This podcast is brought to you by the Albany public library main branch and the generosity of listeners like you. What is a podcast? God daddy these people talk as much as you do! Razib Khan’s Unsupervised Learning
You know that genetics plays a huge role in our health and more people are using genetic testing to determine risk for diseases like cancer for themselves and their kids than ever before. So I want to tell you about Orchid, it’s the only company that does whole genome testing for embryos testing before your child is born. If you're doing IVF, this is a clear choice now, because now you can reduce risk for 1000s of single gene disorders, including heritable forms of autism, pediatric cancers, and birth defects. Check them out at orchidhealth.com.
Hey everybody, this is Razib Khan with the Unsupervised Learning podcast, and today I am here with my friend, Steve Hsu, who is a professor of physics at Michigan State, and also a startup founder, entrepreneur, public intellectual, you've done many things. And you know, I’m going to say friend, sometimes you say friends, and kind of like a generic manner, where it's like, everyone's your friend. And, you know, I honestly, like I am the type of person to do that, like, I'm a friendly person. But you're just to be clear, Steve is a real friend, I've known him in the meatspace, as we would say in the 1990s. Since 2005, we've, you know, kind of worked on projects together here, they're talk to each other, we know each other, and we're aligned in a lot of ways. So this is going to be very collegial conversation. So what I know really well, but, you know, brilliant guy. And I think you guys will find the topics that we're going to cover interesting. We have a lot of history together. So some of this is going to be just reviewing where we were, where we're at where we're going. And so to start out with, you know, Steve, when I first met you, I met you in your office, like in real life, I met you in your office at the University of Oregon in 2005. You had these small children, who are not small children now, like, lots changed in the last 19 years. I didn't have any children. Now I have children, etc, etc. My hair is gray, you know, so, but we still have a passion for the same things. I think, well, not all the same things, but many of the same things that we had back then and so one of the things that we're passionate about, I think is, you know, human genetics, population genetics, behavioral genetics, these sorts of issues. And, you know, when we were first talking, genome wide association was like a kind of like a little bit of a twinkle in the eye, it wasn't really happening to the level that it would like five years later, by 2010 It was like a big deal. 2005, it was like Hot Map had just come out. They were still doing candidate gene associations and whatnot. And so the two of us like we were looking at the future, and we were hoping for great discoveries. And some some of them happened, and some of them didn't happen. Can you tell me what you are feeling didn't happen. I'm curious what you think about that, like, what did not happen?
Yeah. But I'm really psyched to discuss this topic. But I just want to say Razib, it has been a pleasure to know you now almost 20 years. It's hard to believe now. So you had, I think you were one of the founders of something called Gene Expression, which was a blog
It’s still around, the domains all around. Yeah.
And maybe, I think predated my having a blog. So you were an inspiration to me in that way. You're significantly younger than me. So you were a little more like internet - a little bit closer to being an internet native than me. Because I was already an adult, you know, when the internet took off. I just want to say like, you and some of the other guys, I don't know if it's public knowledge, but certain very prominent guy we know. Very big. Many prominent guys actually were participants in GNXP. And I don't know if we're allowed to say that stuff. But it's been a
Well, I mean, you didn't name names. So it's fine. You know,
A lot of prominent people that you would know now. Anyway, so you are an inspiration to me. That was one of the first like really active intellectual community, small, little communities where the discourse level was very high. That existed pretty much only on the internet. So that was kind of a new phenomenon for me. I mean, I have my own world of physicists and professors. But the internet world was a new thing for me. And then meeting you in person because you got your undergrad degree at University of Oregon. So you would pass through Eugene every now and then. Personally, I think you were one of those. One of the first people that I first encountered on the internet and got to know in person. There's a little cafe I think it's called Espresso Roma. I forgot the name
Yes. That is Expresso Roma. Yeah, it's still there.
It's on University Avenue, every anybody who's familiar with Eugene, Oregon will know it. It's a it's an institution at the University of Oregon campus. And you and I met there and had a, I think we talked for hours and the people around us were just shocked, like, Who are these nerds? Like, how can these guys talk for hours at this cafe? So anyway, I have very fond memories. It's been just great to be your friend all these years, Razib and I just wanted to say that because, of course, like, we have a lot of friends in the world. And you know, I'll often refer to a guest on my podcast as my friend, but very few of them have I known as long as I've known you. So I just want to get that out there
Yeah. Yeah, no, that's for real. Yeah.
So getting back to genomics, and what's happened since 2005. Just to remind the listeners, at that point in time, there were only a few datasets like 1000 genomes was considered a lot back then. And, you know, just basic questions like, Gee, I was told by my sociology professor, that I have as many genes in common with a Nigerian as I have with another Chinese guy. And people really believe that social scientists really believe that they, some of them still believe that actually. And when they had the first, you know, relatively small datasets with hundreds or you know, 1000, people, they could at least do some clustering, or distance, genetic distance comparisons between groups, and show very, very definitively that, oh, I have sort of 10 or 15% more genes in common with people that share my ancestry group than people that are of a distant ancestry. So, so lots of things got established right away that were, I think, kind of obvious if you have more of a mathematical or scientific mind. But that were very contentious. They were very contested, new bits of science that came out at the beginning of this period, when we just first got analysis of the first hundreds of genomes. I think now we're at a point where obviously, we routinely analyze sometimes a million genomes at a time. And we're able to build pretty complicated predictors of traits, complex traits, like how tall are you? What color your eyes? Do you have elevated breast cancer risk? Those are all things that are pretty standard now. They've only really become standard in the last five years. You were asking me about what I'm disappointed by I would say, I'm disappointed that we didn't make more progress on cognitive ability, which is a trait that I'm particularly interested in. People often attacked me for being interested in that trait. I'm not interested in that trait, because I'm especially focused on group differences. That's the reason that trait is particularly contentious, because obviously, we've long had these hypotheses that some groups are, quote, “smarter” than others. And obviously, a big part of colonialism, imperialism was based on those ideas or those assumptions. I would still be just as interested in cognitive ability if there was only one ancestor group. So if we were all if the whole world had been nuked, and only some Icelanders who were almost genetically identical, survived, and I was one of those Icelanders, I would still be interested in why Johnny, the Icelander kid who I'm tutoring math is not as smart as Jane, the Icelander kid that I'm tutoring in math, because those individual differences within an ancestral group or within even a single family are super fascinating. So in other words, what is the genetic code that builds your brain? And why do some brains perhaps function better than others? That, to me is the main disappointment that we haven't been allowed to make progress on that, and I say, haven't been allowed, because there's been lots of money to increase sample sizes for targeting medical conditions, lots of other complex traits, but the one that you just cannot get any funding for and in fact, you will be viciously attacked for expressing interest in is cognitive ability. So that that, to me is just a very irrational, crazy view, I understand where it comes from culturally, there's a cultural war over these, this question of group differences. And so I understand why it is the way it is. But I think a Martian, the proverbial Martian, or Mr. Spock, coming to earth would say, Why are these humans investing so much money in genotyping people? And they're not even interested in the main question, which is why their brains are so much better than all the other species on the planet?
Yeah, I was I was assuming that you were gonna say that and I want to get back to that I want to like, take issue or just like elucidate maybe one thing you said, because it's a common thing to say. The connection with these ideas of group differences, of superior/inferior, and colonialism. And people often think, Oh, well, you know, this, this is the cause or of colonialism, and what I would argue, is, you know, in some cases, that's obviously true, like, let's say, like the colonial enterprise of Africa by the late 19th century. But I think that these hypotheses, these, like ideologies related to these hypotheses emerged after the initial wave of colonialism, and just kind of this hierarchical structures that develop between different cultures of different nations. And so it's one of those things where the, the ideology was a kind of like post hoc justification, or, you know, for what was already happening. And then in the later stages, there were some colonial enterprises that occurred where, okay, this ideology was rolled out as justification. But really, that was like the later stage. And I think people need to understand that a lot of times, our ideologies and our views emerge organically from the situation on the ground. And then that's used to justify what the situation on the ground is. It's not what actually determines the situation on the ground, right? So for example, in the United - Yeah, yeah, go on.
I just want to say I agree with you Razib. I mean, I mean, people were enslaving each other and stealing their territory long before there are any theories about group differences. However, maybe as a post hoc justification, or something that came to mind, in the, among the early natural scientists upon encountering different groups of people who maybe were still using stone age technology, you know, it became ingrained in the many of the ideas justifying colonialism and things like that.
Let's talk about like, the cognitive genomic stuff. So just so the listeners know, you are not alone, in being very fascinated about this. I'm saying this because, you know, I can see, I'll tell you guys, I can see which links you're clicking. When I post like link roundups, you guys love that stuff. Okay. When I post a podcast was James Lee and Alex Young, you guys are really pulling that stuff down. Okay. So I know you're interested. I don't know, if you want to acknowledge in public you're interested, but you are interested. You care. So I'm sure that a lot of you are also disappointed. So right now, like where we were, like 15% of the variance, being explainable by genomic prediction. Right. Is that Is that Is that correct Steve?
