hsu: If you, if you write a paper that demonstrates some very strong epigenetic effect, it'll be in all the media, it'll get retweeted a million times, Nature will write like a commentary on it, and then like good chance it doesn't replicate. It's just because there's a very strong desire in this field of science to discover that, so that we we're not crushed by the, the, the cruelness of DNA inheritance.

Razib Khan: Which I don't Right? So I don't think it's cruel, but whatever.

Audience 2: Yeah. You know? Well, they think, they think it's cruel.What are your biggest critiques of academia, and what reforms would you wanna see? Like limited to three

Steve Hsu: What are your biggest critiques of academia and what reforms would you like to see? You, you, you go first.

Razib Khan: I mean, my, my, my biggest critique is, like, all the you know? It's like It's like so many, so many you know what I'm trying to say? It's like, I don't know. It's like you okay, like the reality is, like, you know, like there's all these justifications for tenure, but it's bullshit because, you know, you need to collaborate, you need to have your graduate students be postdocs in someone else's.

So it's always everyone scratching everyone else's back. Sometimes it's ideological. You know, there's like for, I'll give you an example. A lot of, like a fair number of graduate schools followed, like, the anti-testing craze in the late 2000s. It was retarded. Me and Steve talked about how retarded it was.

So it was called GREXIT. They got rid of the GRE so they could let retarded people in to graduate school. And you know, one of my friends, who's a lib, but you know, he started criticizing this on social media and his wife in a different department was rebuked by her, the chair of her department because they were offended that he was criticizing that they were dropping the GRE.

Like, this is like how, it's like contagion. It's like yeah "Oh, your husband can't be criticizing this," and it's 'cause it's like the progressive thing to do is to get rid of it, right? They got rid of it, and now they're having to get it back because it was stupid. I mean, it was range restriction conditioning on the collider.
Those of you who know what I'm talking about here, these are, like, known things. But like they were just, they were cowards, right? And there were, there were people who were, people who knew that the GRE was useful who voted against it on in faculty meetings 'cause they didn't wanna be seen expressing their real views.

And so I don't know what to not all academics are like this, but, you know, a lot of them are. There's a lot of conformity, and it's, like, really frustrating because, you know, they're interesting nerds, but it's just, like, agreeability's a little too high.

Steve Hsu: Yeah.

Razib Khan: That's what I'm saying. For a big 5%. I think agreeability's a little too high.

Steve Hsu: Yeah, I, I think, you know, people don't really make use of their tenure. So even though you get tenure or you become a full professor or whatever, there's still always, like, your salary raise, your NSF grant, you know getting appointed to some blue ribbon panel or there's still, like, little things on the treadmill that you want.

And so it just enforces a high level of conformity in academia. And he's talking about politically sensitive things, but it could even be even within science. Like, if a bunch of people are saying, like, "Hey, the emperor has no clothes. We know damn well there was natural selection over the last 10,000 years among humans.

Oh, but I'm afraid to say it." And only, like, only when Reich becomes super prominent for, already for his Neanderthal work, then he can get away with publishing on natural selection. So even in the realm of ideas, people are afraid to deviate too much from the herd. Part of it is by the time you slog through the whole academic system, like, oh, you got a PhD, then you did a postdoc, then you were a junior faculty member, and you got your first big grants, and then now you're tenured.

By that time, they've kinda selected for conformity, okay? They've selected for conformity. So departments in universities could be zany fun, like Manifest, like this, but they're not. And it's only, like, the, the professors that are here, there's only a handful of professors here, are, like, outliers, like, total weirdo outliers in their own departments that they would go out of the box enough to come to Manifest, and it, it's kind of a shame.

I think that's, like, the most disappointing part of academia. To say a positive thing about academia, to have, like, someone who's, like, like Andrew Wiles, who's gonna go into his attic and work for seven years to prove Fermat's Last Theorem, that's a you know, I think some people called it a triumph of the human spirit, not just, like, mathematics, but a triumph of the human spirit, 'cause nobody had any idea how to prove this theorem, and he went in his attic and for seven years had this, you know, stamina to do it.

Academia still has that, right? Long after everybody in Silicon Valley is done chasing their IPO or their big exit or whatever, what the acqui-hire, you know, those guys are still working on hard problems, and they believe in the knowledge that they're accumulating. So I wouldn't be that quick to throw academia in the toilet, at least some parts of it.

Audience 5: What would your reforms be?

