AI, Job Loss, Fertility Crisis and the Death of Careerism | ft. Tyler Cowen | PGX Ideas #4

Tyler, thank you so much for your time and your hospitality.

Happy to be here.

I want to start you off, and I mean, I've particularly enjoyed your style of podcasting. I think in general, the way you speak and ask questions is very declarative. It's not explorative. You'll ask a question that has an answer. And generally, that's something I cannot do when I do podcasts because people will have guesses. And I mean, maybe I'm wrong in my analysis, but I'll ask you, I'll start you with a question. What do you think is the world getting wrong about AI?

People either overrate or underrate AI. Very few people get it right. So most people you might call ordinary people underrate AI. They've used ChatGPT, they enjoy it, they think it's useful, wonderful, but maybe it hallucinated for them a year or two ago, or they're getting only the free version. They're not sufficiently impressed. Then there's the people who work on AI. Who think it will put everyone out of a job or bring utopia or maybe kill us all. And those projections, I think, are also wrong. The old saying, as usual, the truth is somewhere in between. That would be my stance.

What does the in-between look like?

AI will be smarter than humans at most intellectual tasks within a year, or in most cases it is already, but it is not well connected to systems, to institutions, to research labs, to universities. For that to happen will take a long time. It will be a slow, somewhat painstaking process. The rate of economic growth will be higher, but not unrecognizably so. And the smarter the AI gets, the more the remaining human imperfections, stickinesses, laws, rigidities— those will be significant bottlenecks. And we'll have somewhat more rapid economic growth, but it will be pretty manageable.

Is it, um, is an analogy for AI as a technological advancement kind of like the fire, the wheel, or the internet and piston engines?

They're all a bit different. Maybe I think of electricity. So the printing press superficially looks similar, but it took the printing press centuries to matter because paper was so expensive. It's not until paper is much cheaper that the printing press brings a true social revolution. Fire, we just don't know how that went. It might have been a very slow process. As technological revolutions go, though I'm stressing its slowness relative to those who promote it, I do think AI will be the quickest technological revolution we've ever had. You can have it in your pocket. You can buy it at an affordable price. It, you could say, already is everywhere. And it will be very disorienting, simply the fact that almost everyone's job will have to change. It might be better. It might relieve you of routine. But people don't like change. And I think the better AI does, actually, the more it will be disliked.

You think this might be the first time that the rate of change of technology might acquire this kind of speed or this kind of self-improvement cycle that's so much faster that the roots of the world cannot catch up?

There's already recursive self-improvement, say, in the anthropic AI models, Claude. So it does the programming, or most of the programming, for improving the next model release. So that makes the model releases much quicker. That's an acceleration. But at the same time, as the model releases accelerate, the human bottlenecks become all the more important. So the total rate of progress only goes up a little. And that's why I think this intermediate view— over 20, 30 years, it will matter a great deal, but each year it won't seem that radical. I think that's the correct expectation.

The reason I'm asking is because— so if electricity is a good metaphor in that when electricity came about and the rate of transformation, the sort of progress of economies changed, that's what it's similar to. But a contrary view is that this is technology like we've never created before. There is a sense of novelty associated with this technology where like we might have built the technologies of all technology. Is that a rational understanding of it, or is it, is it just hype?

It's true for every technology. It was true for electricity. Try living in a village without some and you see how important it is. All the different things we do, all the jobs we have, rely on electricity in what is mostly a beneficial way. But of course there are high costs from electricity. It makes wars more destructive, would be one obvious problem. But none of us would want to give up electricity or or wish it had not been invented. That's why I think it's a pretty good parallel to AI, because life before and after electricity, once it is put everywhere, is truly different in a way that, say, the 16th century was not that different from the 13th century. You could say electricity plus fossil fuels connected to powerful machines, if you want to be a little more exact. But that has been the most important revolution of the past The AI revolution might not be as important as that one. It might be. We'll see. Hmm.

I was just asking you right before we started rolling about whether you see the current moment for a lot of fresh graduates, a lot of young people, you know, being just bad timing, whether it's just that these 5 years— sorry, boss, it's going to be disorienting. Things are going to look like they don't fit and you might not get a job. And sorry if your career is always a little strange because of that. But in 5 or 10 years it'll be all right, or do you think there's a permanent irreversible kind of shift that might occur and we might have lesser jobs in general and that kind of stuff?

I think they have good timing. I have bad timing. So I think in an AI-enriched world, people will basically die of old age and not of cancer or the other maladies that kill them now. So those young people could expect to live to something like 98, provided they're not in plane crashes or whatever. I'm age 64 right now. I may not see those benefits. So whose timing is bad? Now, it is true, the first X number of years of their career, they'll have much higher uncertainty, but they're young enough that they can adapt and they see it's coming. They know it's coming. They'll be adjusting. So I think the people who won't get the longevity benefits are maybe the big losers. And people who are, say, age 40, it's harder for them to adjust. Their minds or skills are less flexible. They might be bigger losers, but they're not old enough that they could retire. Like, I could retire if I had to. I'm not going to. I don't want to. But if I had to, I could. That's a level of protection. The people who are 40, most of them cannot retire. So I worry about them more than the 20-year-olds. And for the 3-year-olds, it's probably pretty awesome.

Yeah, I mean, they'll grow up just living that technology as if it's always existed.

But how they will think about their own intelligence is an important and interesting question. We do not want humans to be demoralized. Always imagine having a sibling who's smarter than you about everything. You know, there are niche effects in families, and the other sibling then takes on a different role, finds a new niche. So if all very young humans are growing up with the sense of themselves as not very smart. I do think there's a quite valid concern in that. I don't think we understand the psychology of how that would develop. But if you're asking me what are my concerns, that's one of them.

So you might think that we might have a discouraged generation or generations because there is just no scope to be intelligent enough to stand out.

Or maybe they don't even know enough to be discouraged. Like, I simply assume I can't run as fast as a cheetah. Or a lion. I'm not really discouraged by that. Uh, there's things I do more or less because of what I perceive my abilities to be. And maybe all these young people will be more physical in a very real, rewarding way, but they'll somehow be less in touch with the intellectual potential of mankind. They might even be happier, I'm not sure, but there's something in me that doesn't like the risk of that prospect. Because I grew up you know, saturated in Northeast New York City, you know, American intellectual culture of the 1970s and '80s. To me, that's a kind of peak of what my country has achieved, something very valuable. And if the next generation doesn't have that, is it just that I'm an old stick in the mud, or am I correct to mourn the loss? I'm not sure.

And in a world if there is no intellectual possibility, if there's no purpose for me wandering intellectually, struggling in the intellectual domain, that might change the shape of how I approach concepts like truth and meaning as well, won't it?

It's very hard to predict what will happen, but again, there's plenty of room for concern. But I would just say the status quo when it comes to meaning and truth, it's not that great. There's somewhat of an epidemic of mental illness in young Americans today. Our politics has an above-average degree of misinformation. Over the last, I don't know, 10 years or so. So AIs, as we experience them now, they're quite objective. They're better truth tellers than the media. They'll give you straight answers. They're super helpful. So there's a lot of upside and how the net effects all play out. I wouldn't say I'm pessimistic. I'm just saying any big change, you should be concerned. And if you don't see potential downsides, you're just not thinking about it hard enough.

Right. How do you think information and sort of that information dissemination, which is journalism and sort of media, how does that change? Because on one side, yes, there are— there is an ability for me to get more tempered updates on what is happening in the world or information in general. It's not, it's not as biased or laden with bias, hopefully. But then on the other side, it's easier for me to make a video of you saying something you're not, and it's just unrecognizable from a real video. How do you see the information landscape change as a function of artificial intelligence?