Well, okay, I aim to be - I mean, since you have experts watching your show, and also there are, there are now professional computational genomicists, like professors who, you know, who weigh in very heavily on this stuff, we should be very careful. There's a phenotype called educational attainment. And because that's a convenience phenotype, because in many medical studies, they will just ask the patient or whoever they're getting the data from, oh, how many years of education have you had? So there's a lot of data millions of genomes with an attached education attainment number, it's not directly a cognitive score. It is correlated to cognitive score, but in a very complicated way. The best EA predictor probably does capture something like 15-16% of total variance. There's a whole bunch of debate as to the nature of where that correlation comes from. Is it direct causation in those alleles affecting the formation of someone's brain? Is it stuff that actually you got from your parents and caused your parents to be better caregivers? Or
Are you talking about indirect effects there?
Yeah. So it's very subtle, I just want to say that there's a lot to that EA phenotype, which I'm pretty sure it's not just it's not directly cognitive ability. And I'll tell you about a study that we did some years ago ready, we built a cognitive predictor using UK Biobank data, there's a fluid Int. field, in the UK Biobank dataset in which they administered I think it's 12 or 13 questions in just three minutes. It's a really terrible, shitty cognitive test. But it is a direct cognitive test is called Fluid Int., and it does correlate, you know, with your cognitive score higher fluid Int. scores associated with having a higher real IQ score, that there has been validation studies of the fluid Int. number, but it's a pretty noisy, cognitive score. We built a predictor using that. And I think at the time, there were maybe 150,000 of the UK Biobank people who had done the fluid in test. And we could build, we could predict fluid in better with our predictor, trained only on 150,000 people, then you could predict it with the EA score, which had been trained on you know, millions of people or something at that time, I think it was maybe a quarter million people. And when you then test those predictors, you validate them among siblings. So you restrict we have about 20,000 siblings that we can do these tests on. So you restricted the siblings for whom you have the scores and you say, Well, does the test still predict the difference in phenotype between two sibs as a function of difference in polygenic score is that prediction as good as if you just take two random people in the population and try the same calculation, it turns out EA, the EA predictor decreases in power tremendously. Whereas the fluid Int predictor decreases a little bit, but not nearly as much. And the way we interpreted that we didn't really talk about this much, because this is something that we did. And it was in a huge paper where we validated lots of things using the sibling method, mostly medical traits and height. Our interpretation of that is that, you know, that the deviation between like how much education, brother one and brother two gets, is determined by a lot of things. But there's a little bit in there of like whether brother one is smarter than brother two, or more conscientious, but a lot of times the parents are just pushing the kids to get a lot of education. And so there was a lot of loss in predictive power if you compared kids who had grown up in the same family environment, but that was less much less true for the fluid int. predictor. I say all that just because I want to point out, EA is not the same as cognitive ability or fluid Int
Yeah, that's really good point. I mean, it's, I'm not just saying it's really good point to say it's a good point. That's an excellent point, actually. And, you know, so what I'm getting across from this, like, I actually, not gonna lie, I have not looked, have there been twin studies on EA?
You mean in the sort of classical literature like the old? — I think so. I think there was some prior knowledge about the heritability of EA. Now again, the thing with EA, again, it depends a lot on are we talking about samples from a country where college is free? Are we talking about samples where you have to go into debt. And of course, then it makes a huge difference, right? Because the way that kids accumulate those extra years of education after high school is socially influenced. Though EA is just a very tricky trait. And my hats are off to people like James Lee and Alex who worked on this. I was kind of a doubter, I was like, always, like, I mean, we all go back a long time in this field, right? So I used to joke with James, and the other EA folks at SSGAC, and I'm like, this is such a crappy phenotype, what are you guys wasting your time on, let's just wait until we get more real cognitive scores. That was kind of my attitude. But they were right, that you could still get a lot from just using this convenience phenotype. But they have to do and if you look at the supplements, the appendices of their papers, they have to do tremendous analyses to normalize the EAA phenotype across countries across decades across socio economic status brackets, it's an insane amount of analysis that actually makes the, you know, makes the whole thing a little bit like rickety to me. Like, I just wouldn't put that much store in there thing. I would rather wait until we accumulate a million or more cognitive scores.
go through a few issues that you just kind of implicitly brought up and flesh it out for the listener. So you talked about cognitive scores, obviously, you know, we're talking IQ here. So IQs, not IQs. The general factor, of course, a bunch of, you know, the general intelligence factor, a lot of listeners will know what that is, that's kind of the closest thing that we have IQs kind of a colloquial term, really, we're talking about general intelligence factor, which is measuring the correlation between the all the tests due to a common independent factor, which is, you know, psychometric aptitude, IQ, like you know, colloquially, okay, so that exists, obviously, something like EA, is a phenotype, it's an outcome that goes through multiple sequence of, you know, variables, conditionals, et cetera, et cetera, right, there's a sequence of like, you're raised in this family, you're in this country, etc, and so forth, in terms of what the correlation between educational attainment, and this general intelligence factor is, this is the key of like, you know, just to be explicit, what we're alluding to here, there is a correlation of the correlation could be zero, or the correlation can be one, if it's one, they're just substitutable. And everything you say about EA is pretty much, you know, equivalent for the psychometric characteristic that we're really interested in. If it's correlation is zero, then like, okay, like, what, we're not getting anything out of it. Obviously, it's not zero. People that are more intelligent on the psychometric tests that have a higher score, let's just say that they have a higher score on psychometric tests are much more likely to get a PhD in physics. Okay. I think we know this intuitively. We know there's a correlation. The question is, what is the correlation? What you're talking about within family effects? That I think intuitively makes sense to a lot of people? I will tell you, I will tell you, just like listener. My family is one where there is no expectation that you graduate from college. There's an assumption. Okay. You know what I’m saying? An undergraduate degree is the baseline. Okay? And I'm not saying to brag or anything like that. This is just the way I was raised. It was never a discussion. Be because it was assumed that you'd go to university and get your university degree and definitely not as a studies major, okay? So there's certain cultural expectations within our family, no matter what the range is, like psychometric aptitude, which with its sibling cohorts, the standard deviation is what like 12, like, It's not trivial. There are families that have, you know, okay, like kind of smart kids and extremely smart kids, both of them are going to be pushed to go to university and get those degrees, get the four years of higher education. And the outcome phenotype technically, may be the same for two individuals that are actually very, very different in their cognitive abilities. And everybody within the family knows that. Everybody within the family knows that there's one individual, one sibling, everything is easy, effortless, or the other individual has to grind, grind, grind to get the same output, but they do it, they do it. And that is partly due to parental expectations, assumptions. Also, can your parents help you with university financially, all of these other variables that are going in. So that's muddying the correlation between the psychometric ability and the output, as you know, the outputs themselves? So the correlation, I think, what is it like, correlation between SAT and grades? GPA is was about .5, something like that, I'm assuming that it's actually declining because of grade inflation, but I don't know. But all of these, all of these things are measuring, reflecting something underlying, and you know, psychometric aptitude is not the only thing. You know, there's other personality traits. There's other things going on. And so like, you know, at the extreme tails, of say, you know, like, let's talk about like, I don't know, a theoretical physicist, you know, you even I know, people that want to get PhDs in physics at Princeton, these are the crème de la crème, you know, and I don't know, like, if there's like a special test that they could take that would separate them. But also, like, who succeeds and who fails, like, there's a lot of stochastic effects that are going on at that level, equally brilliant people, one person picks a project that pans out, one person picks a project that doesn't pan out, etc, etc. And so, you know, there was a lot of path dependency there. So I don't want to like overemphasize obviously, the cognitive ability, but it is, it is a big deal. In terms of what's going on in society. Let me just ask you, you're you are in the university system. They're getting rid of the GRE, they're getting rid of the SAT. I mean, like, are you? Do you sometimes feel like the crazy man in the land to the sane? Or the vice versa? You know, at this point?
Yes. I mean, probably, maybe we're going to talk about this. But so from 2012 to 2020, I was the vice president for research at a big 10 University. And that's a position with significant authority. I mean, I was responsible for about $700 million a year and research expenditures. And even more, perhaps, and usually, I was the highest administrator in the hierarchy, I reported directly to the president of the university, who read every faculty promotion file each year, that's a little bit unusual. Because the President who hired me wanted to enhance the academic reputation and capabilities of Michigan State, he made sure that I - that the all the deans knew that not only was I like the guy who was like, responsible for the research funding that went to those colleges, I was also the guy who was going to sit in a little sweaty room with them every year, reviewing the tenure and promotion files of the faculty. And you know, me, Razib I'm not a, you know, I'm a nice guy, but I'm not. I'm not gonna back down from anybody.