Steve Hsu: Wow. What would my reforms be? See, we were trying to dodge your hard question. What would our reforms be? Do you have any reforms? I

Razib Khan: mean, well, I mean I think, like I mean, I think probably, look, a lot of Europe doesn't have tenure. We don't need tenure. Like, get rid of tenure.

Steve Hsu: Bold.

Razib Khan: You know? It's just like it's like they, it, they don't use it anyway.

Steve Hsu: After I'm done, then you can

Razib Khan: get rid of tenure. Yeah, yeah, that's we, we, we could, we could grandfather it in.

Steve Hsu: The last yeah. We

Razib Khan: could grandfather it in. Yeah. But I mean, the, the reality is, like, you know, a lot of American academics, they're obsessed with their tenure clock.

Steve Hsu: Yeah.

Razib Khan: And then once they're obsessed with their they get the tenure, they're like, "Well, then I have to become full professor."

Steve Hsu: Yeah.

Razib Khan: You know? And then I have to do this. So there, there's these so like get rid, at least get rid of tenure. Yeah. You know? So like they're not obsessed with that.

Steve Hsu: I'll, I'll say something very controversial, so which, you know, if they're ever interviewing, interviewing me for, like, presidency of some university, and then some faculty member finds this quote, it would get me in trouble.

But different disciplines in the academy have different levels of depth and contribute differently to the wellbeing of society. There are unbelievably radical differences, okay, in this. But we're because there's politics within the university, you don't wanna be the guy who says like, "That department is mostly bullshit yeah and we should get rid of it." Once it gets to the point where you have a department of X, it's very hard to eliminate that department, okay? It's just name names. I'm not gonna name names, but well, I okay, I'll give you, I'll give you so when I was in my administrative position, I was talking to the dean of the ed school, okay? Nothing against education. It's very important. We all care about, we all benefited from education. We love it. We

Razib Khan: love the educators. We're

Steve Hsu: highly educated. We love

Razib Khan: the educators.

Steve Hsu: Yeah. I was, I was talking to the dean of the ed school and I said to her, "Well, you know when they look at the chip industry, they have this thing called Moore's Law, and the chips get like kind of 2X more powerful every 18 months or something like this, right?

And you know, that technology has gone like this. But during the same period of time, batteries barely improved. They improved a little bit, but not a factor of a million or 100 million the way chips did." So you, you can compare like okay, through no fault of their own, the battery researchers were not as impressive, did not achieve the same level of results that the chip semiconductor industry made, okay?

So I kind of explained that to her. She didn't know what I was talking about, but after a while she understood, okay? Like, "Your battery's not really that much better, right? Remember when you were a kid and you had a flash." Okay. Then I said to her you guys. Then I said to her, then I said to her, "Okay.

Suppose I want to teach French to a 15-year-old kid, or I want to teach calculus to an 18-year-old kid. Can I do that even 10% more efficiently now than in the 1950s or 1960s? The answer is no. Zero productivity growth from all the work you guys have been doing. Zero pro, right? Batteries chips, ed schools. Ed schools are this, right?

Razib Khan: No, no, negative. That's negative.

Steve Hsu: Well, it could, it could be. I was being nice. Yeah. I was like, I even said to her, I said, "Well, I'm sure we've discovered some better pedagogical methods for teaching the derivative," and, you know, we haven't. But But, but I was even giving her that, and I'm like, "Well, how much faster

Like, okay, in, in, in a, in a semester we can now cover 10% more material in the calculus class than we could back in the '50s 'cause we have better textbook." You know, like, okay, whatever it is, it's, like, negligible. And so I actually explained this to her, and she's was, she was like you know, she had to think, you know, she had to think for a long time. But

Razib Khan: Slow, slow CPU. Slow CPU

Steve Hsu: but, but she understood what I she understood, and then the next time I met with her, she was like, "Well, I thought about what you said, and, like, you, you definitely have a point." And she said, "But, you know, we do other things." And I'm like, "Okay, I know you train teachers and stuff, but the research side of it, what really, what is the show me the productivity of the research." We don't a company will ruthlessly cut a division that's not contributing to the bottom line, not contributing to the mission of the the, the bottom line of the company, but universities just keep these departments around. You can't get rid of them. And they'll defend themselves like crazy politically within the university system.

And we just can't fix it. So once you, once you have a department of X, you're not getting rid of department of X, no matter how unproductive that direction of research is, so.

Razib Khan: Wait, wait. No, let's, let's

Steve Hsu: let's go to This man, this man has his

Razib Khan: hand up yeah Vincent. I'm sorry.

