Tyler Cowen on the Future of AI

Well, so do you think AI will lead less people to want to have pets?

I think AI will lead people to have fewer friends. I think you can interact with your pet and your AI at the same time. So maybe pets and AIs are complements.

I don't know. I just, I feel like you, there's gonna be some more like emotional support you get that currently you have these downsides of like having to like pick up poop and whatever.

But the pet is also physical emotional support, right? So the dog sits on your lap, the cat purrs, and the less time you're spending with your friends, the more time with your AIs, the more you might want physical interaction with your pets. Maybe demand for hamsters will go down.

Okay. But you, but like the weight and the heat and the furiness you think is like a key part of the value proposition.

Absolutely. Pets Bar you can use, by the way. I wasn't sure if that was the show.

Yeah. Oh yeah. That one was good.

Sort of spontaneous non-show show element of shows.

So Tyler Cowen is my greatest intellectual influence. I'm reading his blog daily, Marginal Revolution, for over 15 years now. His posts exploring China in the late aughts were a not insignificant inspiration for me moving to China in the first place. Um, full disclosure, Tyler through Emergent Ventures has supported China Talk in the past, but this interview is not payola. Tyler is in fact my dream guest who all these years I have been too scared to invite on the show, uh, not thinking that my skills were necessarily up to, uh, par to compete with all the other fantastic interviews Tyler has given over the years. But you know what? We're doing it. Tyler Cowen, welcome to China Talk.

Thank you for the kind words. Eh, I would have done it sooner. And B, I had no idea you'd been reading for 15 years.

It's, um, scary, but it's like reading you, but it's, it's the reading the links as much as it is you. So it's, it's like your universe living in, um, which is a little different than like, you know, if you were just only writing words and I was only ingesting your thoughts. So I feel like it's, it's actually like a way more broadening than a narrowing, than narrowing experience over all this time.

I once gave a talk in DC, I forget to which agency, but all I did was print out my early China Fact of the Day posts, and I just read them in a sequence, like China, China Fact of the Day. And those were in the times when people hadn't fully realized what a big deal China was. And, uh, listeners were shocked to hear something like every so many months, like China is adding a level of power consumption equal to the economy of Brazil. Now that's old hat and it's no longer true because the rate has so slowed down. Uh, but that was a fun talk. One of my favorite talks I've ever given, and I was purely editor. Editors are underrated is one of my views.

All right. So, uh, you gave us, you gave us our transition, Tyler. Um, we're gonna start with AI. Um, we're recording this December 13th, uh, a few weeks after ChatGPT has dropped. Um, one of the, uh, things that you've said repeatedly in past podcasts, uh, and interviews is that you're sort of on the one hand thankful that you grew up and matured intellectually pre-internet. Um, I feel like sitting here today, I am perhaps the last generation to have grown up and matured intellectually pre-AI tools. Um, what do you think the, um, you know, what do you think people will gain and lose from that transition from having to generate everything to, to, to using these types of, uh, um, types of tools in perhaps more of an editorial sense?

We'll use devices such as ChatGPT to stimulate our own thinking, to help us with lateral thought, to edit, to summarize. They will serve as tutors. For some questions, they're a better Google. So a friend of mine asked me, why is the Ontario Teachers pension system so large? I asked it, of course, right? Easier than Google. Another person asked me, and I'm not sure here it gave the right answer, in the lifetime of Tchaikovsky, did people know Tchaikovsky was gay? It said no. Uh, but even if that's not the correct answer, it's telling you something about what it trained on, correct? And they will do much more writing. I think they'll very much change the traits of successful public intellectuals. It will revolutionize the fields you and I operate in, and just be good for many, many other things we haven't even thought of yet.

But a bigger question of just like what it will— okay, so, so let's say on that public intellectuals thing, um, what will be more and less important, I don't know, 5 years from now?

I think sheer writing will be less important, even if the AI can't quite copy, you know, say Paul Krugman or Joe Stiglitz or, or whoever the person is. But charisma will be more important. Either originality or pretense to originality will be more important. Standing out from the crowd, extroversion, a certain kind of flair, a willingness to engage with one's own celebrity, which the AI cannot really in the same way do. I suspect all of those traits will become more important. The simple regurgitation of information, uh, which you see in a lot of Substacks with some analysis, but that's likely to become less important.

How about coming back to the sort of, um, like intellectual development, um, piece of this. Um, what do you think that not having the internet gate for the first few decades of your life gave you? And, um, how is that potentially going to compare to people born in 2022 who are going to be growing up with, um, you know, all these artificial intelligence brain appendages?

Well, I was born in 1962 and I didn't even have email until I was 30 and sort of some version of the internet as we know it, maybe I had in my late 30s or at age 40. So I just spent a great deal of time reading classic books, rereading classic books, studying them. I would go on long trips and entirely have the time to myself. Rather than having to deal with email or online responsibilities. And again, for 30, 40 years, you learn how to lose, use a library, learn how to do particular kinds of research. Uh, your attention span is strengthened for the most part, and you become more curious about the physical world. And I think I've mostly kept that. At the same time, I've now been immersed for 20 years in the internet, and that's a, a weird combination. It was Ben Casnocha. Who first pointed this out to me.

Um, so you're not gonna try? What do you think? Uh, uh, 2023, the, uh, um, the newborn baby who is, isn't just sort of raised on the internet, but raised on AI.

It's very hard to predict. So if you look at my predictions about the internet 20 years ago, if I even had them, I don't think they would have been especially strong, but that baby will be used to very individualized forms of tutoring. And I think there'll be a bimodal distribution. There will be babies who grow up not, you know, putting their behind in the chair and taking the initiative to learn. And for them, life will actually be more like pre-internet life. And then there'll be those who use the AI as an individualized tutor, and they'll just become much smarter more quickly. And they'll do more amazing things, but they'll do things that don't compete so much with the AI, would be my offhand guess.

You mentioned growing up reading classic, um, uh, novels and scholarship. Um, what do you think will be relevant and not relevant about that sort of stuff in our new AI world?