Show me the money. I mean, that literally money but you know,
Show me the Nature publications and the NIH grants, right. I was often in a conflict with a dean who wanted to promote some faculty member to tenure, who had a weak research record, weak overall record, but they were in a particular ancestor group or diversity category. And so I went through many years of that. And the trend really just surprised even me, that we got to this point, I think it was that summer of Floyd thing where people just went bonkers for diversity. And at that point, like, we just said, Oh, if every college has an internal office called the institutional research office that does all the statistical calculations, how are the students doing, how are we doing on research funding, all these numbers. The kind of things that you would report to US News and stuff like this. And if those guys come in and say, hey, you know what our, if we if we admit using SATs there’s no way we're going to hit this goal of having this number of kids from that particular ancestor group. And then the response for many years was there'd be a fight between the people who were for academic rigor, and meritocracy, and the people who are just for diversity, there'll be a fight, around 2020, that fight just became just became a knockout where the diversity of people just like, they could point to something and say, Hey, this, if I can argue this is bad for diversity, we're just gonna get rid of it. And we're only now just starting to recover a little bit. From that crazy time, this kind of Maoist time in 2020, where some schools now are starting to reinstate the SAT, many graduate departments have gotten rid of the GRE and the GRE subject exam. I hope that comes back. I don't quite see it happening yet. But yeah, we went a little, in my mind, we went a little crazy around 2020.
Yeah, and I want to I want to get back to that, you know, maybe at the tail end of our conversation after we talk more about science. Let me ask you, I know we share a common interest in quantitative genomics and behavior genetics and stuff like that. Honestly, I haven't like, probably since I checked the EA papers. But I haven't kept track of the field too much, partly because it seems like there's a little bit of stagnation. Actually, those of you who have not read it, I'm gonna give a call out to our mutual friend James Lee, who's been on the podcast several times. Steve knows him very well. So he had a City Journal article, eye on the news, ‘Don't even go there: The National Institutes of Health now blocks access to an important database if it thinks the scientist’s research may enter “forbidden” territory’ A lot of it has to do with disparate impact of like, basically, this is literally political correctness, filtering the scientific hypotheses of questions. I can tell you guys privately, and Steve knows this too, a lot of people are griping about this and angry but they don't know what to do. They don't want to target on their back. But yeah, I mean, this is, you know, censorship. They're censoring. That's what they're doing. And so I feel like there hasn't been that much advancement in the particular trait of interest here, compared to what I was hoping, expecting. And I've kind of like, turned away my focus from that. I'm just like, waiting for the big, the big, like, you know, step forward. What do you think about that?
Yeah, I agree with you. I mean, it's just fruitless to spend too much time on this particular phenotype. Everything else is advancing. So the amount of data available if, say, you're interested in diabetes, or heart disease. The amount of data that you can use to do studies of
Razib: or schizophrenia or autism
Absolutely, yes, or Alzheimer's, all those things are just growing at a steady pace. For cognitive ability, it's almost kind of not growing at all. This could change over time. I mean, I don't want to disclose too much, but people are working on this. But the timescale is quite long. And it's quite expensive, obviously, to because to get a good cognitive score for an individual requires, you know, potentially a lengthy testing procedure. So it's non trivial, but what you would like to see as a government that actually cares about, oh, neuroscience, gee, I guess neuroscience might be helped out if we knew the genetic basis for differences in individual ability, cognitive ability, right? Why don't we put, you know, half a million dollars behind, you know, the assembly of a million genomes with associated cognitive scores, like that would be a fantastic outcome. But there's no way to get that funded in today's environment.
I don't think it's going to happen the United States, if I had to bet, I think our political environment is too toxic. For that we are the world's Innovation Center, we are the world's Basic Science Center. So that's probably not going to happen. Certain parts of Europe are actually a little bit more chill about this. Like, you know, the Netherlands, anyone who follows the literature knows that they have like a big twin registry. And now they do their thing. Because of its independence, the Max Planck centers can sometimes do things in Germany, Germany as a whole. So for the listener out there, I don't actually, it's probably already been published by the time that this podcast will be live. I'm reading a piece of genetic history of Germany, Germans are not really interested in their own genetics. There's some history there. There's like a big gap at the heart of Europe when you're talking about like genetic analyses and population genetics, and I think we know why. But you know, the Max Planck centers are pretty independent, just like EMBL is pretty independent. And so there might be things going on there Sweden, as most of you well, not most of you, but a lot of you will know Sweden has IQ testing for conscription. I don't know if they still have conscription, but they had it for a long time. So and those people I think their names are recorded, like you could track them. We have the technology to do this if we wanted to. So do you choose to do it? That is the question and what Steve mistake and I kind of agree is like, you know, we just choose not to look right now, for various reasons. I want to say for the international listeners, because I know like sometimes you get like hyper American and Steve is saying ‘diversity’ and all these things. So he means a very specific thing. Diversity, Equity Inclusion is really focused on race and sexuality. It does not mean diversity of thought, does not be diversity of religion. Well, generally, it doesn't mean diversity of religion, but like, for example, the United States, like Muslims are de facto coded as brown even though they're all different races. And so I can imagine someone saying, like, Oh, she has a hijab. That's like diversity, because it's like racially inflected, right. So it's very, very particular culturally specific idea of what diversity is that he's alluding to, we all understand now. So for example, I'm just gonna make like a joke, but like, it's like, you could have a, you know, a department and if everyone's black, it's very diverse. Okay. We're not, we're not talking like Shannon's index of diversity here. You know, we're not talking about a technical definition. It's like this weird vibes thing, you know. And so I think it's a little infuriating. I know, it's a lot infuriating for people like Steve, me now on the outside. I just shake my head, you know, maybe raise my fist. You know, anyone who follows my Twitter feed knows that I have opinions on this sort of stuff. Right. But I wanted to be clear about that. Before we move on to like some non biological topics. Steve, I want to ask you, as you know, I've gotten really interested in ancient DNA over the last like decade or the last 15 years. So Antonio Regalado at MIT Tech Tech review a friend of yours. You know, I'm not going to stipulate how friendly but, you know, I told him in, I think, American Society of Human Genetics and 2013 or 2014, he's like, what's on the horizon? And I was like, keep an eye on this guy, David Reich, there's gonna be a lot of stuff that's happening and ancient DNA. And he's like, at the time, he was like, oh, okay, whatever. And he didn't follow up. And like, he says that that's one of his major regrets over the last decade, in terms of not following up, because obviously, he would have been ahead of the ahead of the curve there. And so he always asked me, like, you know, what is? What is happening in genetics and stuff like that. But anyway -
If you don't mind on that topic, I'll break some news for you. So thanks, partially, to you and other people, I was following the ancient DNA and Neanderthal DNA stories pretty closely, many years ago. And that partially led to the creation of the startup Othram. So Othram, as some listeners may know, solves crimes using forensic DNA. And I think the record is we've solved the case with only using about 15 cells equivalent of DNA, so found at a crime scene and then we use that to get a genome for that person, say, the killer, and then we're able to find through through a genealogical datasets, we're able to find a close relative of that killer, and then the police can catch the killer. And we think something like 1000 cases now have been solved by Othram. But one of the things that motivated me to do the back of the envelope calculations that told me many years ago, probably seven years ago that this was possible as a crime solving tool, was the fact that I just said to myself, wait a minute, if we can recover Neanderthal DNA that's 50,000 years old, why can't I recover a cold case? DNA from a cold case that’s 50 years old, or 100 years old and indeed cases that old have been solved by Othram. So actually, I have you and others who are interested to thank for some of that.