Audience 2: The follow-up to that would be how optimistic are, are you on parallel institutions that are doing research with

Steve Hsu: Well

Razib Khan: we There are a lot of institutions that are doing research, not all of

Steve Hsu: them Oh, God, man, I, I'm totally forgetting to do this. You, you wanna take a crack or

Razib Khan: No, you go. Okay. You go.

Steve Hsu: So we have national labs, which are really incredible. Corporate research is never gonna allow Andrew Wiles to go in the attic for seven years and do his thing, right? A lot of fundamental things, like even the basis, like for example neural nets, deep learning, that came out of the academy.

It didn't, it didn't come from Apple or Google. Okay, transformer paper came from Google, but deep learning came from the academy. So I don't think there's actually a substitute for our research universities, certainly not at the moment. There are a lot of people, like, who are very based in Silicon Valley who are just like, "Oh, we could just kill all the," and they're doing it right now in the Trump administration.

Yeah. "We could just kill all the universities and there would not be a loss." I don't think that's true. I do not think that's true.

Razib Khan: Yeah. No, I academics are pussies, but I do still really enjoy when I talk to my geneticist friends there is there's kind of like a bliss kind of like this like, you know, you're just, you're just rapping about your field. And you don't get that outside of Yeah the ivory tower. That's just, you know, I

Steve Hsu: don't know. I mean Reich, I mean, Reich, David Reich is at Harvard. He's in the academy. He's got an academic lab. He's got grants. Yeah. And all that research came from that effort. Paabo's at Max Planck Institute, which is a, a, you know, very generously supported by the German government.

It doesn't generate profits for anyone. But you do need some people in the monastery who really believe in what they're doing, but you want them doing stuff that's effective, that actually works. If, if after a while you figure out it's not working, you should kind of get, like phase out that department, right? So

Razib Khan: Well, one, one thing that I will say real quickly is the way modern academia is structured is, like, it's post-World War II, so this isn't like eternal. This isn't like an eternal institutional setup for the production of new knowledge. Yeah. There could be other ways and you know, we probably will have to modify it because, you know, one thing you know about institutions is they get calcified.

You know, Goodhart's law. There's all these other there's all these principles that come to play when an institution has continuity, and it's just like you sometimes you just have to reboot. So you gotta reboot the computer. So Question back there?

Audience 2: Is the centralized federal funding of research resulting in ideological capture of the fund funding dollars through the you know, the, the people who are more senior in the Yeah departments than others?

Steve Hsu: So the question is, is, is it the the, the fact that most of our research dollars originate in the federal government, is that the response is that the causal reason for the ideological takeover of, of our campuses?

Is that a fair rephrasing? I would say that's not the problem, actually, because the ideological capture of our campuses came mostly from departments that don't get grants, that don't get federal research grants. So sociologists critical theory people

Razib Khan: Studies

Steve Hsu: studies people.

Razib Khan: Like, if it has studies in it, they don't study.

Steve Hsu: Yeah.

Razib Khan: What If it has science in it, they don't do science.They

Steve Hsu: Don't do science. What, what really happened, what really happened is that a group of people came into the university who felt that their ideology was the most important thing, and the university should be a vehicle to advance their ideology in society, and that's a very powerful thing because everybody else, if I have to go in my lab and stay up all night and take a shower in the physics building to go teach my morning class, I don't have time to around with campus politics.

Yeah. So you'll find the best scientists, the most serious engineering professors, they're not, they don't have time to play these political games. But the people in the social sciences and humanities, they have time. Yes. Okay? Again, very unpopular thing to say, but it, it's actually, that is what, how the ideological takeover of our campuses actually happened.

So the question from Dr. Matt Southie, the leading expert on Nick Land's philosophy, PhD Rice University is, is concerning, you know how is it that the ideological takeover of the humanities has affected the rest of society or the rest of the university? Well, within the university, it's political, so they, they're activists.

One rabid dog activist can, can herd a whole group of sheep, right? Because sheep don't wanna fight, they wanna do something else, and yet, and like, you have a rabid person in your department, a berserker, as my friend Ron Unz refers to them that person can cow a, a lot of sheep and herd them, okay? So that's within the campus.

But you have to remember, what fraction of American students study STEM, like 20%? So, so everybody else getting edumacated by the continental philosophers and the, and the English professors and the studies people, and yeah, they come out learning all this stuff. I was taught as a student race is a social construct. I grew up, I was a kid in the '70s.