I suspect the classic texts will reemerge in value. So say reading Plato or reading Kant or reading Adam Smith, gives you a sense of a vision and big picture thinking that it, it seems for some while the AIs won't be able to give us, maybe never. Or simply scanning the internet for facts, the AI might give you a very good digest, which you'll consume in less time, and you'll then seek out the thing the AI can't give you at all. And that will again be radically original big picture thinking. So my hope, not really a prediction, but my hope is people will go back to the great classic texts and study them. A little worried people will just say like, oh, Plato's Republic. Oh, please. You know, Mr. AI, Miss AI, you know, spit out the summarized version for me. That's my nightmare. It could go that way too.

So what do you think sort of like middle school and high school history, literature, English instruction will end up turning into?

Well, I think it will be more privatized. There will be greater use of oral exams. Giving people paper topics to write. Uh, maybe you will do it in class, but I think that will radically change quite quickly. It already should change. And maybe there will be more playtime in better schools, but I think the education each individual child gets will be more different than, say, you know, has been the case as of late. That's already a trend with the internet, uh, that you can customize things, what you read. Now you'll literally be able to design your own education. By speaking to your AI and your AI can be trained on datasets that you or the parents want. So you'll have your own personalized, you know, familiar, you could call it to refer to the old world of witchcraft. And it will do things for you. Uh, it's scary too, right? I'm mostly excited, mostly positive, but one would be naive to only see the plus side of this.

How would you think about the net impact on, you know, human psychologies, like what, like what happens to suicides 5, 10 years from now, thanks to, um, you know, trying to isolate AI as a, as an independent variable.

It's very hard to predict, you know, in which country suicide will be high and low. None of my current intuitions seem validated. So you have countries with somehow odd temperaments. Is it, I think Hungary has a high suicide rate. Well, why is that? I don't have any good theory that explains it. Often wealthier countries have higher rates than poorer countries where people are less happy. Maybe it's something to do with the peer effects or your expectations. Uh, but I would say in many ways, people growing up will feel belittled and perhaps they will do more sports or do more gardening, or they'll, they'll do more fencing, or they'll do things to establish their originality in fields where the AI cannot readily copy them. So again, in some ways it may be more of a, a 19th century education in, in the aristocratic sense. You do a grand tour of Europe, uh, you learn how to punt on the Thames or something. Uh, again, that those are just my intuitions. They're a bit weaker than predictions, but I suppose I expect them more than some of the other scenarios.

So like more like coming back into the body than being kind content to live online because if you're just online all the time, you're constantly, unless you're sort of a second or third standard deviation person, you're constantly getting sort of bombarded with the computer being better at you, better than you at anything you could possibly imagine doing.

Yeah. That'd be like playing chess against the computers, which most people don't do. They might in a training sense. So an interesting question is, would you program your AI at a quote unquote stupider level and train yourself in against it in debate to make you a better debater, and then keep on trying to beat it at higher and higher levels of AI intelligence. So is that a way people will use to try to become better debaters? Uh, probably. It's interesting that chess computers are nearly perfect, but it's still the humans we watch. So earlier today I was watching Magnus Carlsen beat, uh, Caruana. I could have watched the computers. But it didn't interest me. But that makes prediction hard also.

So Tyler, um, I played chess until I was like 7, and when I, you know, blundered a queen in the last game of a tournament and like decided I would never play again. Um, and, uh, fast forward 25 years later, um, after the, uh, uh, sort of cheating scandal over the past few months, and I started picking it up again. Um, what do you think are the, what do you think going from 0 to like 1,200 or 1,300, uh, trains you in mentally?

I don't know. Probably nothing. Maybe it trains you in sitting still a bit. Uh, but at a level of 1,200 or 1,300, it's often hard to understand why you lost. You might see that you blundered your queen or a rook, but in some deeper sense, that's endogenous how you got into a position. Where blundering your rook was so easy to do is, is maybe harder to grasp. I think at ratings like 1700, 1800, people really start learning why they lost and they become more meta-rational. They, they realize when they should defer to a better opponent or to a computer. So it does happen.

So how do you think AI's impacts are going to be distributed across developed and developing countries?

I don't have a good sense of that. In many much poorer countries, uh, I would suspect they'll have free or near-free access. And a lot of ordinary intellectual work will be done by the AIs. And in an infant industry argument sort of way, maybe that will limit their ability to produce homegrown intellectuals. Because the idea that you get your start by doing 4 years of journalism, and then you write a column, and then later on you're writing books. A very typical path, especially in poorer countries, uh, that might be harder to do. Maybe there'll be fewer opportunities. Writing will be more the province of the wealthy. Um, again, just an intuition, but that's what comes to mind.

What about from a sort of like, like, uh, institutional upgrading angle where, um, you know, you can sort of short circuit bad teachers and, um, inefficient bureaucracies by all of a sudden sort of like putting AI into the, into the bloodstreams and then you're sort of like gains you get from governance actually are potentially higher in the developing than the developed world.

There'll be some of that, but keep in mind you can do that now with the internet and plenty of people in whichever country don't. So it's the willingness to do it as you might learn from your parents that is scarce, that piece of context. And I'm not sure the AI helped with that a lot. If the AI makes that process a lot more fun, It might, but there's already YouTube and YouTube is the world's greatest educational vehicle of all time. And plenty of people do it. Plenty of people just use it for fun, for music, maybe for bad purposes. So, you know, like how much is AI an upgrade over YouTube? I think at first for poorer countries, not much at all. In some ways it might be worse because it requires more literacy. Now pretty soon there'll be some kind of integration of these AI systems with video, with voice transcription. You know, with search, with, with other things, and that itself will matter if, but if you just take pure text AI, I don't think it's a game changer in those places.

What do you, what do you think AI tools are gonna do at the top end of cultural production? How much better can we hope things, you know, books, movies, novels, video games to get?

People will have access to a lot more ideas and a lot more tools and a lot more techniques. And they will do amazing things with that access. But at the end of the day, a lot of artistic creativity comes from scarcity. Art Tatum sitting down at a piano, uh, a Florentine Renaissance painter, you know, with tempera paint and a board to work with. So I'm not sure that quality will go up. I don't necessarily expect that. There will be a certain kind of elaborateness of creation and presentation. That will be explored to, to a new and extreme degree, new styles that will be fascinating and thrilling and interesting. So in, in one sense, I'm not optimistic. In the other sense, I am. There'll just be a whole set of new styles based on a kind of elaborateness of the inputs.