Yeah, so Mittleman, David Mittleman, CEO, you know, the prime mover there, I guess. He's been on this podcast three years ago. He's a friend of mine. We're still in touch. And yeah, a lot of our conversations relating to forensic genealogy. He was I was remember talking to it was after the Golden State killer thing happened. And, you know, I had a call with David. And I was like, you know, like, they could do a lot of this, right? Because, you know, if Pääbo can pull out, you know, the oldest now is, I think there's something older than 400,000. There's definitely the cave in Spain is 400,000. I think there's some older ancient DNA now, they're probably going to make it to about a million and that's probably with a few exceptional cases, that'll probably be the limit of DNA protein, they're gonna go even further. So you know, I just wanted to bring up the ancient DNA stuff. So because we're talking about stagnation in a field that we're interested in, and others like another field where it's like, literally, I don't think people could, we didn't have a human genome in in the year 2000. When we first had a discussion. There was a handful of human genomes out there. I think literally a handful. They were in an Excel sheet, and everyone knew their names. So now we're getting genomes, we have dozens of Neanderthal genomes, like Neanderthals, some of them hundreds of 1000s of years old. So the rate of technological change in some areas of bioscience right now is incredible. And, you know, we have a, it's just like, there's the synergy between different fields. So it's like computation along with automation. And then there's also some fortuitous, you know, chemistry. So for example, polymerase chain reaction, which I think a lot of you will know, you know, I mean, that wasn't really a thing before the mid 90s. So that's really amplified, literally our ability to get DNA and get at DNA in things like forensics and whatnot. I want to I want to pivot a little bit, I know that you have a new startup, but you're working in a new field. It's related to artificial intelligence, you know, you're a physicist. So some people are probably thinking like, Wait, this guy's a physicist, you're talking about, you're talking about like genetics, that you're talking about artificial intelligence. Let me tell you guys, physicists, they are the, they're the wizards of knowledge. They go where they choose to go, and no one will stop them. So I don't know if Steve, like got into economics too much. But that's a very common thing. They dabble in social science. Physicists are smart. They don't need you to know they're smart, because they know they are smart. I've never had a physicist correct me that they had a PhD ever. It's not needed. A credential is not needed. They open their mouths, and they know, you know, and so, you know, I want to talk about artificial intelligence a little bit. I haven't dug into details, like what your kind of, like 50,000 foot view of this field is, I've been talking to people recently about it on this podcast, like some listeners may be a little curious and confused, because of my background as a geneticist, but, you know, in technology, it's impossible to avoid the specter of artificial intelligence, of what it's going to do in terms of improving technology. Existential Risk, you know, all these other issues, the big picture issues that people are people are concerned about, how is it gonna affect medicine? You know, you know, how's it gonna affect, like, a lot of areas of operations, research and operations science, you know, in terms of like, you know, for example, like me, like you, we get a lot, you know, we both get a lot of email, you know, if we have AI assistance, there'll be like, you know, this, this is like, give us summaries, give us the highlights, and then we can actually, like, focus in and respond. So it can increase potentially increase productivity a lot, you know, but, yeah, so, I mean, what's your what's your general take?
So, I think when a science historian looks back on, say, the last decade, one of the biggest stories will be deep neural nets. And the breakthrough that happened with large language models just in the last few years. So you would, an audience would know that as GPT. That what what that is really accomplishes the following. So I take a very big neural net with, you know, perhaps 100 billion connections, and I train it to predict the next word, given the, you know, n preceding words. And, you know, I, they've effectively done this with to first approximation, all the text ever generated by humans, okay, a trillion tokens - a trillion words of text. And now, you might say, what's, what's the point of that Steve? What's the big deal about being able to predict the, you know, 50th word, given the 49th word in some chunk of text? If you think about what the model has to understand what quote “understanding” has to be encoded in those neural connections? For it to be able to perform that task well. What you conclude is that task forced the organization of the deep neural net, to encode, effectively a representation of all the concepts that humans have ever written about, and the relationships between those concepts. If you're a little bit mathematically minded, and you understand high dimensional vector spaces, I encourage you to go look into this because what has happened now is that they've built a roughly 10,000 dimensional vector space, which encodes all the Primitive Concepts that humans use and their interrelationships. And these transformer models are basically using that all the calculations are doing are within that embedding space. To the casual user, all what they realize is like I have some chatty bot now that understands what I say to it in multiple languages, and can respond appropriately. So it's A huge breakthrough. And if you know the history of AI, there was this question of like, how is the AI ever going to know that when I say, How much is that doggie in the window, the only way the model can properly understand what that means if they know that there are things called pet stores, and that oftentimes there's a little pet for sale in the window, and how much refers to the cost. And the reason the dog is in the window is not because it's someone's home, but because they're at a mall, and it's a pet store, right? So there's all this like implicit knowledge and understanding and relationships between concepts that most AI people until just recently despaired of ever, quote “teaching” their AI, all this stuff. And in just one jump by basically brute force training on huge datasets and using the right clever architecture for the deep neural net, they basically surmounted that problem. So that now these GPT models have an internal representation of basically all the concepts that humans use. So that's the way I describe the breakthrough. Now, the problem with this breakthrough, and the reason we started our company, superfocus.ai is that this technique, you will have noticed, it changes the neural connections in such a way that the model does well, if it properly guesses the nth word, or n plus one word. And when you say it did it well, it's comparing against some training samples where those n words appear. So it's really being trained to generate plausible text, but not necessarily factually correct text. So if you ask the model, is there a United Airlines flight that lands in Paris tonight, around 6pm? It has seen many references in literature to that kind of thing. And it'll say - it could easily just say, yes, there is. There first class seats available, you know, or whatever, it'll say something which is plausible. It's similar to the text it's seen. But in fact, there may not be United Airlines Flight landing at 6pm in Paris. So that's called the hallucination or confabulation problem that models have. And it's really a fundamental consequence of the way that they train these models. Okay. So, to solve the hallucination problem, our startup built something which conceptually we described as an attached memory. So when I talked to you, Razib, you're using the language and reasoning center in your brain, which is very similar to the what a GPT is, what a transformer model is, but you're consulting your memory. So if I say, hey, Razib, remember when we first met, the part of your brain that interprets that sentence is different from the part of the brain where it's pulling out Like, oh, yeah, yeah. Espresso Roma, Steve talked for a long time, I couldn't believe how much espresso that guy can drink, right? That's being pulled from somewhere else, and shown to the language center or reasoning center of your brain, and then you formulate your response. So we've built a software architecture that allows our customer, which could be, for example, a consumer electronics company, they can take the training manuals that they use to train their call center people, that are those people, you know, troubleshoot your Smart TV for you, they tell you where to send it if it's broken, they do all these things, right? The manuals that they use, we can install now, as the memory that the LLM accesses, and the LLM will then faithfully answer any question that the human asks, using only the knowledge that's in the attached memory and not the training samples that it saw relevant to TVs and stuff like this. So instead of telling you that there's some drop down menu that you need to access and click a button for, which doesn't actually exist, that's a common problem for LLM is like, you know, being asked these kinds of troubleshooting questions, it will only respond based on what is in the attached memory, the ground truth that we and our customer have installed. So you were asking about jobs and how this is going to impact a human society and whether you're going to have a personal assistant that sends email for you, that's an AI. All those things are possible once the hallucination problem has been solved, and it is now reliably solved. So for example, this consumer electronics example that I gave you, the technology actually is deployed, and it is superhuman, so it does a better job troubleshooting problems with the smart TVs than the modal human call center worker. So it’s superhuman, and it's actually rapidly in the next couple of years going to replace a lot of human jobs. My co founder and I, Tushar Sheth, have been tag teaming flying back and forth to the Philippines and Singapore. Because a lot of this call center work is done by people in the Philippines who speak English well, but the wages are quite low. And over the last 20 years, a lot of work has been outsourced to the Philippines to the point where a 8% of the Philippine GDP $40 billion a year is exactly the kind of call center work that I'm describing. But that can now largely be done by the kinds of AIs that we build. The kind of AI that we build could actually read through your email messages and formulate responses sticking very closely to what is in the message that you received. And then some base information that you want it to rely on in answering those, like about your podcast, or the state of your startup or whatever, you can tell it like, hey, only respond using these 100 pages of information that I gave you don't deviate from that, but respond to whatever the person asked for, we can easily build AIs like that now. So I think the impact on human jobs is going to be significant over time. And I think it's something that people really actually don't understand yet. People, even economists and such, have not really figured out what's going on how this is all gonna play out.
Yeah, there's so many ways I can go with this. You know, and, you know, as we're recording stuff is coming out with the the one minute videos produced by open AI that are very realistic, except for like a certain weird few things. I do have to say, as you were talking, you were using n words, maybe you should like change that phrasing a little bit. I know what you're trying to say. Yeah. Yes. So. But in any case, I think, a lot of stuff you say is fascinating. It's clearly going to transform the way we work in a lot of ways. I don't know like how accurately GDP is going to be captured. You know, I've seen arguments I don't know what you what your what you said that like that what you believe, but that the internet has transformed our lives in ways that have had economic effects but are not captured by GDP. I think that's probably true. I think AI, open AI, chat GPT and stuff, that's also going to be similar. I was just curious what Gemini - I have Gemini Pro, and ChatGPT thinks of you. Gemini didn't say much. It's pretty generic. And there's also a cardiologist named Steve Hsu. But I'm gonna read what open AI says and I'm gonna tell you what, I'm going to ask you what you think about it. It’s not that long. Steve Hsu is a physicist and entrepreneur known for his work in theoretical physics, genomics, and higher education administration. He has a diverse set of interests and achievements, including contributions to the field of quantum information theory, and the foundations of quantum mechanics. Hsu has also been involved the application of advanced statistical and computational techniques to understand the genetic basis of human traits, including intelligence. He co founded a genetic testing company which focused on deep genetic information for various applications, including personalized medicine, and understanding human evolution diversity. Additionally, Hsu has had academic appointments, including serving as professor of theoretical physics and at times, taking on administrative roles within higher education institutions. Hsu was also known for his involvement in discussions and debates surrounding genetics, intelligence, and their ethical implications. He maintains a blog and podcast where he discusses these topics among others related Science, Technology and Society. It's worth noting that Hsus’ work particularly in areas of genetics and intelligence has been subject to public debate and controversy. Critics have raised ethical concerns about the implications of genetic research on intelligence and the potential misuse of such information. So that was that was pretty good. There's some weird issues in there, but
I think that was really good. I don't have any complaints about that. It's much better than like The Guardian.