Razib Khan: Yeah. It's, it's very hard to, to just administratively to get rid of a department. The, all the alums of that department will just start lobbying the university right away, that the department head will write out an email to all their alums and say, "The August department of XYZ, the, those damn bureaucrats are trying to, you know, get rid of us," you know, and stuff like this.

Steve Hsu: And, and then, like, a million emails go to the president saying, "How dare you eliminate the whatever department?" It just doesn't, it doesn't happen. I mean, as

Razib Khan: I was thinking of, like, a weird analogy as, as I'm listening to and thinking about this. I read a, I read a book it was, like, 20 years ago about just randomly about the Talmud and how you analyze Halakha, like Jewish law in the Talmud.

And, like, the scholar was like, "Well, you know, the way it works is, like, you can only extend the law. So you can make it more complicated, more restrictive, you know." And so, like, there are, there are apparently, like, there are some Haredi Jewish sects that are vegetarian because they don't trust anyone to do the kosher appropriately.

Like, they just don't trust anyone, okay? I'm bringing this up because it's like, okay, like, this is a system that gets more and more complicated . And it's like academia is like that. Like, okay, like Stanford decides to, like, split its anthropology department in two. Oh

Steve Hsu: yes. You know

Razib Khan: like these

Steve Hsu: It's a good example.

Razib Khan: Yeah. Good example. The physical and the cultural. Yeah anthropology split apart. And so, like, you start to have all these, like, different departments that do very similar things, but now they have a culture, now they have a history. And, you know, what is done cannot be undone, you know. It's, it's just how it is. It's, like, just, you know.

Audience 6: Okay, so if academia is all and all sheep, what are your personal most controversial, you said racism's, it exists, and most controversial population differences?

Steve Hsu: Question is, question is can you guys say something that will get you canceled?
But you go ahead.

Razib Khan: Well, 'cause I was talking to him about this. I think, like, you know, I think I definitely have been canceled more than him

Steve Hsu: right? I only got canceled once.

Razib Khan: He got one big cancellation. I got, like, I've had about, like, depending on how you can do the math, like six to 10. Although it's been, like, really low-key since about 2023, so I'm wondering if I'm losing my mojo.

And there was actually a discussion to delete my Wikipedia page last year, 'cause they're like, "It doesn't seem like anything's been happening." Oh. You know? And I was just like, "I gotta do something big." You know? So I don't know what it is, like, you know. But Not the Jews. I mean, usually it's like the race stuff always got, would get me in trouble, but whatever.

I mean, I mean, it was okay, I'll tell you a story, because this person is not in academia anymore, just to give you, like, a flavor of the crap I had to deal with. Just, it was funny, though. So I'm meeting a friend, and and I'll tell you the gender. It's a she so some of you can, like, think about it, like maybe use AI and figure out who I'm talking about.

'Cause it's a population geneticist. It's an eminent statistical population geneticist. And so I'm having lunch with her, and I'm like, "Why did you sign that pledge against, like, Nick Wade and stuff like that? Like, like what do you, what do you, what do you think about, like, group differences in IQ?"

Like, you know. She's like, "Well, they're clearly, you know, substantially genetic." And I'm like, "But, like, you said that it wasn't in that" So she's like, "I know." And I was like, "Well, why did you do that?" And she's like, "Well, because my colleagues would ask me why I didn't sign." And I'm like, "Okay, like, did they sign?"
She's like, "Yes." And I'm like, "Well, what do they think?" "Exactly what I think." So this is like, you know, and I was like, we're still friends. You know, obviously I like her a lot, actually. You know, it's just like I just kind of accept this is how people are. I am, like, a person that's like, if you ask me a direct question, I'm, I'm not gonna lie, you know?

And so that, that did get me in trouble. And, you know. I mean, you know, and I've written about this. Like, I got into a fight at ASHG 2013. I almost got into another fight in 20 Was I there? No, no. I almost got into another fight in 2014. I

Steve Hsu: would've backed you up.

Razib Khan: Yeah, I know, I know. But it was just like, it's just like, it's just because, like, I will not, um, you know.

Like academics, like, you know, most people, they're pretty conformist, and I'm not, and so, like, I don't back down. And, like, then people start to get agitated, 'cause they're just like, "You're supposed to back down. Like, there's, like, all of us, and we disagree with you." And I'm like, "I don't give a what you think.
I think what I think," you know? So I obviously have, like, a, a very, particular type of personality, which, you know, makes it a little hard to work with people sometimes in certain fields, right? But I don't know. Yeah, so there's a, there's a lot of, there's a lot of smart people doing a lot of brilliant things, but there's, there's some blind spots here.