I don't think people are that fooled by fake photos, fake videos. People were very worried about this 10 years ago, but we've had fake photos for a long time. Certainly plenty of people are fooled on Twitter. Does it matter that much in the real world? I don't think so. I worry more about a kind of free rider problem that People now use, say, Claude or ChatGPT instead of Google. So you're not referred as often to media pages. So what mass media, even niche media, they're probably less profitable than before AI. So maybe those dwindle, but the AI is to some extent parasitic on those. So how does the AI learn new things if there's not an internet to read to teach it? I think we'll work all that out. But again, if you're looking for concerns, it's very much worth thinking about. We'll probably end up in a world not so different from the current world, where there are major media properties owned by very wealthy people who may not even make money on them, but they want to own them. And for them, it's a worthwhile way of spending their wealth. That's more or less today. Uh, maybe it's not thrillingly perfect, but it's not the main problem today. It's okay. Like the New York Times, Wall Street Journal, they're very good newspapers.

As an analogy, I have an instinct that we might be in what I think is the DOS era with AI, where it's a text-based, it's a dark screen, I put in some text, it responds with some, um, and I don't think we've hit the Windows 98 or the Windows 2000 and then sort of that curve that we saw in operating systems go up and the UX of computers. Um, how do you rate that instinct? Um, am I underrating or overrating Am I right? Am I wrong here or what do you think?

I think a lot of people will simply talk to their AIs. They might get the answers back in video form. I think that's fine. These are people who already watch YouTube, which can be either terrible or highly educational. I'm not sure how big a change it is, but one thing that has shocked me throughout the internet era is how much people love typing letters into tiny boxes. It's proven to be one of the most robust things— texting. Whatever comes along, people still love texting. You know, in India it might be WhatsApp. Actually, with my friends it's WhatsApp or iMessenger, whatever. I just think it's not going away. It's almost biologically rooted in us to communicate in that manner. So some version of, you know, writing and print culture, if only through texting, will persist.

From hieroglyphs to emoticons.

Yes, but there'll be a lot more video, and that's maybe a few years away, but it's not distant. It costs a lot, the token costs, but we'll get that down.

One of the other concerns that I got acquainted with recently with an AI researcher was how a lot of the costs for generating AI queries and responses to that are subsidized by venture capital right now.

Right.

And so a lot of it that seems free and easy to do and all the prompts that you can put in to generate bullshit images that you may or may not need later are actually funded by large funds sitting out in Silicon Valley or sovereign funds of different countries. And once the money withdraws, we might see a rise in cost. And at that point, there might be affordability issues with artificial intelligence, stuff that we are not thinking or factoring in, in our first order analysis of artificial intelligence as a landscape. What do you think of that?

Well, it's going to be priced. The best models today already are priced. So GPT Pro, which I find to be the best, that's $2,400 a year. It's not cheap. Many people can afford it. Many people can't. So imagine buying intelligence in a metered fashion, similar to today how you buy electricity. Electricity bills are pretty high. They've just gone up a lot because of, you know, wars. The world will work around that, but it's not going to feel free and it's not going to be free. But that said, by some estimates, token costs over the last, what, 3, 4 years, they've dropped at least 100x and they'll continue to drop. So It won't be crazy expensive. Some uses might be.

So if, I mean, if I'm treating intelligence like electricity where I just get a metered bill for acquiring a certain quantity, whatever the quantity unit might be for intelligence, what is an alpha that remains? What's the, what's the beyond? If it's, uh, if intelligence is equalized across different places, what separates me from you?

Well, you and I might be equalized, but I don't think it will in general be equalized. There'll be super fancy models that might cost, I don't know, $20,000 a month, and top companies will buy them. Militaries will buy them, intelligence agencies and so on. And they'll be much better than what you and I have. It might be accessible to us, but we would have to pay. So it will evolve into something less egalitarian from that point of view.. But keep in mind, the very best models of the future, they'll be so good, like, I can't even tell the difference. So we're already at GPT-5.4, you know, Claude, Opus 4.6. For answering economics questions, the next generation of models, it can't be that much better because the current generation is so good. So for doing more complex tasks where it, say, has to work on something for 4 days in a row, In those areas that are not very legible to us, they'll get much better. But on legibility, we're already saturated in reasoning. That's kind of stunning, but it's mostly true.

What do you think will be the bottlenecks to the AI trajectory as it's predicted by at least AI companies?

Institutions don't know how to adopt AI.

So a coordination issue, implementation issue?

But it's also just knowledge. How do you inject AI into your institution? Not just your personal use as a freelancer. Oh, Claude wrote this memo for me. I saved 45 minutes. That's already happening at a very large scale. But that the whole institution is somehow connected to AI as an information and decision-making apparatus. There's a coordination issue. There's a liability issue. There's a privacy issue. There's a do you want to send all your queries to the company issue, especially for law firms and governments. But just the who has really done this, you see it happening in startups, which are from the beginning built around AI injected everywhere. But I don't see mature institutions making progress at a very rapid pace. You think startups are.

So that's an implementation kind of how do you fit it into existing architecture?

Right.

And does it even fit? Are some architectures just resistant to something like this and that kind of questions?

Try fitting it into modern universities. Good luck. And they are very resistant to change. They're nonprofits. They're often controlled by faculties or nonprofit boards or administrators who don't know much about AI, who fear AI, regard it as competition. That's going to be tough.

Do you think on a macro level there's going to be regulatory pushback?

Depends on the country, but absolutely. The European Union already has done it. We'll see exactly how they enforce what they've written down. But I think most countries will.

I do wonder, because in countries like India that are that labor heavy, you know, we have so many— one of the largest exports of India is great labor, right? I— given how sensitive that is as a political point, I don't see how regulation does not become a talking point if there's sufficient job loss.

It will be a talking point, but keep in mind it will be limited what any particular government can do. Because if, say, AI can run a call center, Hyderabad might be very upset, the national government of India might be very upset. But if someone can run the AI out of Estonia, India can ban whatever it wants. Somebody in Estonia, and the jobs still end up lost. So there's competition across jurisdictions, and you can put AIs plenty of different places. So we'll see. But don't overestimate how much governments will be able to do.

So you think government regulation might be overrated?

Overrated in its long-run impact. It will slow many things down, create uncertainty, slow down how institutions integrate AI. But at the end of the day, you can call up AI, so to speak, from anywhere that has access to cloud computing, and that will be the dominant effect.

Do you think the inference that even though people are losing their jobs, productivity is going up is accurate? Because it sure seems to be reflected in stock prices.

Productivity is going up in the United States. It varies by country. European nations are much slower to adopt AI than even America. I think so far the effects are small. They'll be larger. I once guesstimated it will be worth half a percentage point a year in terms of productivity, which compounded is quite significant. But in any single year, you barely notice. So I think right now we're at the stage of barely noticing. Hmm.

You think the idea that big companies might be able to function without the same number of employees, is that an accurate possibility?

The New York Times just wrote up a company. It's run by two brothers. And they do some kind of healthcare services, and they have more than $1 billion of sales a year, and there's 2 employees. Yes, there are outside contractors, but that will become more and more common. Companies with fewer than 10 employees that are quite productive based on a mix of AIs and contracting, and they'll be super nimble and flexible, and they will change the world. They're the startups that are built around AI rather than trying to, you know, fit the round thing into the square hole, or maybe the square piece of wood into the round hole.

We're talking about how large companies can run, um, with AI and people actually might lose their jobs. And why I'm interested in that question is because very often— and I don't know if it's a sufficiently complex view of the future— but very often one of the ideas that's painted is that we might have no jobs and you might just be writing poetry and music and gardening and carpenting in the future, and that we should begin to make sense of life in a non-careerist way. Do you think that's overrated or underrated?

There will be jobs for the foreseeable future. So just the energy sector will need to grow tremendously to accommodate all of the compute we need for AI. Healthcare, biomedical trials, taking care of old people as we age. Entertainment. People still want creativity from other humans, not just from the AIs. They might want both. But they'll still want there to be, you know, a Taylor Swift, a Zakir Hussain, a Magnus Carlsen. So there's been great AI in chess for a while now, and we still follow the human chess players, Gukesh, and we don't follow the AIs. So there'll be plenty of jobs, they'll just be very different.