Um, what do you think? I mean, one of my intuitions is just that you'll have so many more people who would've never started or even explored or thought they weren't, didn't have talent or whatever. And now that kind of like what defines talent is a different thing. I mean, there's this like, there's this like Ira Glass quote that I keep coming back to where he was talking about making a podcast and he was like, the worst part about starting off making a podcast is the reason you wanted to make a podcast in the first place was because you had good taste, but the first, you know, 1,000, 2,000 hours you do making a podcast, creating whatever it is, you're constantly sort of depressed because you're judging yourself based on your own tastes and are sort of disappointed in what you can create. I, my, my hope is that sort of the, the loop between what you can imagine and what you know is good and how long it takes you to sort of develop the skills to get there might decrease, which maybe would like bring kind of more geniuses to the fore. Is that Unreasonable, unrealistic.

It's not unreasonable, but I'm not sure. So I think back, say, to mid to late 18th century Germanic music. You have so many great composers in part because so many families have pianos and people, there's no record player, no CD, no Spotify. You sit down at the piano to amuse yourself, play for the family, and people start getting good and they're trained in music at a young age, and then some of them become composers. So that's an entirely plausible path by which people might use AI to excel in the visual arts. But there is an alternative scenario where the AI ends up being more like an iPad that you just use it to keep busy. Maybe it makes you slightly stupider. Uh, you're more entertained, but you use it in, in ways that for you remain passive. And again, there I find prediction hard. Maybe we'll get some kind of bimodal distribution. Where people go to one extreme or the other. Uh, but I think both of those scenarios are possible.

What's the most, uh, breathtaking thing that AI has done for you so far?

Well, AI is not an entirely well-defined term, right?

Sure. Whatever.

Like, uh, the thing that works so well is a kind of AI, and Google searches have done me very well as a blogger for 20 years now. And there's not any single act of Google that stands out, but the, the mere aggregate of how useful it is, uh, maybe would be the number one contribution of AI. Let's say the role of AI in medicine, you know, is not always transparent. Presumably I've benefited from that. The general role of computation in getting us the mRNA vaccine so quickly wouldn't in the, in the narrow sense count as AI, but it's AI-like in some way. So it depends what you count.

So, uh, I was playing around with ChatGPT-3 the other day and trying to kind of come up with, uh, uh, visual art for a, a podcast that I recorded about semiconductors. And I asked it for visually interesting ways to describe a semiconductor fab. And it came back with me, it came back to me with cleanroom cathedral. Um, which I then put into Google and realized no one had ever put those two words together before on the internet.

And how was the image? Good. It was okay.

But actually, you know, it's interesting because like my imagination was actually better than what Midjourney gave me. I think I was, I still needed to kind of like, uh, fudge it a little bit. Boost.

Yeah. Boost your creativity. Like, oh, I should run with that idea. So that's the positive scenario. I think we'll see a lot of that.

In ancient Greece, it wasn't easy to ask the oracle a question. You had to like hike for 2 weeks. You had to pay all this money. You could only go a few days a year. Um, uh, what does it mean that we now have, you know, on-demand oracles?

Some people will worship them as gods. Keep in mind, in the ancient Greek world, you have more local, lower status oracles that you consult, right? Even just soothsayers. So in some secular way, we'll become more religious. We'll be more obsessed with these things. Uh, some of us may fall in love with them, as in that movie Her, which is an excellent film. And I think, uh, you know, who loses influence is a question I think about. Well, do you go to your priest or rabbi less, right? Do you go to your friends less? Maybe you go to everything else less. Uh, will people be asking, or are they already asking, you know, ChatGPT, well, who is it I'm most likely to marry? Maybe you need to reword the question to get through the filters, but there's a way to jailbreak it, right? Yeah. Will I marry before I die? How, how should I find my true love? You know, is, is John Smith the love of my life? The possibilities are unending, but people are going to believe the thing in part. That's what matters.

Yeah, no, I mean, it's, it's wild how much it will guide people's decision-making, you know, what classes to take, what job to take, um, you know, where to spend a Friday night, where to spend the rest of your life. Um, uh, you mentioned religion. I want to take an antisemitism detour. Uh, what does it mean, uh, for, uh, America that we now have, uh, sort of antisemitism back in the, I don't know, somewhat normalized political bloodstream. Is this a leading indicator for worse things?

I suspect there's a pretty low ceiling on antisemitism in American political life. I don't really see it as a winner, so I'm fairly optimistic on this count. I think we also need to ask, how much did it ever go away? It is definitely more open now. Uh, I haven't done a formal count. Perhaps likely there are more acts of violence, but in terms of number of mainstream institutions switching, you know, to that side of the debate, I just don't see it. So I think the best prediction is that it will asymptotically dwindle once again. I hope.

What is What's your riff on ChatGPT and aliens?

You mean, should you ask ChatGPT if UFOs are alien vehicles? I could do that right now. I haven't tried doing that.

It might be worth something about you. You were saying something about like, uh, well, while we're waiting for the answer there, you were saying something about, uh, uh, ChatGPT making you sort of think in a new way about the Fermi paradox.

Oh, that post. I get now your question. Well, look at it this way. ChatGPT and other devices, they're smart in a way humans are not, right? So you have to up your estimate on the possibility of other kinds of intelligences being quite impressive. And ChatGPT gets smart in large part because it has so much training data. So just having access to a lot of training data in your environment, it seems a lot of animals, plants, whatever. Have that. So your estimate about how easy it is for intelligence to evolve should go up somewhat. So that means the number of intelligent aliens out there should be higher than, you know, say it was a few months ago. And, you know, correspondingly, you should think the chance that aliens or alien drone probes have visited this planet in some form whenever is somewhat higher than you used to think. All of those to me seem like valid inferences. I don't know if they're big revisions or not, but it seems those are the directions that the revisions should go in. Uh-huh.

I want to come back to AI, uh, like HER and companionship. Um, will this put therapists out of business?

A lot of people like the human therapist as a kind of placebo, even if they know it's not more effective than the AI. But I would guess this will cut the demand for therapists, I don't know, by a quarter, maybe a third. So it will matter. No, I don't think it will put therapists out of business cuz most therapists probably aren't effective anyway. And you're paying for the feeling of going to a therapist. So the notion that the AI is as good as the therapist, maybe it just doesn't really matter that much. Yeah.