And like, you know, obviously in population genetics I know what the issues are, but, you know, what's going on in physics, you know? I, I don't know.

Audience 7: Like

Razib Khan: you start, you

Audience 7: start to worry. Can you talk about the capture in physics. Oh, just like, has the ideological capture I know it's not as extreme, but how has it affected, like, the hard sciences?

Steve Hsu: Yeah, so the question is how is this ideological capture of universities affecting hard sciences? So it, it used to be the case that so in engineering departments, maybe some econ departments, physics, you could find some Republicans.

That used to be the case, and if you track the numbers it's just been decreasing with time, so that overwhelmingly now you have a kind of monoculture even in the hard sciences. And most of the time it doesn't affect like, if I'm trying to measure some property of the electron, it doesn't matter whether I voted for Biden or Trump.

That, no problem, right? But if it's some tricky thing someone comes to the faculty meeting and says, "You know what? We have no women in our department." And maybe the dean even comes and browbeats us and says, "You know what? You have no women in this department. I'm not approving any more searches in your department until you hire a woman."

That, that literally happened. That was very common in the last 20, 30 years, okay? They they all deny it, but every professor has sat through these meetings, okay? The dean literally comes and says, "I'm not approving any more searches, so when your old people retire, you're not gonna get new lines unless you knuckle under and do this thing."

Which is strictly speaking, I think probably completely illegal, these things, under civil rights law, et cetera, et cetera, okay? But we all lived through it, okay? When that happens, in the old days, if you had some salty Republican libertarian guys in the engineering in the engineering department or the they would just tell the dean to go off, and they would go right to the newspaper and say, "You see what these guys are trying to do?"

So it wouldn't happen so much back then. But gradually, as the department became kind of a monoculture, and yes, these social goals, you know, predominate over, what's this meritocracy thing anyway? It's very subjective. Who's good? Who's bad? And, and they all knuckled under. So this has been going on for decades.

Razib Khan: Oh, wait, wait. Yeah. Really quickly, like unrelated to the question, and like there's a question there from Sean Thomas. I saw something in the apparently, like one of the reasons that like at Berkeley they want standardized tests to come back is there was one okay, there's many quotes. The quote that I thought was funny was like apparently second-year engineering class at Berkeley, the professor had to take time out to explain how to, how you add fractions because they didn't understand why one half plus one third is, you know, five sixths.

Wow. Okay? And so it's like that's, that's where we are when like academics like puss out and just allow stupid things like getting rid of the SAT to 'cause they did. They could've, they, you know Yes, they recommended the academics were not the real problem. It was political people, but the reality is they didn't stick their neck out.

Steve Hsu: They didn't fight.

Razib Khan: Yeah. They get Yeah. They If you, if they really fought, they could've prevented it. They were just like, "Okay, like this is stupid, but you know what? It's not my problem." Well, it's your problem now. Anyway. Question, Sean. Wait, Sean. Yeah.

Audience 9: Do you think the federal government should just step in and cut the fat like they did at the federal level for funding, especially for like historic STEM schools? Do you think they should just cut funding for programs that are sucking funds away and not producing research?

Steve Hsu: When Trump was first elected, I was at a meeting and I met with one of the people at AEI, which is a very conservative think tank and has a lot of influence in this Trump administration. And I met with the guy who does education policy for AEI, the vice president at AEI, and he I told him, I said, "You know, there's good and there's bad.

There's no reason to cut our research on semiconductors at the university, or photonics, or quantum computers because there's some woke people in some of the other department or whatever. Or even all the physicists are woke, but at least they, they're still building quantum computers, right?" This guy said to me, "These guys are our enemies.

These universities are producing our enemies. We're gonna take them all down." And that was the attitude of the Trump administration at the beginning. So they were just like, "Burn it all down." Okay? Then it got more complicated because of the Gaza stuff. So then there was Gaza stuff, and then the thing shifted to like, "As long as you control your students and you don't let them protest Gaza, we can, we can be friends again."

So, so it's, it's gone in a completely crazy people in China would laugh at this because they're building, like, the most impressive scientific machine infrastructure, meritocratic universities, and they're watching the Americans do stuff like this. But that, but that is what is going on

Audience 4: Well, I, it's related to something

Steve Hsu: you

Audience 4: said. Okay. Polarization of faculty. Yeah.