Wait, what did the Guardian say?
Oh well, so as your listeners may know, our company Genomic Prediction, which does embryo genetic screening, The Guardian has, I think, written several articles attacking us over that kind of thing, as you know, it's it's eugenics, it's evil, etc, etc. So, also, they don't like this is in the weeds, but the Guardian really doesn't like me, because I'm friends with Dominic Cummings.
Oh, yeah. You're a Dominic Cummings stan your public about that. So I mean, it's weird, just so the people out there know, once you have hit pieces written on you, it kind of exposes, all of a sudden, you're focused on what the structural parameters of journalism are, or what the motivations are. So you find that a lot of it is like personal beefs. Like, why are people that are just like you not attacked? Well, the personal beefs are not aligned. You know, it's so it's not like, totally random, like, there are reasons behind it. You know, I still have the recording, I should post it. It's legal. With the guy that wrote the hit piece on me on Undark, like, there was like a moment in our multiple hour conversation, where he almost apologized for what he was about to do. Because Deborah Blum the editor clearly had told him what he was going to write in terms of like, you know, it's not going to be like, you know, so he didn't use remedy quotes from me because like, he didn't get too many bad quotes from me, you know, so I never cooperate with journalists, by the way. If there's any journalists listening, unless I know you really well, I never cooperate, I will never give you a quote, I will ignore you, I will block you. And then also, if I do trust you, and you do turn on me, I will destroy you. Just just to be clear, I will come after you personally, on social media and with every person that I know. So come at me, I will come back to you. I've been canceled so many times, it doesn't matter. You can't kill the dead. Sorry, that was kind of a dark segue. But like, I'm serious about this.
Im glad you have that perspective. I think if more people had that perspective, I think a lot of these journalists are just irresponsible. I think - Well, this is a long digression. But I think that the standards of quality and truthfulness in journalism obviously are pathetic these days. So yeah, I agree with you.
Well, so in terms of AI, we're obviously in a hype cycle right now. You're, you know, you're next year, you're doing a startup, you're being an entrepreneur, and actually just random, it's a small world. You employ a guy who is like, I've got to know recently, a very young man in Austin. And then, you know, he was like, Oh, my boss is coming into town. I'm like, Oh, who's that? I'm like, Oh, I've known him. Anyway, everyone out there, there are 1000 Interesting people in the world. You know, that's it. That's what I've learned in my, you know, into my early middle age. That's what I'm going to say.
It is kind of a small world of, you know, if you put various cuts like this person is, you know, pretty intelligent, this person is willing to do original things, even if they're against the grain. This person has the determination, and willpower to carry through a lengthy project that might last years, you're already down to a fairly small number of people. And when I meet a young person that I think could be in that category, I'm very interested to get to know them and understand what the world looks like to them. So it's just a treat. I'm in the bay area right now. And we just had a meet up dinner. Actually, the guy that you mentioned, I believe, has since moved from Austin to Hayes Valley, which is sometimes called Bayes Valley is all the AI guys are here. That's where I'm staying right now. And he was at dinner last night, we had this whole meet up of people last night. So yeah, it's a, it's a great time to be alive. Like, just to come back to the AI thing. There's this hype cycle where like, when ChatGPT launched, people were blown away. And then you immediately had people talking about Doomer stuff like singularity and terminators killing us. But if you listen carefully to what enterprise CEOs said, CEOs of big companies that might use this AI, just listen to their quarterly call all throughout 2023. They were always saying we can't deploy these things at enterprise level, because they hallucinate, we can't take the risk that it gives wrong information to one of our customers or books of flight that doesn’t exist, you know, all these things are the problem. So in the background, in of this hype bubble, you have companies like in the trenches, grinding, trying to fix all these problems. And so my own plug for our startup is that we actually have fixed these problems. We have AIs deployed now that do not hallucinate and are superhuman in their capability. So you're just going to see more and more of that. You're going to start to see less and less of statements by enterprise CEOs, big company CEOs about hallucination, because their people are gradually going to propagate up to them. Yeah, it's solved. We have a, you know, we have an AI that basically handles like the equivalent of the FedEx communication about packages and stuff like that. We've actually built one that's interested in testing right now. So you're going to see these AIs deployed at scale, and you're going to you're going to actually, they're going to become an everyday part of your life, the way the Internet has become, you know, very big part of you.
Yeah, I will say, you know, it's, for example, I would love I think a lot of people would love an AI that would, you know, do Excel work for them, right? The issue is like, you are not going I mean, especially like all of these, like, you know, McKinsey Excel, you know, jocks like they would never trust an AI to do the Excel work because like a hallucination could be a massive, massive financial hit, they're gonna get, you know, they're obviously not going to have a career after that, whatever. Okay, so I get that. But I was also like talking to - So I’m gonna have him on the podcast at some point, my friend, Nick Gray, he does these, like, you know, he's an excellent founder. And he does these little cocktail parties where people will be new people and everything like that. And I think like when it comes to something like party planning, hallucination is not actually necessarily a bad thing. Because what it does introduces some wildcards and otherwise everything is over optimized when it comes to - So my theory of like, I'll put my theory out there because as you know, Steve, I am a very social person. And, you know, my theory is like, basically, like, I have a core group of friends. And the periodically like, I'll sample out, you know, now that's an opportunity cost. That's time. I, well, I'm not gonna get into that, but until recently, I didn't have very much time. So it was a very big, it was very big opportunity costs, but I would still do it because one out of 10 times I meet someone really cool, and then they're going to be one of my new friends. Otherwise, you know, friends move away, et cetera, et cetera. You're just like pool kind of, like, overtime declines. You know, that's just how it is. And, you know, a lot of people - I don't talk about my personal life too much. But there's a lot of people that say things online, like, men don't meet new friends after college and all that stuff. You know, me, Steve, that's not a problem for me. You know? Yeah, it's just like, we just like, we just put ourselves out there. And we're like, Hey, what's up? Like, are you interested? Are you not interested? If you're not you move on, you know, maybe you're interested someone else. This isn't rocket science. It's not physics. So I think I think like a lot of these interpersonal things, AI could help. I'm not introverted. But I think introverted people would be able to actually benefit from some of these nudging technologies. I don't know how they could get rid of their social anxiety, like maybe they need some sorts of pharmaceuticals. I don't know. I don't want everybody in the world not to be introverted, though. That's another issue when it you know, just kind of looms over like, you know, embryo screening and other issues. Other questions? You know, I call it the Korea problem. In Korea, South Korea, everyone wants their kid to be a doctor or lawyer. But most Koreans are stupid, because most humans are stupid, you know. So that's not gonna happen. I'm being very like blunt, like an East Asian mom right now. But you know, most Koreans are wasting their time trying to be a doctor and lawyer. So you have a nation where people have like all of this, like investment in education, but their outputs are not commensurate, you know, in terms of how much effort they put in. And I've talked to many Korean people of like Midwid level, who are clearly extremely resentful about all the effort that they put into education, because they're managing a hotel now. They have a middle class life, they're happy, they're fine. But like, they have like a really weird, freaked out attitude towards learning. i It's hard to describe because I didn't go through that. And they still revere learning in a way, like they know, it's super important. But they're obviously going through PTSD. That's what's happening, when I talk to them, when I talk to them about their kid going to college or something like that. They're having PTSD. I only bring that up. Because you know, if everyone embryos screens for like, extremely tall, good looking, intelligent, whatever, however you define it, if it's like total homogeneity, you know, I mean, United States like and we talked about the Steve - The United States is an extremely wealthy, productive country, but like, on the metrics, like, you know, IQs, not as high as Taiwan, or South Korea, like what's going on, like, we have, like this unique mix of, we talked about diversity earlier, right? Like, we do have diversity, it's not optimal, in some ways, okay. Like the crime, other things, everyone knows that. But like, look at the innovation centers that we have, you know, we have San Francisco, that's all. That's all that’s needed.