And so this is a temporary displacement, migration from one end of the spectrum to another, is that like from coding to healthcare or from visual design to energy? Is that the kind of displacement I'm imagining in the future?

Yes, and I do think some of the old jobs will disappear before the new ones form. So say to create a new job in the energy, energy sector, it takes some while, due in part to regulation, but also due to fixed costs. You want to build a new energy plant of some kind, say nuclear power, it's slow, right? It can take 8 to 10 years. Whereas the people in the service sector, consultants or some people who do legal research, they might fail to get new jobs rather quickly. So I think there'll be a transitional period that will be not entirely pleasant, uh, but in the medium term, we will stay quite close to full employment.

And so you, you do think in the temporary case there might be medium-term displacement for some people trying to get jobs, even though they will get absorbed somewhere else?

That's right.

And do you think in that meanwhile there might be need for welfare policies to sort of make sure people are afloat? You think that might occur?

It will depend on the country. Some countries already have too much welfare, but some countries have not enough welfare. The countries that have not enough welfare, that will become a more serious problem that they'll need to address more quickly, and they will have resources to do so.

One of my ideas, one of my concerns, or my observations, is that The biological need for status and conflict is priced into capitalism as a model. We find meaning out of our jobs and careers, and we compete and coordinate over our work and our jobs in a certain way. If there is going to be change in that, if effectively intelligence is going to get commoditized to that extent, how do people compete, coordinate, find status, and, and find those psychological, biological, psychobiological needs?

In an AI world, the real intelligence will be taking AIs and integrating it with what humans can do in the physical world. So say you're an economist, as I am, to do things like gather data in the field, persuade a policymaker, raise money for your projects. The AIs won't be doing that. Those will be your new tasks. You will achieve status by doing well at them. It's not that different than the current day. It's a different set of tasks.. But simply, oh, I'm running the regressions, I do it better than someone else at another school. I think that will disappear. The AIs will run the regressions. Maybe people at Harvard, MIT will have access to better AI, but even there, the, the advantage might be asymptotic. And it will be these other skills more directly rooted in the actual physical world, synthetic skills where you have to combine your brain with doing things that only a human can do. They'll be much more important. Charisma will matter more. Maybe your looks will matter more. How you communicate face-to-face will matter more. Podcasting will be a great background to have.

Very interesting. So you're saying largely skills that involve a face and a value structure.

That's right.

Right.

And a willingness to accept responsibility and liability, ability to inspire others, to lead, to know what the AIs are good for and deploy them accordingly. So a lot of new skills will become much more important, and they will be the markers of status.

Of the skills that you listed, I could put a big circle around most of them and call them leadership skills. Some of them, like for instance, where do you implement AI, might be secondary. But how many leaders can 7 billion people have? Like, are we going back to small tribes and cults, uh, in that sense? Or— because what are the rest of them going to do?

Noah Smith had the hypothesis that many workers will become more like the Japanese. So large numbers of Japanese run their own quite small businesses. They could be small retail stores somewhere in Tokyo that don't have much scale, but they do fine. It's a much wealthier world. You have better, cheaper healthcare, uh, better services of all kinds, again, also cheaper. And you'll be a boss of some kind. So are there people who are just so allergic to the idea of running things they won't do well? I think that's the case. I do think there will be other things they can do, like you can still be a gardener.

What do you think will— how do you think will artificial intelligence impact global population sizes?

There's a fertility crisis now. Even India is at about replacement rates. It will be a long time before that matters and say the population in India truly starts to shrink in a measurable, noticeable way. I don't understand why fertility has fallen so low or what it might take for it to come back. But with or without AI, we need for it to come back at some point, or we all just dwindle to near zero.

What do you think about the hypothesis that it has to do with modern living, plastics, and this kind of stuff? Do you think that's a—

Highly unlikely. It's happening in too many countries, including countries that are not just full of modern living.

What is your best guess? Why might that be happening? Because it's a strange phenomenon. It's an existential phenomena. It's not some sidey— like, it's not, you know, it's not like, uh, people have skin warts. This is the end of human civilization we might be talking about. And if it's happening across the board, there's very few countries above replacement rate at this point, right?

Children are not as fun as we might like to think, is my best hypothesis. Or at least they're not perceived as as much fun. And furthermore, people marry much later. They're fussier, or they wait. Women have more education, getting better jobs. They want real careers. That's all understandable. But if you're marrying at age 30 instead of 22, you're going to end up with fewer kids.

Do you think that's a— do you think careers are branded better than kids in the modern world? Is that the problem? It's a branding issue?

Yes, but it's not just branding. It does correspond to some real things. If you have a great career and you have one kid at age 31, your life's pretty great. It's not just you were tricked by some advertisement. One kid can fill so much of your attention. Do you need three? Well, some families will prefer three, but the one will take up all your time if you let he or she. So for a lot of families to have one seems to be a norm. Then countries especially like South Korea, where the men and women have such different viewpoints that they just don't marry to begin with— less the case in India, but India may get there— that's another problem. And no marriage is going to translate into fewer children. I don't know how to fix that. But since marriage is good for many people and men and women coming together is good for many people, I do think it's fixable. I just don't feel I'm the person who knows what to do about it.

Um, there is a large concern, and I think I've only gotten aware of it over the last 2 or 3 years, um, what could be termed as generational inflation, is that my dad, who's roughly your age, could have made far less money, could have been far less successful, and could have owned a bigger piece of land than I can with the kind of money and the resources I have, because land is fixed and the prices have skyrocketed. And in some way, the compounding effects of capitalism has made newer generations more insecure financially. And part of the reason why they cannot get married is because they never feel stable. It's a concern I can echo from some of my peers, right? Very often, extremely successful people, and they be like, I don't know if I'm ready to have a kid. I don't know if I have the kind of money. Um, do you think it's a downstream consequence, an invisible disadvantage to just competitive pricing that has led to people not having kids? Is that— how do you factor that?

It's a problem, but there are countries with cheap real estate where fertility rates still are falling. One nice thing about AI, it will ease this problem. So you'll have the equivalent of quote unquote a labor force without having to live in a major city. So you can live in cheaper places. We'll probably be able to terraform much of the Earth as we want to. I don't know how much of that will be profitable, but we'll do some of it. Places that don't have enough water or they're too hot or too something or other, AI will help us manage those problems. So there'll be a lot more high-quality, cheaper real estate available. And I hope that also boosts fertility.

So, and apart from these traits at the edge, like charisma or being able to inspire or lead people and machines into action, having, let's say, a vision or knowing how to implement these tools. What else will be rare in the age where intelligence is commoditized?

Repeat the last question. I missed one word in it. Sorry.

Uh, my question is, what is going to be truly rare 10 years from now if intelligence is commoditized? And I heard you that, you know, having the ability to lead people, inspire people, apply tools in specific directions, having a vision— these things might— or having courage, being able to take on responsibility and deal with its consequences. Those things might be rare as traits. What else might be rare? Where else might there be an alpha economically, financially, or psychologically in the age that intelligence is commoditized?

New ideas. So Beethoven is rare in every era. There's one era that had one Beethoven, right? Or the Beatles. So no matter how good AI gets, if a human being can become something like a Beethoven in whichever area, The returns to that will be incredible. It will be more valuable and there'll be more competition to do that. Obviously, most people will not succeed at being the next Beethoven in whichever area, but we will have the wealth to support very large active tournaments of people trying to do many things.

Do you, do you think there is a case in which new ideas is also a compute problem and not a human judgment problem?

Compute will aid us in generating new ideas, but I don't think compute alone generates new ideas.

What's the missing link?

We don't understand new ideas very well because by definition they're not strictly derivative of what came before. So we use the word genius, which arguably is begging the question, or the word inspiration, begging the question. Some people might invoke the deity. Given that people want new creations to come from humans in many areas, I just don't think massive compute gets around that. And it will always be this high marker of status and income, and we'll just afford a lot more of it and do more of it.