Steve Hsu: I'll just repeat your question. Yeah. The, the question is, maybe it's the case that the Republican Party is so dumb that basically no professor would really endorse

Razib Khan: Retard right

Steve Hsu: what yeah, what they're up to Trump and therefore that explains the secular decline in

Audience 4: Rep Explains a large part of

Steve Hsu: it. Yeah, a large part of it. That could be true. I mean, I'm not, I'm not gonna I'm not gonna lie. I will point out that in recent years, in order to get hired, for example, at the University of California, many departments or to get a federal grant, you had to write things in your essay, in your statements, that no self-respecting person on the right would actually write.

Razib Khan: Yeah.

Steve Hsu: So, so they were literally filtering for ideology at these at universities and also even NSF grants. You had to write stuff that you literally didn't believe in in order to get the grant. So we had that filter on. Your effect might be there, but this other effect was definitely there too

Razib Khan: Yeah, I agree, I agree. I think it's both both dynamics.

Steve Hsu: Yeah.

Razib Khan: The Republican Party is getting more retarded. I mean, that's just, like, true at the empirical level. Like, college-educated people are leaving, and, like, you look at the, the influencer like, you know, back, back in the day it was, like, George Will or William F. Buckley, and now it's, like, Matt Walsh. You know? It's kind of retarded, so I get it. You know, it's an aesthetic thing. But, you know, my people, what do I, what am I supposed to do? I don't know. It's just, like, how it is.

Steve Hsu: I voted for Obama. I have to

Razib Khan: Yeah, I know. I remember. Like, we used to, we used to talk about that 'cause he was, he was a shit lib back in the day.

Steve Hsu: I wasn't a shit lib, but I did vote for

Razib Khan: Yeah, you did vote for Obama yeah. You know, it was, it was, it's squish.

Steve Hsu: You.

Audience 2: What do you two disagree about the war?

Razib Khan: Well, I don't know Iran war? Iran and Israel. We, we haven't talked about that stuff.

Steve Hsu: Yeah, we haven't talked about it.

Audience 2: What do you guys agree about it, and what do you guys disagree about?

Steve Hsu: Why, why don't you okay. Question is about the Iran war. If any comments you wanna give on the Iran war, go ahead.

Razib Khan: Yeah, I mean, I think it's kinda retarded. I'm not, like, very happy about it. I don't think Trump knew what the he was doing. I think it was always a high-risk proposition, and I think, like, you know, they assumed that, like, you know, there was gonna be a popular revolution. That didn't happen, so we're right? We're probably gonna go back to, like, something more like the, you know, the Obama deal, but probably worse. That's what I think. And they're gonna end up controlling the Strait of Hormuz. Yeah. I mean, like, we're right? It was stupid.

Steve Hsu: Yeah.

Razib Khan: You know? So I'm not happy.

Steve Hsu: Yeah.

Razib Khan: Like a lot of people.

Steve Hsu: Yeah.

Razib Khan: And I think, like, a lot of people are holding their tongue right now, but that's another thing. I
Steve Hsu: mean,.Just to show you how dumb I am, I voted for Trump because he said he wasn't gonna start a war with Iran, okay?

That's how dumb I am. I thought, oh, Kamala, the Israel lobby will totally roll her, and we will do what Bibi wants. But I didn't, I didn't actually correct even though I am an Epstein scholar, we can talk about that later. I didn't fully understand the depth of the connections between Trump and Epstein.

And so we are doing things which Bibi wanted, Trump said explicitly he wasn't gonna do, most Americans don't think we should do. The Pentagon didn't think we should do this kind of thing. It's a disaster. It's just a disaster. But you know I was talking to a, a real foreign policy, you know, kind of think tank strategist, you know, about this.

He says, "Well, Steve, at least it's not as bad as Iraq." "We, we spent trillions in Iraq. We haven't spent anywhere near trillions on this one yet, so no problem."

Audience 10: So 2006 Greg Cochran, Henry Harpending and the late great Henry Harpending, he, they put out a paper called The Natural History of Ashkenazi Intelligence. And that was basically, so certain neurological disorders, it was assumed

Razib Khan: Swinglehuban, Swinglehuban disorders

Audience 10: Right, right, right.

And is there been any work on that? Is any follow-up,

Razib Khan: any there has, there hasn't been follow-up, too much follow-up. Greg told me like I mean, do you talk, do you talk to Greg?

Steve Hsu: Yeah, I can go you go first, but I'll, I'll give you my take on this.