But I mean, if you can have a, a voice that's like a better therapist than any therapist that's ever been. In existence because it has more data about you. It's sort of like feeding off the entire internet. And you know, in not too long we'll probably have a sort of visual, a sort of human avatar to go along with that. Isn't that close to sort of strictly dominating, sitting someone in a room who's like thinking about their own problems?

I don't think most people want a better therapist. So that's part of the issue. Like what role do therapists serve? You want it to be somewhat scarce. You don't want it to be available at your fingertips. Uh, you want to lie to it to some extent. You want it maybe to be reassuring in a particular way with just, just enough challenge injected that you don't feel it's just a placebo. Now maybe an AI could copy those features of a successful therapist, but there still seems something about the vividness of face-to-face interaction. Uh, that will make the human therapists valuable in some way, even if they're not actually more effective in fixing your problems.

Are people gonna still learn languages?

I think they will, oddly enough. Uh, there's something about wanting an intimate interaction with a culture, as you have achieved by knowing Chinese. So in some future, you could be walking around with your universal translator. And it would make every conversation twice as long. But in addition to that, it wouldn't feel the same to you. China, Chinese would not become real to you in the same way that they do, given that you can understand and speak Chinese. So I think demand to learn other languages will be more robust than people think. What I think will vanish is the demand to learn a bit of a language, you know, to order in the restaurant or ask for directions. And there the AI is just gonna do the work for you. And if that takes twice as long, no one really cares.

I mean, it seems like it's gonna very much turn into a luxury good even more so than it is today. Absolutely.

But luxury goods, as people get wealthier, remain consumed, right?

So Tyler, you've had another, a number of, uh, blog posts, uh, alluding to the fact that you're surprised people underrate the risks of great power war. Um, why do you think so?

If something has not happened for a long time, most people simply forget about it. The last time a nuclear weapon was detonated against human beings was 1945, right? It's a, a, a great deal of time ago. So we simply start assuming it can't happen, it won't happen. It's not even within the set of our consciousness. So terrorism at different periods of time becomes thought of as not really possible. Uh, not recently. Of course, it became salient again with 9/11. A major pandemic that affected everyone, HIV/AIDS, although it did affect everyone, was not perceived that way. Uh, now that's salient again. So we just underinvest in catastrophes if they haven't happened in a long time. I see this pattern in history again and again and again. So I think a major war between great powers falls into that category. And now with Russia, Ukraine, China, Taiwan, it's no longer the case. People talk about it a lot. Uh, I'm not sure how much the talk helps, but at least there's some basic awareness these things can happen.

Uh, if you, if China does invade Taiwan in the next 5 to 10 years, what do you think the odds are that the US would join the fight?

Uh, if it happened very soon, I think the odds are extremely high and I've, They asked people who ought to know, and they confirmed that answer. Now, maybe some of that response from them is a bit strategic, a kind of bluff. But I don't think it's only that. I think the United States really would be obliged to respond, or you're faced with a world where South Korea and Japan get nuclear weapons. There's quickly a war between Israel and Iran. Uh, countries in the Middle East, Turkey, Saudi, UAE want to get nuclear weapons. Our other commitments become much less credible. And, and we don't want that. So we would do something quite significant in a military way. Now, how things are 9, 10 years from now, you know, say there's an evolution of Ukraine-Russia situation. Well, could that change our views on Taiwan? Absolutely, it could. But at least anything close to the status quo, I'm pretty sure we would respond in a significant way.

So the, the sort of Trumpist isolationism rises again as the thing that changes the dynamic. You're, you're giving pretty low weight to.

I don't see the evidence for that. I don't think it's the view of the American electorate. I think American elites for the most part accept and embrace the idea of some degree of American hegemony. And no one wants to be the president presiding over a world where American global credibility is gone kaput. And just, oh, you lost Taiwan, it looks bad in the election. Wars break out other places. There's chaos in Asia. Uh, there's a problem with semiconductor chips depending on exactly how the war goes, possibly a recession or depression as a result. Politicians don't want that. Uh, they will fight back.

Um, the other sort of scenario I worry about is, uh, a victory happens fast enough that then there's a, uh, a choice that a US president will have to make between a war, which would slice 10% off global GDP, or, um, uh, sort of just shrugging your shoulders and allowing trade, um, in East Asia and the sort of Chinese, uh, Taiwanese semiconductor ecosystem to continue. And that—

I think if the war is over that quickly, you're right that it becomes a real dilemma, but I don't think it will be. As you know, Taiwan is not easy to invade, no matter how superior your numbers. Chinese military does not really have proper training, especially proper training with supply lines. Supply lines are remarkably difficult to pull off. It's one of the amazing achievements of the current and recent U.S. military. So there's going to be bumps along the way, even if you don't think the Taiwanese put up a Ukrainian-style resistance. And U.S. can deploy its military assets really very quickly. So the idea that we wake up one morning and Taiwan has fallen and then, oh my God, what do we do? We have excellent intelligence, satellite information, so I think it's very unlikely that's how it's done.

How much impact do you think the recent protests will have on Xi's governance into the medium term?

Well, we're speaking on what, December 13th, or maybe December 12th. Uh, for a very short period of time, what we're seeing is reports saying zero COVID is over. If anything, people are being pushed out into the workplace, out into the world. They're being told non-truths, shall we call them generously, that Chinese traditional medicine, you know, will save you from long COVID. And the flip is very much in the opposite direction. Now, this being China, I'm reluctant to predict that will last, but it certainly seems like a possibility that it will last, that you go from one extreme to the other. And above all, what's important to social solidarity. And so thus we'll get an overreaction. Toward openness rather than zero COVID and everything being, you know, sequestered, locked down, or closed. So it's likely those protests will have mattered a tremendous amount. Sure.

Um, I guess my question is more like beyond the specific COVID policies, um, if, if like concern about social stability, uh, changes, you know, the tightness or looseness of, or, you know, foreign relations?

Well, I would, you know, put countries or regions into two categories, those that are easy to predict and those that are hard to predict. If you look at the history of India, it seems to me relatively linear. You have some big changes, like it's colonialized, then it gets its independence, but those happen at around the same time a lot of other places are colonialized and get their independence. Growth rates are pretty steady. Uh, if you just are predicting with India more of the same for centuries, you'll do okay. You can't do that with either China or Japan for whatever reasons. If you predict more of the same, you miss the Taiping Rebellion, right? You miss the Boxer Rebellion. You miss, uh, the Communist Revolution. You miss Deng and the liberalization starting in '79, or some would say earlier. Uh, you miss what's happened recently with a move toward greater autocracy. So China is one of the hardest countries to predict. You periodically get these very major sudden flips that most observers are not predicting. So I'll just predict that China will stay hard to predict.