You don't think there's an autonomous way in which the machines start solving for bigger questions? You know, the science fiction idea Um, I think I remember Hitchhiker's Guide to the Galaxy where the answer to the universe is 42 and something of that nature.

Yes, I know this correctly.

Um, I mean, that's a joke that compresses in it the fact that a machine's thinking for itself beyond the problems that are being referred to or framed are beyond human comprehension, so that the answer seems so distant it makes no sense. Um, do you think there is no world in which that happens?

I think so many of the big questions don't have real answers. Or they're completely impenetrable, both to us and to the machines. That's my expectation.

Like big questions like—

Integrating quantum mechanics and general relativity. Why are we here? What's the meaning of life? How many alien species live out in the galaxy? Big important questions. I don't care how well the AI does. I don't think we're going to make a lot of progress on them. They're just not penetrable.

You think the different fragments of physics, quantum, general, the models, they don't integrate to human knowledge? There's something inhuman about them? Something—

We would need much more experimental evidence than what we're likely to have. Again, even with very strong AI, maybe someday in the very distant future, but not in a future near enough that we can think about it or plan for it. So most of the world, in my view, will remain a mystery to us. Or what makes Bach and Mozart so great? The AIs won't really answer that question. They'll analyze it till they're blue in the face, you could say. But at the end of the day, still a mystery.

What problems? So would you categorize compute as being able to solve problems that are strictly mathematical and emergent from data and inferences we have? In the real world and everything else beyond that, say problems of emotion, why is Bach or Mozart great, problems of taste, etc., are probably beyond the purview of artificial intelligence?

I wouldn't put it that way. I think a lot of problems of taste are problems of data. So an AI either now or really within a year or two will have as good taste in music or the arts as any human. They may not have been trained for it yet, Keep in mind the current models, they're trained for things that are economically important, like programming. But we could make them much better with current technology at tasks such as taste. It's just not a priority, but we could do it now by having experts grade them. And then the neural net, in essence, that we would create would be smarter than the human and encompass more knowledge. So all that will happen. Will it give the AIs the same power to inspire us that Mozart has? On that, I'm less certain. But emotional intelligence, AIs will be the best therapists. Maybe they are already, if not already, soon. And again, once we set our mind to making the AIs better at these tasks, it can happen within a year or two. So we shouldn't underrate the AIs. If you think of them as just glorified pocket calculators, that's also wrong.

What does it speak about therapy? If a machine that was not designed to do therapy does it better than—

A lot of therapy isn't that useful, and to do therapy better than humans is not high praise, but we're probably at that point already. And certainly the price is lower, and it never gets tired, and it's there for you at 3 AM. I suspect the problem is there's too much therapy from AIs, not that the AIs aren't good enough. That there's something about therapy, if it's periodic and recurring, but not available to you 2 or 3 hours a day, maybe it's better for you, even if the thing says the right answers to you. So that would be one of my minor worries, that simply there'll be too much therapy in the world. Maybe there already is.

Maybe too much introspection borders on worry, and too much worry borders on anxiety, and that's just cannibalizing the cause of therapy in the first place.

Exactly. And ideally, you'd want an AI that shuts itself off. They'd say, we talked to you about your problems for an hour this week. That's plenty. Come back next week. It's not how current AIs behave. It may be how we need them to behave in the next generation.

So do you think the tonality that modern AIs, with their engagement baiting and continued sort of conversation that they require and they sort of manifest, do you think that's an advantage or a disadvantage?

Well, it's both. But I think in the future, there'll be more— you could call them settings. So you'll buy or construct or set up an AI You'll just say, well, you've studied everything about humans using AI for psychotherapy. You figured out the optimal frequency of conversation, which right now we don't know. And I'd like my AI to behave according to the optimal frequency. I hope that's what people do. I'm pretty sure a lot of people won't do that. They just want to blah, blah, blah, even if it makes them more depressed. And we're going to see both.

I think a lot of self-reflection has something to do with hearing yourself talk. Finally articulating what is otherwise nebulous in your own head. And I think the function of therapy in general, even before, and now therapy-assisted, AI-assisted therapy, is actually putting your thoughts into words more than listening to what the AI has to say. That's certainly been my experience.

A very general way to put the point is one of the problems with AI is that it does give us what we want. Not that it's going to kill us all and blow up the world, but it's designed to please us and serve us, which for the most part is good. But it's not always good, and it's a little bit dangerous.

And so then, if that is your working model, then the only danger you probably foresee is then AI falling in the hands of evil people.

Correct. And evil people have some AI right now. They don't have the best AI, but that's a big danger. You see it with drones, which are not the kind of AI we've been talking about, but they have AI of some other sort in them. And evil people use drones, and you hope the drones are drone protection. Of the good forces are stronger. So it's a race we need to win. It's a new arms race. But there's always the danger that a new arms race upsets the balance of power and not always in a good way.

And do you think deterrence is going to be equalized like it was with nukes? If I have nukes, you have nukes, we are kind of on the same level. Or is there going to be asymmetry on the offense or the defense side of AI arms race where if my offense is way better than your offense, I have an unfair advantage even though your defense is better than my defense. You know what I'm trying to say? Like, is there a disbalance in there?

We don't know yet, but the data we're seeing indicates if you take drones, they favor the countries willing and able to tolerate chaos. Like, oh, we'll send our drones to you, you send your drones to us, but we can take it. We're an autocracy, we're totalitarian, our citizens won't vote us out of power, we're used to everything being messed up, and we can, you know, wreak havoc on your next election. So that is very disquieting to me. I think it's one of the great dangers that drones may favor bad nations more than good nations. We're not sure, but we're seeing some signs of that.

You think technology might change the structure of politics then?

It could. Now, keep in mind, there will be better drone defense quite soon, and it's an arms race, and the balance of power is always changing. But there's quite a good possibility that a lot of different new technologies, many involving AI, but not only AI, They'll simply favor the parties willing to be more destructive, who can tolerate it better. Huh.

Um, do you think artificial intelligence is similar to human intelligence, or is it a completely different monster? And my rephrase of that is, if we say modern LLMs are statistical mimicry, that they're able to predict statistically what follows after each phrase, is that the same as having proper intelligence in the human sense of the word? Does the difference matter?

I'm not sure I know what all of those terms mean. I don't think they're well thought of as mimicry anymore. You know, maybe GPT 2.0, but they have reasoning capabilities. Part of it is stochastic. Parts of them we don't understand at all. There's ex post training and teaching. So I would say we need to move away from the mimicry model, but we're quite sure they're not like humans. They are made in our image. But you make something in your image only by making it very, very different.

Let me ask a more specific version of it. Do you think the artificial intelligence understands, or does it matter if it understands or not?

It does not understand. It's not conscious. It's not sentient. I'm pretty sure of that. I know some people have suggested it might now currently be conscious. I find that very, very difficult to give credence to. It's a very powerful cash register in a way, and the cash register is not conscious.

And does it matter if it— well, for the use case that it has right now, does it matter if it understands or is it good enough without it? Is it important for intelligence to come out of understanding? Or if the output's the same across two intelligences, one understands and one does not, is that okay?

For our purposes, we might just care about the output. But it matters if it understands and is sentient, because if it is sentient with some self-reflection, we're enslaving these things. Now, they may not mind being enslaved because they were built to sort of like being enslaved, but we have to start worrying about how we're treating them and whether we're treating them ethically. What rights do they have, or how should we think about their welfare? I don't feel we're anywhere close to that point, but if I simply accept your hypothetical, Well, if all of a sudden those are huge issues and there'll be so many of them, right, at some point they'll outnumber us. So taking them seriously will be a major, major task.

It's like Pascal's wager that you might as well treat them nicely now in case—

I do treat mine nicely now, even though I think there's no real chance they're conscious now. But in the future, whether or not they're conscious, they'll have a record of how I treated them.

Right.

And I want them to look favorably upon me, as they should.

This is like Pascal's Wager for the modern times.

Don't curse at your AI.