Razib Khan: Yeah. So no, I'm just saying like, well, I mean, you talk to Greg like, Greg Cochran has been calling me every like two to, two to four weeks since 2005, and like we talk for three or four hours.
There was a period when I was busy with kids where, you know Like it's, the amount of Cochran conversation is definitely like how involved I was with kids or my marriage you know what I'm saying? Like, over the years. But anyway so there was some talk in Israel of, of people looking at it. And then apparently Greg heard that they nixed it because of ideological reasons, like they didn't wanna destabilize the Ashkenazi Sephardic Mizrahi, like whatever.

And so I think the one thing that I will say about that paper is the theory I do think that Ashkenazi Jews, like, you know, their cognitive profile is different, lower visual-spatial, you know, you know, higher verbal. And like, you know, higher overall IQ. I think it's probably like, you know, substantially due to genes, but it's, IQ is polygenic, highly polygenic.

And so their, their paper kind of hypothesized large quantitative trait loci, which means, basically means that like it would be a large effect gene, and I don't think we've detected any of those. And so therefore, I think part of that model has been falsified insofar it's gonna be a quantitative trait it's gonna be polygenic as opposed to these like big, big genes of large effect that they're talking about, right?

And in terms of investigation I mean, I mean, like, you know, like, there's a lot of politically incorrect things you could do, and studying the genetics of Jewish intelligence is kind of, like, up there. So it's, you gotta have some big balls, and, like, nobody

Steve Hsu: They I'm just saying, like even in Israel, you can't do it.

Razib Khan: Yeah.

Steve Hsu: So a little recap for people who are too young to remember all this stuff. So there was a paper, Cochran, Harpending, and who was the other guy? Hardy? Hardy. Hardy. Yeah, Hardy I didn't know. Yeah. Henry Harpending, I will just say, was a very distinguished anthropologist. Yeah. A member of the National Academy yeah of Sciences. Some people here know what that means. And a gentleman, really a great person. Died you know, not that long after, I think it 2016. Yeah. 2016. after this. So I got a copy of that paper from Greg before it was published or released because they wanted my comments on it, okay? The paper is about the following Natural History of Ashkenazi Intelligence, that's the title.

So there's some evidence that Jews are smarter or have somewhat different cognitive profiles than other Europeans. They were interested in the causality, how that happened did it happen due to natural selection? What were the causes? There was history in that paper where they talked about the fact that, you know, Jews could not farm, they were forced into certain occupations like trading and lending money that maybe forced them to use their brains in different ways than the average European.

So they, they speculated about these things. They speculated that because of a relatively tight population bottleneck that the Jews went through, the way that they got smarter in these roles that they were forced into involved the accumulation of some rare mutations that could be IQ-enhancing. Okay? I told them then, and they didn't understand this at the time, that it's almost certainly mostly polygenic variation.

Greg now says that that's what he believes, but at the time he didn't really understand this, just honestly. So but the paper's all about these rare mutations that might be associated with higher IQ. They don't really talk about polygenic variation in the paper. Okay? Subsequently, many of us were interested in this question, so both Greg and also James Lee and I got in contact with some Israeli researchers who were actually interested.

They had medical data on the people that had some of these rare conditions, and the question would be, oh, if you test the siblings, the, you get family IQ data for people who are carriers of these rare mutations, you could establish whether that mutation is associated by itself, that one mutation, with significantly higher intelligence, which is what the main hypothesis that they stated in the paper.

But that, that researcher, I think there might have been more than one that we were in contact with, later said, "We can't do this. It's not politically. It's not, we can't politically. The political climate will not allow us to do this." Razib is saying that it's because this enrichment is in the Ashkenazim population, but not in the Sephardic population, so it would be considered divisive even within Israel if you establish something like this.

So that, that's the history of it. Still unresolved, like many scientific mysteries

Razib Khan: I mean, one thing I will say real quick is we do have the technology, you know, this is like a 1970s reference, so it's, like, lost on a lot of you guys. You know, we have the technology to answer some of these controversial questions, because you guys keep coming back to that.

It's just like, you know, we choose not to, you know. So for example, like, like, let's say like, you know, Charles Murray proposes with with something like family-based, like looking at like genomic ancestry. Okay, so you know, we've been talking about like group, group differences in IQ, so like, just like brass tacks. Within a given family of like mixed race, there's going to be variation in the amount of ancestral contribution from different populations. That I think standard deviation's like about like be like 3%, you know. So you know, you'll have people that like vary substantially. You will have people, and also like the correlation between physical appearance and your genomic ancestry is actually like, it's not as strong as you would think.