Hmm.

Um, or if you have riots or demonstrations in India, I'll just say we're gonna get more of the same. Like we're still gonna have some riots and demonstrations, but India will still be India. I'm not gonna say that for China.

So as a sort of corollary to that, what's the point of regional studies?

Well, look, you need smart individuals in the State Department who have credentials, the ability to command facts intelligently, the ability to be persuasive, and they are credible. If you've done regional studies on China, are you in a better position to fill those jobs? I would say absolutely. It's not just signaling. Some of it is signaling, but you're actually better able to produce the kind of image and persuasiveness that the job requires, or working as a diplomat. Now, people who do regional studies do not only work for the State Department. That's just a very simple example that helps you see that it still can be quite useful.

Um, so in the early, you know, in the first few decades of the Cold War, the Ford and Carnegie Foundations basically created American area studies. I'm curious, Tyler, as you are thinking about, you know, as, as you've been following the sort of new wave of billionaires, aside from investing in, you know, progress studies stuff, what sort of like new intellectual outgrowths will come out of new philanthropic money coming online in the coming years?

Well, I think donations will become less regular. So the old model of donations is you have a long-term relationship with the donor. They renew every year. It's quite predictable. I think younger donors, they think more in terms of venture capital, like hit and run, give something a boost to get it started and then hope it can find a way of supporting itself. Be more non-traditional, take more chances, invest more in individuals, be more skeptical of universities and nonprofits. Those are all trends I see with tech money and crypto money. Now, how much more tech and crypto money will be generated? On tech, I'm pretty optimistic. Crypto money, I guess I'm less optimistic. That will depend. Uh, but I think those are some of the trends we're already seeing in philanthropy. And I think as it relates to China, a lot of donors would like to do something, but they don't know what to do. So say somehow you want to improve American relations with China. Well, what exactly do you do? It seems like there's a lot of ways you can spend a lot of money and just get marginal impact, like make a big donation to a regional studies program at a major university. It seems not very leveraged. Um, and people are looking for something to do and they're also afraid of becoming a target of China. So we'll see, but I'm not sure that area will change so much.

How would, how would you spend money to prevent World War III?

I don't know that spending money is the way to do it. If I were the U.S. government, and of course I'm not, what I would do is go to some of the countries that we're not so friendly with and shore up, you know, boost the quality of their early warning systems so they don't fire nuclear missiles thinking by mistake that they've been attacked, which indeed happened several times during the Cold War, and we almost had exchanges of missiles. So you want to give those, you know, evil state parties better information than they probably have now. Now this is all, you know, high levels of national security. It could even be we're doing more of this than we let on. I don't think we are, but it's possible we are. But it seems to me that's some low-hanging fruit that we could be doing more of. It's not something I can do. The thing I do is just talk more about the issue and try to raise consciousness.

So recently, uh, Blizzard Activision just, uh, was, didn't renew its deal with NetEase, uh, its Chinese domestic distributor. So shortly, uh, you know, millions of Chinese nationals who've been playing World of Warcraft their entire lives will no longer be able to. Um, I'm curious, you know, how important shared cultural touchstones like, you know, video games, the NBA, and Marvel movies are. To keeping the peace?

I don't know. We had plenty such touchstones with, say, Germany before World War I, World War II. It didn't matter, but it's certainly worth trying. You know, I had my own project to improve relations with China, which failed, by the way. I wrote a manuscript for a book, and my plan was to publish it only in China. And it was a book designed to explain America to the Chinese. And make it more explicable, more understandable. So I wrote the book, I submitted it, uh, to Xinhua, which gave me a contract, even paid me in advance. But then a number of events came along, most specifically the Trump trade wars, and the book never came out. They're still sitting on it. I don't think it will ever come out. But that was my, you know, you could call it misguided project to just do a very small amount to help the two countries get along better.

Wow. What were your, uh, what were your themes?

Well, if you think of Tocqueville, he wrote Democracy in America so that Europeans would understand America better, right? I thought, well, if we're trying to explain America to Chinese people, it's a really very different set of questions, especially in the 21st century. So I covered a lot of basic differences across the economies, the polities. Why are the economies different? Why is there so little state ownership in America? Why are so many parts of America so bad at infrastructure? Why do Americans save less? How is religion different in America? That was, I think, an especially sensitive topic. And just tried to make sense of America for Chinese readers, but not defending it. Uh, just some kind of olive branch of understanding. Here's how we are. And, uh, I don't know. I don't think they'll ever put the book out. And of course, by now it's out of date.

Yeah, but there's, I mean, there's plenty of other people, other like countries on the planet who, you know, could use a little, you know, Civics 101.

They could. I mean, this is a book written for Chinese people with the contrasts and data comparisons to China.

Yeah.

So to sort of send the same book to, you know, Senegal, I don't think would really make sense.

Yeah. But if you publish it in the US, it will like, You know, Osmose out. I don't, I don't think it needs to be published by Xinhua for Chinese people to read it, Tyler.

I've thought of having it translated into Chinese, distributed Samizdat in some way. Yeah. Uh, haven't ruled that out.

Um, not a lot of downside, I don't think.

No, I, no downside for me, but, uh, you want to do things right. And I kept on waiting for Xinhua, and now I've really completely given up. The book is out of date with facts. That's not a big problem. Facts you can update, but it's very out of date with respect to tone. So right now everyone feels you need to be tough with China. You can't sort of say nice things to China about China. You're pandering. You look like LeBron James, or you're afraid to speak up. And the book would've made a lot of sense, say in 2015, uh, that its current tone doesn't make sense in the current environment, even though I still like the current tone, but it would be misread as something it's not.

Yeah. Well, I think it's, I think it's a more important book in 2023 than it was in 2015.

It probably is. So, you know, it will have some future. I'm still thinking about it, trying to get that right. By the way, this is the only time I've ever discussed this publicly.