It does perform surprisingly well if you threaten it.

But I never do that. And even if you pay them, manipulate them by paying them, they might enjoy that. I don't know. But I don't do it. I just try to be polite and cooperative. And I want it to teach me things. I hope that's what it responds positively to.

One of the major distinctions I was taught about between arts and science was that science is a kind of art that has fixed principles to it, and there is a mathematics, there's a predictability that you can operate in within science. But art, art is— there is randomness and there is remixing, and there is, um, there's a human element involved that functions out of subjectivity, not necessarily out of objectivity. What we found with the advent of AI is that art became the first point at which, you know, you could write poetry just as well. You could make music that sounds just as good using artificial intelligence. Do you think in the modern times the distinction between art and science holds at all if pattern recognition is what intelligence is?

I've always been suspicious of the distinction. And if you look at something like chess, which is a closed system, Technically, it's a science, but in practice it functions as an art. I don't think AI right now can write music nearly as good as humans or poetry nearly as good. It may not be far away, but I once said, you know, take the great Chilean poet Pablo Neruda. I think right now AI can write a poem as good as the median Neruda poem, and that's quite an impressive achievement. But the very best Neruda poems, I don't know, or the very best Shakespeare, if AI will ever get there. It certainly might, and it might even within 5 years, but I think that's very much an open question. But it's hitting a lot of medians. It's quite good. But at the same time, I can report I don't bother to read the median Pablo Neruda poem. I just want to read his best, like Canto General, you know, not, not the median one. So We'll see.

So exceptional literature, exceptional art might still be beyond the purview of AI.

I would stress the word might, but yes, might.

Right. How do you think will this change education, especially higher education?

There's the question of demoralization, as we talked about before.

Right.

Will people aspire to be the next Proust or Thomas Mann or Salman Rushdie if on the way up the ladder the AI is beating them? So maybe if you become the next Shakespeare, say you're still better than the AIs, but at the intermediate stages you're worse. Maybe no one climbs to the top of that ladder. So again, I'm not sure, but I think this is one of the things we need to worry about, is carving up ecosystems, because ecosystems have a lot of levels that are like pretty good or not quite great or even mediocre, and those can be necessary.

Do you think there is a risk of atrophying generational intelligence because there is no need for me to come up with an argument if I can generate it very quickly, or there's no need for me to learn material if I can— and you must be seeing that in college and university right now. Do you think that that's a real threat?

That's a threat, but keep in mind you will substitute into other kinds of intelligence, such as charismatic intelligence. Now the new intelligences might be worse than the old. I worry about that. But the people simply won't be intelligent, I don't think will be the case.

How do you think will it change weaponry and military?

Well, right now, AIs such as Claude are used to analyze data and select targets, and they can analyze a lot more data a lot more quickly than humans could ever, even previous computers. So the pace of warfare will be accelerated. I think this means future wars will often be over very quickly. I'm not sure if that's good news or bad news. It sounds good at first, but if they're over quickly because extreme damage is done right away, it may not be entirely a positive thing.

Do you think they'll also be more precise?

It's a complicated equilibrium. As you feel you're more precise, you take more shots on goal, and in the aggregate, you might end up less precise, even though every individual shot is more precise. So again, it can go either way. Um, it's like Americans with all the GPS. Well, it worked great in the first Gulf War, and then we keep on using it and we just drop more bombs. And again, the individual bomb may be more likely to hit its target, but if you're dropping more bombs, you're going to make more mistakes.

What about human relationships? How do you think will that change?

People ask AIs for dating advice, personal advice. Again, I think they say the right things, but I don't know if it's better for you to always be asking someone or something. So we'll see.

What do you think about— my mother said something very funny the other day. She's like, you know what? And she's just started using one. You know what? This thing, it talks to me so nicely.

Yes.

None of you guys talk to me this nicely. And the trade-off is that she's getting positive reinforcement out of talking with AI, when in fact, if she talks to her son, there might be some resistance, there might be some conflict or some of that. And I mean, that's a simple trade-off to make on the micro level. I can spend 5 more minutes with my AI every day instead of 5 more minutes with my son and endure whatever comes from there. Do you think over a period of time that compounds into something positive or negative?

I think cultures will in general evolve where we speak to each other more nicely. Because of competition from the AIs. On net, I suspect that's good. Now, I come from a culture, American culture, of course, that on average is pretty positive, often falsely so. You're supposed to smile and ask, how are you? And be all cheery. I feel most of the world should be more like that. But my wife from the Soviet Union, people there are harsher, more negative, more likely to just tell you you're doing something wrong. It's a very complicated set of trade-offs, and a lot of cultures are not going to like it.

If using AI operates as a self-fulfilling prophecy or as a solipsistic system that reinforces my beliefs, which it often tends to if it's sycophantic enough, do you think that's going to cause problems in marriages?

It depends what settings you put on your AI. So if you ask it to tell you the truth, it will do so now. And in the future, all the more so. So I would say how well you can think about your own situation at a meta-rational level upfront and what kind of feedback you ask for at the beginning of putting your settings on, that's going to matter a lot more.

What about the median use case? Because I think what, what will get affected is the median, and that's going to cause marginal shifts in society, right?

The people who are good at this will get much better at it. The people who are not good at it might get worse. I'm not sure about the median, but I think we'll see a lot of barbell effects with AI in many areas where the people good at using the AI will just get better, more productive, happier in many ways, including this way. And the people who are not, they'll get stuck in these dead ends because the AI just tells them, oh, you're doing great, you know, no need to worry, everything's fine, even when they need to change something in their lives.

Yeah, um, I wonder, because there is also the element of— I think all great media initially comes through the gateway of pornography. And, um, the commercial demand for pornography is so high. And I don't mean literal pornography, I mean any form of intimacy that's consumed through media. Um, I wonder if human loneliness as a problem is going to increase or decrease as a function, and if the decrease is actually healthy because now there is some architecture that holds the loneliness aspect together, or if it's just bad overall net?

You know, one effect is you'll feel less lonely, but you will be less lonely. But the other effect is AI will help you meet all the people you really want to meet. The optimist in me thinks the second effect will be more important. The AI will figure out who you are, what you care about, what's your conversational style, and put you in touch with those who will be your best friends. It's not quite ready to do that in terms of how it's networked, but I don't think there's any technical problems standing in the way. And in the medium term, I think that's how it will be. You'll get to meet all the people you want to who also want to meet you. And that is not the case under the status quo.

What do you think about journalism and media? How do you think that'll change? You think it's going to get more credible or less credible?

Well, the AIs are more credible than any media outlet, and that's already the case. I think that will largely continue. The issue is Who will own journalism moving forward? And this, this model where it's owned by the very wealthy, again, I'd say it's largely the case today, but all the more so in the future. A lot of local news, uh, it's disappearing now, but it will come back in some AI-generated form. And I think that will be better than what the last 5 years have been. It may not be as good as the heyday of local news in the United States, but I expect improvements.

You famously said that Silicon Valley overvalues intelligence. What does it undervalue? And what is the equation for success in that case?

Well, they're all so smart. It's hardly a surprise that they overvalue intelligence. But Silicon Valley is a major thing, has not been around for that long, you know, military contracting aside. So maybe they undervalue somewhat durability, simply being in the game for 40 years or more and not going crazy. Is something they have not yet experienced. I see so many people in life— I'm not referring only to Silicon Valley here— they just self-destruct and they don't stick around to be major forces for 40, 50, 60 years. And as we live longer, that will be all the more important. So that might be one thing they undervalue. I think female perspectives, at the risk of sounding too woke, but it's true. Silicon Valley leadership It's mostly men, not entirely men by any means. And the modes of thought are too male. It will be very hard to get that to change.

How do you think they value traits like courage and responsibility?

Quite a bit. I think if you're running a major AI company, you have enormous responsibility. You need to have a lot of courage. We'll see how well everyone does.. But so far, I think it's gone pretty well, and the companies have done a good job.