Yeah, the US is a special case, because first of all, US absorbs a lot of the top brain power from the rest of the world. Up until recently, we had, by far the best educational institutions in the world. And so people would come here, and then the top people would stay here. So we're in a unique situation. Also, you know, as you know, for in the post war era, we were the only undestroyed industrial economy in the world. So the game was here. The infrastructure was built here, the great institutions, the great laboratories, particle accelerators, all that stuff is here. Supercomputers. But less and less so over time. So if you actually ask, like, with four times the population, and slightly higher, at least by test scores, cognitive ability in China, if you put some cut, like someone who's able to, you know, do an advanced degree in engineering, or physics or something, how many people did they have in the generation that's just entering the workforce. It's almost an order of magnitude higher than the United States. So and those people are no longer required to come to the US, they can work at the cyclotron in Beijing, and work at the battery research company in Shenzhen. So everything is shifting. So this this idea that the US can have, like, a lot of people that can barely read and write, but still be the richest country in the world that will be challenged over time. I think that model is going to be challenged and but looking backwards, yes, it's worked great for us so far.
So, you know, Steve, I want to talk about higher education. And, you know, go into, like, what happened to you a little bit, at least in terms of what you're comfortable with. But I want to talk about like China also really quickly. I am like, much more bearish on China than I was three years ago, partly just because like bizarre governance and whatnot, obviously, like the human capital stuff that you said is totally true, you know, but it seems like it flourishes a lot more when it's in the United States, in terms of like what could come out of it. Like in China, you could create a unicorn, and then the party decides not a good sector done. Okay. So I think this sort of like instability is a major problem. And the second issue I have to say is like, we've been looking for innovation centers outside of the few that exist in the United States for a long time. Don't look to Europe, everyone knows this. I have a lot of European entrepreneur friends, there's a reason they're fleeing. Okay, Europe is where civilization is going to retire. As far as Asia, you know, there always has been an issue in a lot of East Asia societies in particular, because when we say Asia, that's what we're really talking about. I don't want to talk about the permit Raj, okay, just I don't wanna talk about that, okay. But like in Japan, the hierarchical system and science, some of that also is true, and like other parts of East Asia, and that's obviously dampening innovation. I still think China's gonna be great, obviously, when it comes to engineering, and gaining efficiencies on a lot of the processes. But in terms of like basic breakthroughs, I'm still now more bullish on the United States, partly just because we have a crazy chaotic, not over engineered social systems. So that's what I'm gonna put out there.
Yeah, you may be right, it may be the case that the East Asians are a little bit better at optimizing stuff than, you know, zero to one and completely novel new ideas. I think it remains to be seen, because that civilization, you know, despite any headwinds coming from Xi, over a longer timescale, they're getting richer, better educated, and in the sense that I think they will eventually become more free. So I think it just remains to be seen, what will happen. I will say, I often say this, when I'm interviewed on podcast, if if you if you really want to know what's happening in the world, you should just write down on a sheet of paper 10 key areas, industries or technologies that you think are important. And then spend 20 minutes on each of them see what is actually going on in China. Oh, is solar energy important. Hmmm, what are they doing in China? Well I guess they dominate the industry. Okay. Energy storage, electric vehicles, batteries? That could be important in the future. How's China doing on that? Oh, I guess they're dominating that industry. You know, AI? That seems pretty important. How's China doing? Oh, actually, they have a number of models that are comparable, maybe a little bit behind GPT4. But you just go down the list? Oh, space exploration? Oh, I guess they have their own space station. And they've landed, you know, rovers on Mars and on the dark side of the far side of the moon. I guess I didn't know that it wasn't really reported much in the US media, you know, if you actually just go down the list, oh, jet engines? Oh, should they have good? Oh, I see now that they no longer require Russian or American jet engines. Their J 20. Stealth fighter has a domestically produced jet engine, you know, you just go down the list and you realize, Wait, something is happening here that you could lose track of, if you're not paying attention.
That point is taken. And also like, you know, I'll just like explicitly, you know, people like, you know, there's some skepticism of Chinese Science sometimes, for various reasons, some of them are not unfounded. But I do know for a fact that so in an area that a friend of mine works in, and I'm just gonna say what it is, because I'm not outing his name. There's many people works in this area, agricultural genetics, applied agricultural, molecular genetics, and the guy knows the area. And he's like, yeah, they're catching up and surpassing, it's hard for American institutions now to get postdocs, Chinese postdocs, it's hard to get them. And if they get their PhDs, they leave, they don't stay in the United States. This is very different from 10 years ago. So that's just that's just fundamental reality. So we got to keep that in mind.
A fast economic growth rate in 10 years, your country could easily become twice as wealthy as it was. So the rules of thumb that applied 10 15, 20 years ago for US China relationship or, or how attractive it is, for a highly trained scientist to be there instead of here, all of that can change, you know, over a period of time, where oh, there, they became 4x richer than, than they were before. So everything is going to be a little bit different. You have to actually watch the frontier in China, because moving so fast.
Well, I do have to say, just the last word before, because I know you gotta go. And I want to talk about the Michigan State issue. But I will say Not everyone is as fecund in East Asia, as you, Steve. So that's another major headwind, that high IQ, but they got to disappear, bro.
That's it, that's probably their biggest challenge. They have to solve that. But they have about 20 years to solve it. So the people that are going to enter the labor market in China over the next 20-25 years, they're born already. We know the numbers. And they will continue to have a deepening of their human capital pool, not in raw numbers. But in terms of the level of number of trained engineers, number of trained scientists, number of MBAs that will continue to grow actually over the next 20 years. So people who do demographic forecasting of China, half are usually like innumerate people. If you do it carefully, you actually see very different things, at least for the next 20 years than what most people are predicting.
Okay, so I want to end on - We've been having a good collegial conversation, obviously, we're friends. And it's not like pleasant topic, but I want to touch on what happened to you, in 2020 . You know, I was emailing you, then a lot of people were chatting about it, like, you got, famous. Some of the people that went after you were people I've been friendly with, like people I knew, personally, like people who our kids have socialized together, you know, I'm not as close to a lot of those - Mostly, they're not as close to me, let's be candid. A lot of the people that could do that I used to be close to that I used to came up with because, you know, they decided in 2020, that they were going to take a stand for anti racism by canceling people like us, because, you know, we're white supremacist. I'm just saying like, it's just like, the whole thing, I don’t know. Like, the whole thing doesn't make sense outside of an American context. People are really confused. I did have some non American,
You’re the brownest white supremacist, I have ever seen.
There's a reason that the KKK they have those hoods, okay. Just sayin’ But if you're like anyone, I mean, I had a friend who supported Elizabeth Warren, and he was being attacked as right wing in his lab. Because like, he wasn't a Bernie bro at the time. So I mean, this is what was going on in academia. You know, some of the people that were observing what happened to you were not surprised. And you, you were kind of surprised, right? But actually, like, let's just give me like, give a capsule summary. Give us a capsule summary. And then tell us your perspective of what happened.
Let me give you a summary. Because most people are not aware, even if they even if they followed superficially, what was on the internet or in the media about what happened to me, they don't know the whole story. So I'll just go into it a little bit. There are two forces that were trying to get rid of me. One was a bunch of radical leftist grad students from the Union, the GEU, Graduate Employees Union, and actually led by a guy, I won't say his name, because I'm much more gentlemanly than him. But you know, him, he works actually in genomics, and got his PhD at MSU. And he was basically the ringleader of this whole thing. And basically trying to create a kind of Twitter mob to go after me. The second thing, which is invisible to people who are outside the academy, is that if you work eight years as the hard ass guy who is trying to impose rigorous standards, and grow the university research budget, you make a lot of enemies. Anybody who is the chair of a department or dean of a college, that isn't getting the money from me, because I'm putting it into, you know, STEM stuff, is going to be my enemy. Anybody that have clashed with because they wanted to promote a professor who had never gotten a research grant, and not published very many papers, simply because that person was quote “diverse” We would clash over that, because frankly, that is illegal. A lot of illegal things are done in the name of diversity, okay, just to be totally honest. So, because this is not affirmative action, admitting a freshman to your school, this is a legal employment matter where you are perhaps denying tenure to an Asian research professor, assistant professor, and you're giving it to an assistant professor from some other ancestry group. And the disparity between their track records as professors is enormous. This would happen routinely. It is happening routinely at every university right now. And I was fighting against that. So I had a lot of enemies. Okay, so that's the backstory. The grad students rather comically link to a bunch of blog posts I had made over the years and claimed that the showed I was a racist. But if you read those blog posts, most of those blog posts are me discussing a published scientific paper, not by me, but maybe by David Reich. Okay, so that these comical guys who couldn't - I think their reading comprehension isn't very good. They would often characterize, or maybe they did it deliberately, maybe that's more charitable either they're liars or they're dumb. They would miss characterize some blog posts that I made as a racist thing. But of course, people never go and bother to read the actual blog post, they would just assert something. Right. So anyway, that was that was a campaign that was going on, and they started a petition to get me fired from my position as the Vice President for Research. And meanwhile, The people that were my enemies within the MSU system, obviously, were very happy to sign on to these petitions. Then there was a counter petition, the counter petition, if you look at - if you're a credentialist, and you say, well, what's the level of prestige of the pro Steve petition? Wow, Steve Pinker sent a letter to the President University supporting me, Sam Altman signed my petition, the former dean of Harvard Medical School signed my petition, All kinds of so that the level the, you know, huge numbers of STEM faculty at Michigan State signed my petition. So anyway, that's where it was left. Now, the guy who was my boss, who was not the president who hired me, since she had since been replaced, he was not nearly as committed to the things that she hired me to do. He was scared. He was, I mean, literally scared. I mean, I can say this, because I had many conversations with him about what was happening in the wake of George Floyd in that summer of 2020. He was scared. All corporate leaders, all higher ed leaders were scared that they could easily be overturned by this wave of craziness that was sweeping the country. And so actually, he said, Well, I'm going to ask you to resign. And, you know, the way that position works, Vice President for Research, I report to the President and I do - I'm basically the instrument of the President, to do the things on campus related to research and academic integrity, stuff like this. And if he doesn't want me working for him, I have better things to do in life. So I just said, if you want me to resign, I'm going to resign, because I'm not going to work for you, if you're swayed by what these idiots are saying. So that's how it transpired. And actually, I'm quite happy. Like, when I look back, I can't believe I actually did that job. Like, why was I working so hard to save these, you know, corrupt, horrible institutions that our universities have become? Why should I be doing that? Let other people do that. I've got better things to do. Right. So that's the capsule story of it.