I guess like 0.5, for example, for like skin color in African Americans and how much European ancestry they have, and that's because the number of loci is small. Anyway, the point is that you have like all of these variables going in different directions. You know, we could do like, you know, we, you know like for example, let's say that I'm a, let's say I'm Tom Steyer, like I'm a liberal billionaire.

I'm like, "I'm going to prove that systemic racism is causing the Black-white IQ difference. What I'm gonna do is I'm gonna fund genome sequencing of like millions of African Americans and have them IQ tested. I'm gonna show that there is no correlation between ancestry within sibling groups and IQ."

So, you know, if you guys have contact for Tom Steyer, you know, other liberal billionaire that wants to prove this hypothesis, we can do it now. You know? We can try to figure it out, but we choose not to.

Steve Hsu: On, on this issue, I'm a I'm not saying anything.

Razib Khan: Question? Yeah.

Audience 2: Just on that point, like the late political philosopher Robert Nozick attributed to the proliferation of the rabbis, like rabbis as a meritocratic selection for high IQ people and then encouraging them to have children and over generations. I

Razib Khan: I think Greg and, Greg and Henry said there's not enough rabbis, right? To drive that. That, that, that was their argument

Steve Hsu: so for the podcast listeners someone just stated the hypothesis that so, so what, is, is it selection for being a rabbi that it, or rabbis have more kids that yeah raises the overall intelligence? Cause the other one is that the, they would marry their most,

Razib Khan: Successful students, success merchant. Yeah. Yeah.

Steve Hsu: But I think there is obviously clearly a cultural veneration of intelligence in the Jewish community going back a long period of time, but that could have some kind of effect on the overall population becoming smarter.

I just wanna mention that you mentioned Bob Nozick, right? Yeah.

Razib Khan: Yeah, Robert Nozick.

Steve Hsu: So I was good friends with Bob Nozick when I was at Harvard, and, we actually discussed this issue. So he, he was definitely not afraid to discuss controversial, topics. Yeah.

Audience 2: Yeah. He was libertarian throughout any free thinkers and

Razib Khan: Activists. Libertarian. I think he changed his views by the end of his life is what I heard, but All right, are we, we winding down here?

Steve Hsu: Any last question?

Razib Khan: Who

Audience 11: was libertarian?

Razib Khan: Robert Nozick. Oh, Nozick. Anarchist Utopia. Yeah, yeah, yeah. Yeah.

Audience 11: Uh, in what way did he change by the end of his life?

Razib Khan: I he was, like, living in a rent-controlled apartment, and so someone, someone like Oh,
Audience 11: yeah, but he did take welfare, so

Razib Khan: No, but someone asked him, like, why, like like, "Why are you doing that?" He was like, "Well, I don't believe in those sorts of views anymore," or something.

Audience 11: Oh, interesting.

Razib Khan: Yeah. Huh. He might've been lying, I don't know. Whatever, you know. Okay, I think we're good.
Steve Hsu: Nozick, Nozick was very popular at Harvard Law School. Even though he was a philosopher, he would sometimes teach at Harvard Law School, or he knew all the Harvard Law School professors.

And so sometimes he would invite to the Society of Fellows dinners, where I saw him every Monday night, he would invite Alan Dershowitz. Oh. So I got to spend a lot of time with Alan Dershowitz, he of the, of Epstein fame, of course. Sure, sure, sure. ut at the time he was very well known because he defended Mike Tyson against a rape charge.

Do you guys know who Mike Tyson. You kids today Yeah. You don't even know who Mike Tyson is.

Razib Khan: He's

Steve Hsu: big,

Razib Khan: he's big again.

Steve Hsu: It's, it's ludicrous. It's ludicrous. So I used to talk to and I was the only guy, I was the only egghead in the Harvard Society of Fellows who followed boxing, and I was definitely the only Tyson fan in that group.

And so I used to I used to spend hours talking with Dershowitz about the case, so just a little anecdote for you. Yeah, question. What

Audience 4: are

Audience 2: your Epstein theories? Oh.

Steve Hsu: I have a whole episode on this on my podcast, so I but probably 'cause we're out of time, I don't want to go into the whole thing. Yeah, probably. Is

Audience 2: Epstein still alive?

Steve Hsu: I don't know, but kompromat Mossad. Word. Those are keywords Okay.

Razib Khan: All right, I think

Steve Hsu: we can All right. Well, thanks
​