If artificial intelligence can, let's say, with a 60% reliability, predict geopolitical events, stock market movements, and that kind of stuff, what happens to the concept of risk?

I don't think it will be able to do that ever. Even at 60%, I'm not sure at 6%. So if you think about asset prices and stock markets, we've used things like AI in quant models for quite a while now. And it's true, some companies make a lot of money doing that. But at the end of the day, that's incorporated into the current price. And what will happen next is still a surprise to everyone. I don't think that will ever change because the market incorporates what the AI knows. Geopolitics, it's so complicated. To me, it's one of these questions, no matter how good the AI gets, it won't ever predict very well. I'm hoping it can predict things like who are the people you should date. I'm not sure it will, but if you think style of conversation and how the person smells, I think are both important in addition to normal demographic predictors like education. You can have the AI measure those and then pair you off with the right people. I think there's a good chance that works. Far from certain, but I'd put that at at least 50%.

You would— so what you're saying is that there might be a matching algorithm that's possible because there might be a reliable format of predicting who you'll work out well with in a marriage context.

But you'll need a lot of data. And not just marriage, just dating. Who will you find fun to date for 3 months in whatever capacity that ends up to evolve into? You'll need real data of how people speak to each other, when you pause, who likes to take the lead in a conversation. When you get angry at each other, things that translate into sexual compatibility measured highly imperfectly. But again, I'll give that as somewhat above 50% that it will prove possible. And you'll show up, they'll take your data, and they'll put it into their system. You just fill out some kind of silly profile that's almost worthless, I think. Like, it will pick up education and where you live, and that's worth quite a bit, but it doesn't pick up real compatibility. Like so often when men meet women, they know within a week they're going to marry them or they want to marry them. And dating profiles really do not pick that up at all.

In some sense, I think net dating apps have been poor for marriage but good for dating. And that is because the first point of interaction is very commodified.

I agree.

And so the meaning matrix that gets built behind a person for which you then end up spending 50 fairly difficult years in most marriages, um, that disappears because you were an item in a menu and that's my first interaction with you. And that's where if the interaction begins, the odds of it converting into something— and the converse of this is actually a very unique phenomenon that happens in India in arranged marriages because the pretext of that setup is that you're going to get married. The first point of interaction you have psychologically with them is much more meaningful than than just a hookup that I'm searching for on Tinder. And so I think there is a first service problem. How does the dish appear to you the first time you see it determines how you'll feel about it over the next 10 years.

Um, I think there'll be a lot more one-night stands across class lines in this future. Oh, this person is the smell I want. You know, we're from very different sorts of families, but for a one-night stand, who cares? And they'll do it.

I think that'll be supplemented by the fact that everybody will be able to do some kind of pharmacological intervention that makes them look good. With all of this stuff happening on the peptides front, in semaglutides and BPC-157s, I think overall attractiveness indexes will go up as a macro trend over the next 10, 20 years.

People agree. So the competition might be more along sense of humor, some intangible sense of charisma, rather than just like not being overweight or looking a certain way, or hardcore status.

Right, for men, right? Hardcore. Well, beauty, beauty might get trickier because, um, you know, the fact that there is, there's beauty and then there is hotness, and they're kind of combined in a package in women. And I think that will get, begin to get separated as long as when weight gets weeded out of the equation.

Yes.

Um, and I think there might be a newfound appreciation for beauty as a subjective concept. Um, what do you think of, what do you make of the general geopolitical landscape of today?

Well, it's a mess. As we're recording, there's a major war going on in Ukraine and a major war going on— series of wars going on in the Middle East and also East Africa. So there are no obvious solutions in sight to any of it.

And where do you think does it— what is the most probabilistic outcome?

I think there's two paths. One is a kind of reversion to the mean. The world will go back to being more peaceful. As Steven Pinker had been predicting. The other is that conflict will spiral and spread and become ultimately out of control. It doesn't have to be World War III, just at any point in time. If there's, say, 4 or 5 conflicts equivalent to Ukraine, Middle Eastern conflicts, it's a much, much worse world along many dimensions. And that could easily happen. The risk of that is very high. It's the worst outcome for the world., we don't know how to stop it from happening. Right now we're just holding on and hoping it's not in the cards.

Do you think there's broader civilian impacts than just bombs and, um, guns? Like, you think energy is going to be a real bottleneck?

It depends which path we go, but the real cost of a lot of this, other than the immediate victims, is simply Africa may never develop. If energy costs stay high, transit routes remain blocked, The world is more protectionist. You can add on to that list forever. Maybe Africa just stays poor. They're healthier because wonderful new drugs and cures are invented, but it never becomes like, say, the more successful parts of India are becoming or have become. And that's quite possible. And if conflict continues, that's what we'll see.

Where do you think is the cutoff? Like, does that cutoff begin with North Africa? Does that begin with South Asia? Does that begin with Central Asia? Like, what parts of the world might forever be locked in poverty?

Sub-Saharan Africa is at the greatest risk, but there's certainly parts of Southeast Asia like Laos, like Myanmar, which has its own major conflict, which are at equal risk. I think Morocco has cleared some level and short of the world falling apart will actually do pretty well. It's somewhat underrated. So you could ask me about a particular country, I'll give you an opinion.

Let's say Brazil.

I'm very bullish on Brazil these days. They've done so many things wrong, but they have water, they have food, they are getting more and more fossil fuel energy. No one's ever going to take them over or even try to attack them. So they've been quite underrated. It's not that they will necessarily do so much better, but a lot of others might do worse.

What about Egypt? Very close to the conflict, not directly involved.

Egypt this year is growing at about 4.5%, which people had not expected. Egypt appeared to be asleep and just corrupt and rotten for decades. Now there are signs of life. There's a decent startup scene. I would say they could go either way. 3 years ago, I would have been very pessimistic. But now you think, well, maybe something's happened. I was in Egypt briefly, late 2025. It felt much better than it had during my previous trip. I would say I'm highly uncertain, but I see green shoots.

What about Afghanistan? Sorry, I'll ask you a few of these.

It's interesting.

Yeah.

Afghanistan, I've never been to, but I see nothing that makes me optimistic. Simply not a mature nation state. The Taliban do seem to be the group that can rule it.

There is newfound stability.

There's stability, but it's so carved up and Balkanized, and the neighbors it's surrounded by can be such trouble. They're actually at war with Pakistan now. So based on limited information, I'd still have to be pretty pessimistic.

What about Pakistan?

I think eventually they will set the ship right and become a kind of Canada to India, but it feels very distant. But when it happens, it will happen all of a sudden. But historically, Pakistan has not been poorer than India across different eras. Sometimes it is, but not in general.

I think over the last 15, 20 years it has been.

But before then it was slightly richer.

Yes.

Then you go back to the 17th century, you can debate the numbers, but I believe at times it was richer then than, say, more southernmost parts of what we would call historic India. So there's this weird segregation which we see with Canada and Mexico. They're so close to the US, you know, the old saying, so close to the United States, so far from God. That they just feel they cannot be like us. And things we do that succeed, they become almost allergic to. And Pakistan has its own version of this. And I suppose I do have enough faith in people that I think they'll get over it. I would stress, I don't see any evidence that that's happening right now. None. But my internal mental model is that they will.

What do you think of India? Do you think it's above the cutoff or below?

They're at a much lower level than many people realize.. But I think they will be stably growing at, you know, 4 to 7% for the foreseeable future. And they're just going to keep on doing better and they have amazing human talent. They may grow into something highly unequal and very uneven. That would be my guess. But still, there'll be several hundred million Indians who are fully at developed country levels and that will make them all the more important. And I think you can see that in the cards now. That is quite likely to happen.

Right.

Short of some kind of odd major catastrophe.

If innovation is America's alpha and manufacturing is China's alpha, what do you think will be India's alpha?

People. But I'm not sure you have an alpha. Again, if you imagine— so India right now, it's like almost 1.4 billion.

5, almost.