Well, you know, I know Steve, you have often said that. And other people have said this, you know, I have friends. Also, I want to say this because I think people don't understand it. Don't understand who you are sometimes or who you were, where you come from. I remember talking to you, I think at Berkeley, we had coffee 2008. And I mean, everyone out there just so you guys know, I've never been a lib. I've never been liberal. I've never been on that side. Steve, you were like, you're in exultation about Obama. And I was like, okay, you know,
This is the funny thing. Like, because of the weird woke thing that happened. Many university professors like myself who voted for Bill Clinton, Barack Obama, John Kerry, Al Gore, you know, suddenly we’re right wingers? This is how insane it is, if the average American could understand how far left our Professoriate our grad students have become, they would just be shocked. It's totally at variance to the standard distribution of political views in this country. Because to normalize the country as a whole. I'm a liberal. I'm left of center. It's just insane. But but on campus, I'm a right wing nut, compared to the
Razib: Benito Mussolini.
Yeah, I'm Mussolini compared to the woke sentiment on campus. But how can we allow our institutions of higher education to be that far at variance to the beliefs of the people in the country, there is something wrong here. And what has happened is that the university shifted so far to the left, that now it becomes an uncomfortable environment for the engineering professor who's a Republican. In fact, they even feel justified in not offering the job to the young applicant, as assistant professor, if that person is Republican, it’s revealed he's Republican or even worse, like he thought about voting for Trump or something. The whole thing is just in a runaway nonlinear feedback loop where it's just getting worse and worse and worse. And smart, talented young people can figure this out and say, like, Hey, man, after I get out of grad school, I'm going right to Silicon Valley and start a company I'm not going to be within this crazy circus that the university has become. So that's unfortunately the situation we're in but even like Republican state legislators and congressmen don't really understand this DeSantis understands this a little bit, actually. But and he's done a lot of things in Florida to try to correct the situation. But most even Republicans, I think Republicans are a sorry lot, actually, honestly, they just don't seem to be very effective in the world, but but they don't really understand, like, how far things have changed, and that they need to reach in and correct these things before it's too late.
Yeah, I 100 percent agree. I've talked to people, they're clueless, they think - I've talked to people where an aide. I don't know Gonna say who. He was an aide to a prominent Republican senator. And I was just trying to like kind of get across, like, shits going down, you know? And he said, Well, you know, a lot of Republican senators don't want people thinking that they're Neanderthals and I'm just like, doesn't matter they're gonna put them in cages. It doesn't matter what they think so, um, you know, you said, Steve to people before that academia, you know, you wish there was a place intellectually vibrant, you know, academia, people are just really interested in these deep questions. And there's no other place like that. Would you still say that 2023 2024?
Well, unfortunately, it's still largely true because professors are paid to think and explored issues in great depth over many years. And there aren't that many other positions in society, which allow you to do that, you know, there a the National Labs. There are some new alternative universities like the University of Austin, which I'm involved with a little bit, which are trying to create more based institutions of deep thought in higher education, but it's a heavy lift, it's going to be a difficult road for those new institutions.
Yeah, I guess if the issue is like what is the net Good? Because, there were always communist in academia, though, Americans were not communists, who understood that that was the cost for this, you know, vibrant institution that produced all these ideas and spin.
If you watch Oppenheimer or whatever, during the Cold War, there were some communists at the universities. But now, a lot of those beliefs which were only held by those communists in those earlier times have become the standard woke beliefs on campus. That's the shift that really shocked me that, that people can openly say, like, dude, if there are too many South Asians and Chinese guys in our department, we're gonna discriminate in hiring against the next candidates that come through that are from that ancestry group, like, where did that come from? That is? I call that racism. But this is this is actually practice, no matter what these guys will tell you, if they're on the record, that they don't do it, they do it, and they think it's okay. And they're basically judging people by groups. So it's okay to discriminate against the Chinese or Indian researcher who's trying to get tenure in favor of a guy who's black or Hispanic. They think it's okay. They think that is justice. That is the woke mentality. But I call that racism. I actually can't I cannot believe that people accept this, but it goes on every day in our universe.
Oh, yeah. Quite literally. People tell me what happens in searches and the conversations that happen, and there is a university that you are very familiar with, that you may be affiliated with, in a department that you're familiar with, that may have hired someone based on their race and gender. To speculate
Oh, I’m sure they have, I'm sure they have the part that's really verboten to say is that this has been happening for a longer period of time in certain hard sciences, in favor of women. So it's also true that, again, a very noble goal. We'd like to have more representation of women in physics and math, engineering, let's hire more of them. But if you actually look at it from a legal perspective, it's actually illegal what they're doing. They're discriminating on the basis of sex.
I'll give a concrete example, for the listeners out there because this was told to me by department chair in mathematics at a research one university, not a top 10 In math, but not like bottom tier. So let's say like mid level, and this guy, he's a man, conventional liberal, not quite woke, but not anti woke definitely. And he said, basically, on the faculty searches for math at his university, and I wanted to stipulate the prestige level, just so it's like, it's not Harvard, but it's not, you know, I don't know, like bumfuck state. Okay. This, you know, in any case, this individual said, usually the top female applicant, and they have to hire a certain number of women, the top female applicant is ranked 50th Out of the total pool. Okay, that's, that's pretty sad. And this is not again, this is a pretty liberal guy, actually. That's what he said.
So there you go. Yeah. So this kind of thing is going on. The rest of society is not aware of it. These institutions are still in part funded by our tax dollars. And they play an outsized role in shaping the worldview of students who go there to vote, get educated, edumbacated. And so it is important that when the variance becomes so large, between the belief system that our institutions of higher ed adopt and the general population Something has to be done. It is not right for, you know, what would be considered a centrist political position to be considered grounds for disqualification in hiring for a faculty position at a major university, but that is roughly the situation now.
We've been talking for a while I know you gotta go, I want to close out this discussion, particularly, you know, I hear a lot of things privately, from people just because I'm a big mouth. And I will tell people what I really think. So they know that they can be safe talking to me, that I won't denounce them. Anyway, what have you heard that surprised you or shocked, you, or has taken you back, or has been sobering since what happened to you? Because you became, I mean, you were always you will already out there. But like, you know, for like a hot minute, there. You were the main attraction to the academic circus. So what did you hear, like, just give us obviously, like, no names needed, but I just want to get a general sense of what you heard, like, What was you were exposed to after that?
Oh. So I'm not sure exactly what you're looking for in this question. But I'm sure there are people that still hold a grudge against me, and that they would try to block me if I applied for a certain grant or did something. So there's probably a lot of that that like, now that they know who I am, they're probably going to go after me. On the other hand, you know, there are plenty of people that support my position. There are actually many people who emailed me and said, I support you, but I'm not going to sign your petition, because I'm afraid. Is that the kind of institution that you want your kids to be educated in? So, I'm not sure exactly where you're going with the question.
That's what I was wondering.
Yeah. Just to show you. So, in Michigan, like in California, we have a state law that was passed by referendum against affirmative action. But I know for a fact Michigan State and Michigan, we're practicing affirmative action against in violation of state law for many years during the time that I was actually in office. Do the Republicans in the state of Michigan, though even the ones who got that law passed in the first place care or know about this now? Basically, not. So the whole the whole system is pretty malfunctioning right now.