So, uh, Breaking news here on, on the China Talk. No, but seriously, someone, someone out there who's listening to this is a translator, um, or works at a publishing house that might make a little more sense than the Chinese state, uh, you know, apparatus. So, uh, yeah, get in touch with me or Tyler. Let's, let's, uh, uh, peer pressure him into, uh, making this thing happen. Industrial policy. Should the U.S. be doing it? And if so, what are the kind of institutional, uh, bodies or know-how that it would need to not waste everyone's money?

People mean so many different things by those words. Better, I think, to be specific. I would say our industrial policy of the past has been A, the military, and B, strong governmental support for universities. Rather than a lot of explicit or even indicative planning. I think that's worked pretty well overall. If someone says, well, there's a crisis in semiconductors where just, you know, doing things to help universities is far too indirect, I strongly suspect that's true. Now, I don't know exactly how to fix that problem. You have Intel laying staff off, you have Intel not always stepping up to the plate in terms of like delivery and efficiency and quality. So simply throwing money at it, I don't know. I'm pretty convinced we shouldn't be doing nothing. But I would just say it's more useful to talk about specifics. Industrial policy per se, it doesn't really mean anything to me anymore.

So Boeing, for instance, I think, you know, feels, or Boeing or like major military, your Northrop Grummans of the world, um, when I sort of read about them and their dependence on, on U.S. government contracts and like, feel a lot like Chinese SOEs, um, uh, how worried are you that more government money towards, you know, further away from, uh, you know, labs and universities and closer to firms themselves will, uh, lose, uh, you know, help, like, lead to more firms losing their edge? Uh, and is that something that we can, should be concerned about? Or are there certain industries where scale is so large that you kind of gotta hold your nose and, and bite the bullet?

I think we're stuck at the moment. So the number of firms that can get through our military procurement process, you know, often it's no more than 2. The process itself is so cumbersome. And the companies you mentioned are pretty good at that, even if they're not always good at other things. But it does seem to me our procurement cycle is broken. A lot of procurement can take 12 years or more at a time when technology is changing so rapidly. We're in essence depending on companies to do things outside the procurement cycle, as is happening with AI, I might add. And then like jumping on board very late and commandeering it in some fashion. Maybe that too is just what we're stuck with. It's just hard for me to believe that what we have now is the best of all feasible worlds. Uh, there must be a way to improve it. Uh, I'm not the person who has the knowledge to tell you how to do that, but it has to be broken and wrong is my belief.

Uh, Tyler, what, you know, bureaucracies or organizations throughout history would you have most liked to work at or be a fly on the wall for?

Well, something like the Aztec bureaucracy before Cortés arrived. I think I would learn a great deal. Wasn't a very pleasant bureaucracy, but it must have done a lot of things very well, right? Probably underrated from an efficiency point of view, like managing all those canals. Were they all private? I'm not sure we know, but I doubt it. It's a lot of infrastructure in Tenochtitlan at the time. Almost certainly a strong public role in that. The Roman Empire, ancient Greece.

Oh wait, let's stay, let's stay on the Aztecs for a second. The Fifth Sun. I think that was a book you recommended, was the most, one of the most fascinating things I've ever read. It was this, I think anthropologist who, you know, the, the, the source material for the Aztecs is like first generation folks or like people who were talking to the Spanish.

They had codices, right? And the libraries of the Aztec Triple Alliance were burned, so we're not sure what they had. Yeah. But it was a truly remarkable civilization. When the Spanish arrived in what we now call Mexico City, they were impressed and astonished, especially by the network of canals. And if you look at what we now call biotechnology, they breed, they bred corn, a remarkable achievement from, you know, what was a weed called teosinte. That was very far from modern-day corn, and they turned it into a food that drove numerous civilizations, including a lot of European civilization. And these were just untutored in the formal sense, untutored farmers in Central Mexico, just playing around with, with plant breeding. Yeah. And they, that led to achievements that really no other civilization has come close to. Actually, including ours, as far as I can tell. So, uh, extraordinarily impressive set of people and culture. The language, Nahuatl, is beautiful. Uh, the poetry, what, what we know about them, uh, it all just seems incredible. Quality of the food.

Uh, uh, you mentioned the Roman Empire.

Well, look, you're curious, how good was it really? How far did its reach extend? Someone writing another history book is not really gonna teach me that. But if I could go back and, and work for the empire, you know, say AD 150, or, you know, pick your year, I feel I'd come away with a pretty good sense of things. How long did it actually take to get a message to the other side of the empire? How were grain deliveries managed? How binding were the price controls? And so on. I'd learn all those things. How can you not want to do that?

All right. Uh, let's, let's do the past millennium. Pick a few.

Current Singapore. You know, I've gone to Singapore many times. I love my conversations with people who work in the government of Singapore and, uh, to work there, you know, I'm not a citizen and so on, but that for me would be a great honor and I would learn a lot.

How about a US bureaucracy?

Well, something DARPA-like. Would be my pick. Now, maybe DARPA itself. As you probably know, the NIH is introducing its own version of a DARPA-like institution for healthcare. I know some people who are working with them trying to make it better. I think those would be some areas where maybe I could have some positive input. But you put me somewhere like the Fed, even if you like my views, I don't really think I could contribute to what the Fed is doing. They have an excellent, numerous staff. I don't per se know anything those people don't. I think my marginal product at the Fed would be zero.

And what about a, not, not your, less your marginal product and more your, oh, this would be really interesting. I get to learn a lot or learn something interesting.

Well, yeah, this is more related, right? What's interesting is to be able to improve or change something. So there's bureaucracies that are hopeless. That I wouldn't want to do. There's bureaucracies that are pretty good, like the Fed, where my marginal product would be zero. Uh, but basically a new institution or a new branch of an institution is where I would want to be. I gave a talk recently to ARIA in London. Of course, that's British, but that's their new science funding agency, and they have latitude to take a lot of chances. And, uh, the people at ARIA, they're all reading my book with Daniel Gross, Talent. And trying to think, well, how can we use this book to shape who it is that we hire next? So to be somewhere like that, I would be excited about.

Do you have some book recommendations about technology and bureaucracy?