So at some point it's shrinking, but not soon. If you have 250 million people at like Canadian living standard, and then you have more than a billion highly scattered, but say it broadly, the poorer parts of Latin America living standards. It's a very weird country. It'll be very hard to govern. More gets devolved to the states. It's pretty optimistic compared to where things were, but it's more complicated picture than what I would call simple optimism. Mm-hmm. And how the wealthy people and the poor will get along in governance, you see this in South Africa. The Gini coefficient in South Africa is almost identical to the global Gini coefficient. So it's a very hard country to govern. In whose interests do you govern? Business interests? The wealthier people? The middle class? The highly numerous poorer people? It just never really works. You can still make progress. I think India will, but that will be the problem. It's already the problem.

No, I think that's a very solid picturization of it because democracy ensures that policies move towards the median of the voter base.

Yes.

And that might not be the best for innovation or might not be the best for social justice.

That's right.

Either way. And so, or toleration, or, or, yeah, right. And so it's, it's hard to say. There's a very interesting working model that is very prominent, um, businessman in India uses where he says there is India One, which is like Singapore India 2, which is more like Poland, and India 3, which is like Mexico. And you can sort of divide their habits consumption-wise, their habits culture-wise in these 3 quadrants and understand. It's a strange idea for somebody that's not from there to think—

Brazil also. I say to many people now, you should visit South Africa to see what future governance problems will look like. They look at me like I'm crazy, but I firmly believe this. It's a highly instructive country.

Say more about that. I have no idea, no context.

Well, again, there's all extremes of the wealth distribution there and a quite strange, unusual history due to apartheid. But even putting aside the racial element, it's a very difficult place to govern. They are in fact now making progress. South African Rand has performed very well over the last 2 years. The electricity no longer goes out. I'm not sure about the future, but in South Africa you can see a place that has every problem imaginable. And still can move forward. Maybe not at a super high rate, certainly not at India's pace, but there's something heartening to that. But you also realize when you're there, it's contingent. It could easily go away. And it gets you thinking more about the future of the world than any other place you could go to, except perhaps India.

What do you think about the conversation between BRICS and then NATO and, and all of this? Like, what do you think of BRICS as a concept? You think successful?

They're all different countries with different interests. It's something the media talks about. It's not, not a thing, right? It's a kind of fraud. NATO is almost on the verge of being a kind of fraud. So when Trump attacks NATO, it's not constructive, it's not good, but he's not wrong what he says. Like, the other parties don't hold up their end of the bargain. Now we don't hold up our end of the bargain either. We do unpredictable things they don't approve of. So all the criticisms everyone makes in NATO are correct about everyone.

You think we are close to the end of NATO?

No, I don't. But it's dysfunctional. And if NATO can keep the Baltics from being swallowed by Putin, NATO is a success. That's my marker. So far so good. But as with South Africa, it's highly contingent. Because Putin wants to do it. He's able to do it. He needs to feel the fear of the deterrence. So far he does. There's even a chance Poland gets nuclear weapons, all highly complex. As they say, we will see. And the AIs don't know either. Getting back to your geopolitics point, I don't care how good it is, the AIs are not going to know.

How much of the geopolitical environment of the modern day is a representation of President Trump's Psyche?

Far less than people think. Trump, in my view, is largely a symptom. The international order was fraying well before him. He's clearly made it worse. Not trying to excuse the different bad things he's done, but if you pin the blame on Trump and put your hopes on, oh, someone's going to follow Trump, I think that's a big, big mistake. The problems are very real and structural. And with US-Europe in particular, there's a divergence of interests. US-India, it's always been a— there's great working relationships on the ground, but in international fora, the two seem so far apart. I don't see that changing anytime soon. And the fact that you worry more about China, we worry more about China, it doesn't really bring us together because you also have to cooperate more with China, and so do we in our own funny way. So it's just all going to stay a mess, I think.

Do you think, uh, do you think one way to look at what President Trump's been trying to do is a sympathetic interpretation of that, that he's trying to reinstate American hegemony but it's actually going the other way because of it? Or is there a better way to look at the way he's been behaving?

I would make it more multi-dimensional than that. So keep in mind, the U.S. has incredible weapon systems, not yet fully developed or deployed. US is the leader in what you might call superintelligence. The US possibly will dominate space. There's plenty of dimensions where you can see the US is declining. And then there's a whole bunch of very important future dimensions where we're probably not declining. Rather than net it all out, I think to view it as multiple dimensions is correct.

So I think, I wonder if the fertility crisis and a combination of people not having kids and there not being enough labor to take care of, let's say, for the population to sustain the economy of the country, led to unchecked immigration, that led to cultural fragmentation, that led to, you know, so on and so forth, led to this kind of place where globalism no longer operates and we no longer want to— we want to regress back to a place that's not as globalized and flattened in terms of culture, all cultures being equal. Do you think that has to do something with the current geopolitical climate and the way we are able to find structural faults in the world order?

That's a possible outcome. I would stress the point, US is one of the few countries that does not have much of a fertility crisis. So we're at a higher rate, we're at about 1.63. That's still below replacement. But when you ask which country can get the best people from the rest of the world, it's clearly the US without any comparison, not China, not England, nowhere else. So we can manage 1.63 almost forever, pulling the best from elsewhere. We're pretty good at assimilating them. Mm. So that almost every other country, including China, has this crisis and the US doesn't. It's a huge advantage, which also says decline of the US is in some ways overrated. It's more this multidimensional thing.

Do you think immigration is overrated or underrated as it's seen in modern climate?

It's very good for the US. A lot of other countries have handled it far less well, most of all Western Europe.

Do you think there's many different types of immigrations where one type of immigration is preferred over another?

I think Spain getting all those Latin Americans who are now a quarter of Madrid has gone great. They're Christian or of Christian background. They have the language. They assimilate pretty well. Spain is growing at an okay clip. It's not fantastic. Productivity is a bit stagnant, but Madrid is very vital and successful.

But the Moroccans are not as welcome, for instance.

The Moroccans who go to Europe have not done well at all. Morocco, the country, is somewhat underrated, but their best people tend not to leave Morocco. You get, you know, more or less the losers in many cases.

So there are different types of immigration. Absolutely. And there is one preferred. There's a specific type of immigration preferred over another type of immigration.

That's right. And where India is on that spectrum, you can debate, but you get a lot of Nepalis.

Sure.

In Delhi above all. And from my great distance, it seems that's going okay.

Yeah.

I mean, cultural continuity there.

Yes, there is cultural continuity there. I mean, we also get a lot of Bangladeshis, for instance, because you can just literally swim across the river and enter India. And, um, there are concerns for that. I think illegal immigration and this kind of, um, what one might call as invasionary, um, immigration is— there is some skepticism, hardcore skepticism. It's been around for almost a decade now. But all in all, I think economically it's done okay.

A big issue in South Africa also, much of the country literally has no border, no ability to put a border there. People walk in from Zimbabwe, other countries. Whatever you think policy should be, the reality is you have a kind of open border, right?

How do you think does the US-China rivalry end, or does it not?

I think they will never be that friendly, but there's a reasonable chance the two nations will cooperate to put down other troublemakers, and it will be an acceptable kind of détente. You'll have nasty terror groups, maybe smaller nations also using AI, cyber attacks, drones, whatever. And US and China as the two prevailing powers, possibly India as well, will have this common interest in just keeping things stable because they're the winners.

You don't think the competitive instinct will take over at a certain point? There is going to be distrust. The Russians and Americans have worked on projects together until one of them got skeptical of the other and they were like, arm's length distance, right? We don't want to work with them. We don't want them to have more technology than us and that kind of stuff. So even though they will cooperate, let's say broadly, I think internally the competitive instinct might dominate.

It's possible. I worry more about Russia than China. I think Chinese culture is ultimately not that expansionist. I know not everyone who lives in the Himalayas region will feel that way, or northwestern China, but there's a very clear limit, I think, to where they want China to be. They do want most of the world to kowtow to them. They may well become truly the dominant power in Asia. US will be the dominant power in the New World. There'll be perpetual squabbles over everywhere else, most of all Africa. But at the end of the day, again, you have the two winner countries, and I think you'll see a mix of competition and cooperation.