Yeah. Okay. One thing that I'm gonna say here, I was talking to a friend of mine, you know, you're talking about like people who are scared. Academics, you know, are generally cowards. You guys could disagree, but I know you're a pussy you won't say to my face.
I totally agree with you Razib. Most people in academia are careerists. And they just want to get ahead, the best thing that I could say for the majority of them that lack courage, the best thing I could say for them, as they would say the following ‘Oh, I support you, Steve. But you know, my main goal is to push forward our understanding of dark energy in the universe. And I can't jeopardize my big NSF grant by supporting you publicly, but I do support you’ And okay, that that person at least has an articulated justification for the cowardly behavior. Fine, I respect that I maybe would not have been a hero during the Holocaust. I maybe not would not have, you know, risked my own life and my family to rescue people persecuted at time. But most of the people in academia aren't even thinking like that. They're just like, go with the flow conformists who they don't have an original thought, you know, in their heads, which is kind of surprising, because they're supposed to be the leaders, intellectual leaders and researchers in this country. But that's, that's after a very long career in academia. I would tell you, that's an accurate characterization.
Yeah, I mean, a lot of people, you know, not just in academia tell me like, you know, they're based on all this stuff. I'm like, okay, like, uh, you know, just for the listeners out there “based” is kind of like the opposite of woke, like you say, what you think you don't give any fucks? You know, and, but they're like, Well, you know, I have this startup, , I have this that, you know, I'm just like, you know, when the time will come. And people have been saying this to me for 20 years, because I've been a loudmouth for 20 years. And I'm gonna tell you, most of you are pussies, you're never going to actually, you're never going to actually like reach the stage of comfort and security where you're gonna stick your neck out. That's not how you're built. Most people are conformist, they are cowards. The goal has to be to push that small margin and just increase it a little bit. And, you know, it's not always easy. You know, Steve and I put ourselves out there, people come after us. You know, like, I get like social media mobs, probably about every six to eight months. All its kind of calmed down recently. I don't know, I think people are just getting probably because academic Twitter has kind of died. You know.
Here's, here's just a brief comment. Because I know we're running out of time. Academics do value argument, and intellect, and they claim that they decide their beliefs rationally. But I've never ever had a Wokster academic come at me on the facts. It's never happened. It's never happened. So they say like, Well, I didn't like what you wrote in this blog post. I'm like, let's talk about it. You think this paper published by these researchers at Harvard, you think the paper is wrong? Tell me why you think it's wrong? Or did I misrepresent what I said about this paper in my blog post? Tell me tell me where. Which sentence do you disagree with? I have never, ever had on a campus where I was the top research person on this campus of over 2000 professors. I knew people in the neuroscience department, the psychology department, the engineering department, you know, everything, I knew all of them. Not a single - This is for your listeners. This is about the academic integrity of professors and scientists in this country. I have never had anyone, even from the sociology department come to me and say, Hey, Steve, let's talk about what you're saying. Because I think these are racist statements. You know, when you say that, Oh, IQ measured age 13 is relatively predictive of like earnings at age 50. Right? Oh, I criticize. I think that's a racist statement. Let's talk about the data, why you think that's true, or whose study you're quoting, when you when you reference that result, I've never had that happen. It's just absurd. Like, within physics, if I if I make some tiny little statement about some theorem about the mass of a spin one half quantum field or whatever- Guy will come into my office, and we'll argue about it for three hours until one of us is convinced that the other guy's right, that's what really should happen in academia. On this particular thing, I've never, ever had a wokester or come to me and have a serious scientific discussion over a point of disagreement. It just shows how little depth there is in that other position. And you know, your friend, Paige Harden, whoever it is, I'll debate them right now, you know, on a video, and we can talk about the science and we'll talk about where do I agree with you? Where do I disagree with you? It will turn out that we will probably agree on almost everything and where we disagree, they may look like a fool. Okay, so that's, that's what I have to say to these people.
Yeah. You know, we ended with harsh words to the world out there, but sometimes it's necessary. This is why this is why Steve - the knives came out for him. But I mean, this is reality. And that's what we're trying to understand. That's what we're trying to get at whatever domain it might be. You know, and, you know, I hope I feel like Steve and I can, we can at least hold our heads high and be like, you know, we didn't misrepresent who we were to the world for it to be safe. And I hope that you guys appreciated this conversation. I know that I got a little bit harsher than I usually do. But I feel strongly about this. I feel strongly about what happened to Steve, I feel strongly about what's happening to academia. And I feel strongly about the truth. And if you go against the truth, you go against me, and I'm gonna come at you, you know, and so I just let myself go a little I apologize for some of you. I'm usually a little bit more congenial and I avoid the harsh words, but sometimes the harsh words are the truth. And so it's why it came out.
Hey, Razib, if you give me two minutes, I want to tell you a story. Never told before just to illustrate this.
Okay yeah. This is for your ears only listener.
So I’m new in the job. It's one of the first years that I'm running this promotion and tenure process at Michigan State. About 150 professors go through this every year. And one of the cases was a promotion to tenure case, there was there's a thick file, all their research publications, student evaluations and letters of evaluations from outside people in their field, okay, and then letters from their dean and chair, all that it's a very thick file that I have to read through. I'm in this room. I'm looking through the file. I've already read the file and I'm meeting with the dean and the associate Dean's and all the people from that college. And you know me, I give zero fucks, right? So I'm reading it and I said, this guy's record is weak, he doesn't have any mains. He's not the main PI of any big grant. His publications are pretty weak. I don't understand why you guys are saying this is a strong case. Okay, I'm just totally flat out. This is how it works. Almost nobody in the world knows how this works. If you're not poor, even the professor's don't know how this works. Because you're only if you're a senior administrator, would you ever be in a meeting like this? Okay. So I'm going through the file, I'm saying this to the dean. I'm saying like, this is really not a strong case. I don't understand why you guys are writing about it as if it's a strong case. So just asking for clarity, right? And what am I missing? Did I miss some major award he got or some big breakthrough that he made? Did I miss read the file? There's extreme discomfort among the people from that college. Okay, the dean is uncomfortable. The Associate Dean is uncomfortable. Everybody's like, kind of like, why is Steve pressing this matter? Doesn't he understand? So I'm like, totally confused, and the guy had what I thought was either a German or Jewish surname Okay, I'm not. I'm totally colorblind, ethnicity blind, whatever. I was just judging the file the way I judge it but the guy's name was a kind of German slash perhaps Jewish last name. And they were just going around and around just telling me why he's a nice guy. He's really important to the department to the college and like, what the hell you what the fuck are you guys talking about? Finally, I opened my laptop, I avoid looking at the people. I want to judge by the file. Okay, I opened it up. Turns out that guy is black. Okay, but he had a name that, probably he was adopted, which didn't signal that at all. And I didn't look at the fucking color of his skin. I looked at his file. So then I realized what these guys were saying to me. Like, they were like, surely you're joking, Mr. Feynman, you're not going to deny tenure to a black person. Are you? Like, sure? I'm like, What the fuck is wrong with this system? Like was I supposed to know he was black, it doesn't say he's black anywhere in the file was I supposed to like do a skin color test of all the people whose files I'm reading, that's the system I landed in. That's the system that runs your higher education in your country. I just want everybody to know that.
Well on that note, which is a real note. And I have heard not in those details as an administrator, but that's happened so much. Also, I'm just before we close out, I'm gonna say like the east Asian people? Immigrants? Let's just say immigrants from other countries, they sometimes do not understand the difference between text and subtext. And so they will think that minority means minority, but it means a very specific type of minority, you know? So they get really confused, and they have to be pulled aside and in whispered conversations, and they're like, okay -
There are many hundreds of say, Chinese faculty at Michigan State University, and they among themselves, they may be telling some newcomer, hey, by the way, you're going to be judged much more harshly than some other people in your cohort. Because if we have to keep some Chinese guy in our department who speaks broken English, that guy better be a fucking superstar in his research. That’s not really the mentality that people have in this country? So it is what it is.
Well with that, I will link to your blog and everything like that. Steve it was great talking to you. Great catching up. I'm sure we'll see each other IRL soon. I'll let you go to your next meeting. And I got another meeting as well. We're busy guys. And like if you guys got problems with us, come at us and we'll talk. Later bro,
Whole genome sequencing is used for adults and children every day to assess risk of thousands of diseases. Orchid, a genetics company led by scientists from Stanford is able to do this for IVF embryos. Now instead of waiting for a diagnosis, parents can assess if their embryos have genetic variants known to cause severe conditions before their child is even born. Noother tests can detect these issues so thoroughly or so early. So check them out at orchidhealth.com
Is this podcast for kids? This is my favorite podcast.