You know, with books, it really depends where you are starting from. I don't know any magic quality book on technology and bureaucracy that is simply good for everyone. I like the idea of reading in clusters. Like you could pick a bunch of books surrounding the mobilization of technology at the start of World War II. That's a great cluster to read in. Or you could pick, uh, you know, history of the National Science Foundation or history of the internet or the Pentagon and information technology and just read through the cluster. I think that's better than looking for the book.

I've been searching for 2 years now for something as good as Men, Machines, and Modern Times. I don't know if you've come across that one.

That's very good. I think it, it comes down to reading in clusters. Books on technology maybe don't attract the very best writers. They attract people who love technology, which is fine. But there are really many wonderful biographies of literary figures that are captivating. Virginia Woolf, Samuel Johnson, Keynes, who in a way is a literary figure. So many. But when you get to technology, I don't see anything comparable.

So yeah, this is, this is actually, this is one of my questions, um, of the kind of lopsided ratios of book quality relative to topical importance. Um, and you know, I don't want to say that a, a marginal Shakespeare, uh, analysis book isn't important, but, um, where are others where you think there's like overrepresentation or underrepresentation of, of great books?

Well, if you take Samuel Johnson, I believe there are 3 biographies of him that are just phenomenally good, like a book you could just push on someone, not someone totally uninterested, but you can just say, this is a phenomenal book. And there's 3 for one guy. So, I mean, that's great, but it's also a little screwy. Most major tech figures, there's not one such book. And the people who could write great books on tech, like take a Patrick Collison, well, he has a high opportunity cost. Of course, it probably won't ever happen. Uh, in literature, people usually have very low opportunity costs and they're rewarded for writing about literary figures. So I think that's the pretty obvious division of where books turn out very well and where quality's underprovided. High opportunity cost. Technical subject matter that a bit repels the, what Mark Andreessen calls the word cells. And that's most of tech. Uh, also history of science is not the best area to read in. It's extremely important. There's plenty of books full of facts, but if you're looking for, say, a truly charming book on like the history of botany or geology, I'm not saying there are none, but you have to actually look pretty hard. There's a few areas like theoretical physics, Einstein, string theory that are, are enough popular interest. You have a lot of very good books by super smart people like Roger Penrose that are very appealing just to read. Evolutionary biology and Darwin, again, a lot of great books on that one thing, which have sold very well. So you, there are these few areas that everyone goes crazy over, and then the rest of science. You know, the development of the science of taxonomy. I did some reading in that area. Very important actually. Linnaeus and, and so on. How we classify animals and plants. I'm not saying those are bad books, but there's not one I would push on you.

Yeah.

I mean, there's no Richard Dawkins of taxonomy that I know of.

What are like, how do I ask this? Like the worst books for people to read. At different points in their life? Like, how, how have you seen books ruin, like, like latching onto a certain book kind of ruin or spoil someone?

I don't know. I, I'm very reluctant to make those judgments. So I certainly know people who have read books and then become involved with cults, either formal cults or informally cult-like institutions. But maybe that was the best they could have done, and maybe they end up doing something useful. Look at Alan Greenspan. He was in the Rand cult and he ends up running the Fed. Now, opinions differ on how good a job he did at the Fed, but you wouldn't say he just like blew his mind on LSD and was a cult stoner and did nothing. He was super successful. So it's very hard to know when a book is truly harming someone. Maybe people should read more bad books and fewer good books. Just the contrarian in me is inclined to wonder if that isn't the case. I wish people in tech would read more humanities. I, I mean, I definitely have that wish.

But did, I mean, did, did you say like the SPF is the, um, uh, you know, the, a, a great example for, maybe this was, this maybe this was Byrne. I think Byrne said that, you know, SPF is the great, um, uh, uh, example of like, Okay, what happens when you get a tech founder who reads a lot of books? I guess he said he didn't read books, right?

Tyler Cowen posted that he hated books and didn't read books, right? He said if you write a book, you're a fool. It should be a 6-paragraph blog post.

Um, how should people think about giving away money? Or not, not billionaires. How should like normal people with like, you know, 4 or 5 digits to spend?

Give2GiveWell is one simple answer. Uh, they put a lot of serious effort and energy into figuring out which projects should get money, but it's not just about the rational side. You want to give to things where you will feel involved in a way that motivates your own altruism. So it should have personal meaning for you. I don't think it should be a waste. It's often a waste to just give to your alma mater. But the notion that you can sit down in a perfectly rationalistic manner, figure out the best donation, that to me seems a bit off. You need to take your own imperfections into account and to give to things that will remain vivid to you. I think giving to political parties usually is a waste. Overrated. People do too much of it.

Tyler Cowen, what do you think about audio quality when listening to music?

I have invested a great deal in audio quality, so I have a very expensive stereo sitting right next to me now, and I very much enjoy listening to the stereo really every day, uh, quite a bit of time. Uh, people who just listen on a computer or who listen with earbuds, uh, to me that's aesthetically criminal. Now, if you have a high-density file and the right kind of earbuds, The sound quality might be good, but it seems wrong to me that most music was not composed to be heard by earbuds, and you're injecting it into your brain in a way that is contrary to the intentions of the music makers. And it's a very different kind of perceptual and emotional act. I think you should listen to music in matters closer to how it was intended and also go to live concert. So it seems to me with music listening, American society has become much worse over the last few decades.

Um, have you tried IEMs?

No. No.

Um, so these are, uh, wired headphones that, um, for like $100 can actually blow your mind. Um, in terms of, in terms of quality, uh, I've had very good headphones.

I don't think they were IEMs. And the sound quality is good, but ultimately it's not how I want to listen to music. It's somehow taking music out of its proper context.

Um, is there like a skill you wish you had?

Uh, reading and speaking Chinese. How about that? That would be high on my list if China truly reopens.

Okay.

That's a skill I wish I had. I prefer not speaking Russian. Actually, some of my family speaks Russian and I find it, it's better not to understand all of it.

Um, what else did we get to on China and AI that's on your mind right now?

Well, how's China doing in the areas of AI, right? That's a perennial question. I'm not sure I know the answer, but I'm surprised they didn't ask me.

No offense.

What?

That's why I didn't ask you, Tyler.

Well, I can tell you what people tell me.

Okay.

And this is just secondhand, but they say when it, in two areas, AI and surveillance and AI and management of a drone swarm, serious people have told me China is ahead. Other serious people have told me the US is ahead in virtually all other areas. Not quite an endorsement of those views. I do consider my sources to be highly credible, so it's the best view I have.