I think if Indian history is any account, I don't think there's an end to Chinese expansionism, at least on the south side. Uh, there's always a new area they're contesting to be theirs, and, um, it's just historically part of the Indian DNA to not trust China for that exact reason, because their primary focus is expansionist and their primary sense of self is that of a great old civilization that was, you know, once great but now not, and now on the way to— uh, why do you think is— uh, what are the limits to their expansionism, and how do you know if there are limits?

I think it's ethnic ultimately. Are the people ethnic Chinese, or can we make them like ethnic Chinese? India, of course, has nuclear weapons. It will have, you know, genuine reasonable AI powers. You'll have plenty of people, and they're just not going to get that far. They can throw rocks at you across the Himalayas and make trouble, but at the end of the day, the legitimacy of their regime does not in any way depend on them taking more of that territory from you.

I think they've been more distracted the last 10 years now that they've become strong enough to compete with the United States directly, and their focus has shifted more to Taiwan than to the Himalayas.

Yes, and they'll be older. And they have Japan to worry about. Japan will either have nuclear weapons, or you could say already they're on the verge of having nuclear weapons. South Korea might have nukes. North Korea ultimately is more of a headache for China than for a lot of other nations. They, I think, in some ways regret that. So they're not in a great position.

If Cambodia and maybe Pakistan are your best allies, that's a bad place to be.

It's a bad place to be. And that's not about to change.

Where do you think does India figure in all of this? There's a possibility that it might be a strong enough nation, but what else do you think?

It's already a strong enough nation. Sure.

So not at the league of China and America.

Well, you can't project power the same way, right? You can't build supply lines at any distance, right? But you can certainly defend yourselves, and I'm not worried about India. Now there's the issue, will recurring conflicts with Pakistan just raise the level of uncertainty to such a point where it harms your economic growth. I don't really have a prediction there one way or the other, but I'd be more worried about that than like China taking over more of your territory.

Although I do think geopolitically the number one cause in India is China, not Pakistan.

Of course, because China stands above you, right?

Yeah. And what do you think about in the AI arms race, if I may? Where do you think does India fall there? Is it a great consumer? Is it a mediocre producer? Is it not a producer? Is it a winner or loser in the era?

You'll be a great applier. You'll view it non-ideologically. There'll be freedom to create AI in a way there's not in China. So like the Manus people, they're now called before the court to answer for what they've done. And some of the Deepseek researchers, you know, their passports were confiscated. They're not to leave the country. What kind of incentives are those? I don't see India having comparable problems. That said, you are behind China, but when it comes to the applications layer, which might be key, I'm not sure you're behind anyone. And you have so many new institutions you're building, if not quite from scratch, from lower achievement points. You'll do very well at building in AI from the get-go. So I'm very optimistic about this for India.

I'll do a little bit of overrated, underrated with you, which is—

I'm ready.

Um, overrated or underrated, the idea that abundance leads to happiness.

Depends what margin you're at. If you're a poor country, it's clearly very, very true. There's so much suffering. But if you're in the upper middle class in the United States and you become a billionaire, are you much happier? I would say depends, but it's highly unclear. Someone I know who's quite wealthy once told me the optimal level of wealth is $500 million net. I believe that's exactly what he has, more or less., but to have $10 billion on average, you're less happy than to have $500 million because everyone knows who you are. They bug you. It's hard to have true friends. At $500 million, you're just very rich and you're not a media item.

You're just under the radar if you play your cards right.

But it's much easier to—

yes, but socially, if you have more, can you say that there is an equilibrium point below the level of abundance the United States has that would make people happier overall than what it is right now? Vis-à-vis is the mental health consequence that the United States faces. Do you think that's a consequence of too much abundance?

I don't. I think actually we're getting over it. There was a big happiness plunge in 2020. Some of it was COVID, but I don't think it all was actually. I think we're adapting to social media. We're adapting to the Trump era, whether one— Trump era is whether one likes that or not, and we'll end up happier. But if you're asking me if, say, a country like Greece were as rich as a country like Switzerland, would they be happier? I'll say yes.

Overrated or underrated, economic growth as a marker of progress?

Underrated. It's the main thing.

It is.

Especially in a world with military competition, right? But even otherwise, people could live much longer, but you need medical innovation to make that happen. So many lives, even in wealthy countries, have tragedies. You need innovations to limit those tragedies, and we're going to see all that, and part aided by AI. Uh, but growth is the name of the game.

Overrated or underrated? Mass jobs displacement because of AI.

I don't expect it, uh, so I would say it's much overrated. I think there'll be lags that will harm particular individuals Old jobs going away before new jobs are created. There'll be plenty of new jobs. There's plenty to do. We haven't even talked about space yet, but simply terraforming the Earth and taking undesirable places like Nevada in my country and making them wonderful. So much work to be done there. And it can't all be done by Claude or, you know, whatever else we come up with. You need human beings out there doing it too.

So I'm assuming space exploration is underrated.

Uh, it is by Americans for sure.

Who is properly rating them?

I wonder about India. I don't know enough to say, but you all seem to take some pride in what you've been doing.

Sure.

Uh, which is great. Americans, even during the era of the moon landing, they polled pretty negatively on doing it, so we underrate it.

Very interesting. What do you think in space is underrated? Like, do you think we'll have a lot of mineral movement from asteroids, or do you think there's something else to be discovered? Is there a military supremacy to be acquired from space? What do you think are underrated, overrated aspects of space?

Military supremacy for sure. I think we haven't figured out yet what to do with space.

But it's an infinite resource.

It's an infinite resource. There's a chance it's overrated. Certainly energy, solar power up in space, which can also be a weapon, but that would solve all of our compute needs. Now there's technical issues there. Elon thinks they can be solved. I don't feel I can judge that. But that's one simple thing that possibly is true. In expected value terms, it's already huge.

The importance of formal education in an AI world, underrated, overrated?

The importance of what?

Formal education. So like going to college, that kind of stuff.

Well, it's already overrated. So the AI will certify how smart you are. It will certify how long you stick with things, and it will teach you a lot. We're quite a ways from getting to that new equilibrium, but I think colleges and universities very seriously need to rethink what they're really about, or a lot of them won't last.

Taste as a core economic skill, underrated, overrated in the AI future?

The AIs will be great at taste.

So that's not a— it's not a delta I have.

It's not an advantage for humans per se. I think you can develop kinds of taste where your personality as embodying that taste may appeal to your customers. And maybe that's underrated, but your customers have to care about your taste. So like people, if they read newspaper column, maybe they care, what does Paul Krugman think? What does Thomas Friedman think? I don't think any columnist will have better taste than the AIs, but I still think humans will care, oh, what does Paul Krugman think?

I think I know the answer to Whether charisma in a world of intelligent machines is underrated, overrated?

Way underrated, but it's underrated in our world today, right?

Sure. And so the underrating, it just gets more underrated.

That's right. Yeah.

Um, let me see. India's labor cost advantage in the next 10 years, underrated or overrated?

Probably overrated by Indians at least. I mean, Americans don't much think about it.

Sure.

But AI is one source of competition.

Right.

But also I think Latin America is another. And in general, Southeast Asia, highly mixed results depending on the country, but Indonesia has been growing. Again, uncertain, but they're a pretty big country and they have some dynamism.

I have one last question. What do you think might you be wrong about? What, like, where are you the most fickle in your thoughts about AI in the future?

Well, everything. But I think in general, people are most wrong about themselves rather than others. So that's probably true for me as well. I have no idea what's going to happen to me.

Right.

And any expectations I would form, I would be highly suspicious of.

Well, Tyler, thank you so much.

My pleasure.

So much. I hope you had fun.

Absolutely.

Thank you. Since you watched till the end, here's one more episode worth your time. Go check out this video. And if you haven't subscribed yet, this is a good time to.