Um, my kind of base case is that winning on those, like, other things will end up swamping— well, I mean, I guess the big question for me, which no one really has an answer to, is just how much this stuff will diffuse.

I think it all is diffusing quite rapidly. Now, surveillance doesn't diffuse per se unless you do surveillance, right?

Yeah.

So, but if that doesn't diffuse, I'm quite happy for the most part, even though it may make society less safe in some ways. And China not just is more willing to do it, but has a much larger database, more people, and they record more interactions. So I think their lead there is pretty safe. Uh, drone issues do concern me. If drone swarms are a big part of the future of warfare, They could beat us, right?

Yeah, but maybe it'll be something else. I don't know. It's, it's kind of banking on one technology being the—

sorry, not attracting the highest quality immigrants is for China a big problem in the AI space and many other areas.

Anything else?

I think, you know, you asked what areas have good books. I find very good books on China hard to come by, almost impossible. That it's one country where traveling in the country is of very high value compared to reading books about it.

I think like political science is a crime on humanity in terms of the quality of books that get written by people who could have been writing better books. Um, or at least political science insofar as it's been done in the past 30 years. Um, Uh, I think sort of that, that's where I see the most like missed potential is early career folks who have done a lot of research. Um, but, uh, to get tenure just had to write things which are painful.

I agree. There's far too much writing for tenure, but it's in most areas. Not only political science, maybe more obvious in political science cuz you have a clearer, more vivid image of what the book should look like.

What percentage of, uh, sort of artistic consumption do you think is gonna be, uh, I guess what I'm thinking is like what people have on their walls 10 years from now. Are you still gonna want, you know, a print of Picasso or a, or a, you know, poster of Bob Dylan? Or is it gonna be some AI expression of something that you thought was really cool?

I suspect AI will be at or below 10% of that market, which is a lot, to be clear. I don't view that as a pessimistic take, but people want the posters for their context. That don't even mainly want the posters for the art. There's a signaling value for the poster. There's a how does this relate to other people value for the poster. And you'll have a pretty large number of nerdy types who choose AI-created art for those reasons, but it will be a clear minority of people.

How different a world will we live in if AI ends up getting dominated by a few countries and companies versus being more of a, um, distributed, uh, utility?

I think it's extremely likely that AI is dominated by a small number of countries. That's the status quo, right? US is by far number one. China is strong in some areas, but not in ways that are exported. So for now, at least, that doesn't matter. The UK has a role in kind of joint US-UK system, but it's ultimately subservient by the US system as evidenced by Alphabet buying DeepMind. And the European Union regulates AI in such a tough way. I don't think we have natural rivals at the moment, and it may assure, you know, yet another American century. While I am an American citizen and patriotic, I do actually find this worrying. I don't think it's entirely healthy.

Um, and what about from a sort of, uh, many versus few company or sort of like a, from an industrial concentration perspective?

5 years ago, I would have thought it would be highly concentrated and I can still see the abstract reasons. Well, you have to buy a lot of GPUs, whatever. But what I'm observing in practice is it spreads and democratizes very, very quickly. And I suppose that's my new expectation.

And how does that make the world a, you know, how might that make a world a different place?

Well, the old Peter Thiel line, you know, crypto is libertarian, AI is totalitarian. It's probably wrong. AI is looking more libertarian and more boosting of individual creativity. And more putting individuals on a par with institutions. Crypto is unclear, uh, but I don't really see how it's been libertarian so far, or not in a good way. Uh, maybe it helps you, you know, get your capital out of China. That's somewhat libertarian. But, uh, crypto also gives the potential for everything to be traced. So crypto forensics is a growing field in part because of The FTX collapse and crypto may lead us into central bank digital currencies in a way that are a lot like Chinese surveillance, but for everyone and all their financial business. So I'm not at all sure crypto will end up as libertarian. So I've become a lot more optimistic about the social properties of AI and their political properties.

How so?

It just seems there's more latitude to do things on your own. That it is put out there in the wild and then cultivated and copied by many people. And it's available, not literally to everyone, but look, anyone can go to Stable Diffusion, right? Yeah. Now, is that the very best thing we'll have? Maybe not. It's pretty good. And the ability to copy is just much stronger than I would've thought a few years ago, even one year ago.

What are some like Like things you would like to create that you don't have the talent for, um, or the aptitude, like, uh, like a, an opera you'd want to compose about a certain topic or like a play you'd want to write or a movie you'd want to direct that you're probably not going to, but like you just think should be out there or would be really cool to be a part of.

No, I don't want to do any of those things. What I want to do is see the entire world. And the person I envy is this guy named Paul Salopek. I have a podcast with him coming out soon, and he has a 15-year project to walk around the world. And right now he's in Western China, and he, he started in Ethiopia, sort of cradle of mankind. He will end up in Tierra del Fuego. Among other things, it's a remarkable physical feat of endurance, and he's not that young. It's also a logi— a logistical feat. And the ability to do that, that's what I wish I had. And I don't for a bunch of reasons.

Okay.

But I don't wish like, oh, I could make a movie like Spielberg or song like the Beatles. I, I, I just want to consume it. So if I have a long life, I can consume more of it. Like, that's my other wish.

Gotcha. Um, well, I just want to thank you, Tyler. Uh, I mean, I did the intro a little bit, but you've had a really enormous impact on my life and a number of, you know, hundreds, thousands, probably tens of thousands of other people who've sort of looked up to, learned from you, benefited from your kind of encouragement and random email responses over the years. So, yeah, I hope you got Plenty more decades in you to, um, uh, to see lots of corners of the world and continue, um, uh, helping people on their journeys.

Thank you for all the kind words. It's been an honor to be on your podcast and I'm very excited by what you are up to.

Okay.

All right.

Um, we end every episode with a song. Do you have a—

Strawberry Fields Forever.

If it's like a top 200 song of all time, it might be copyright stricken. So we gotta be a little—

copyright, copyright. Uh, I don't know, a, a Beethoven piano piece. Can we do that? The performance will be copyrighted. Are you gonna sing it or like, how do we do this song? Oh, um, does the AI do it for us? Play some Beethoven. It'll do it.

All right. I'll, I'll, I'll make, I'll make that happen just for you.

Okay, great. See you around.

Beethoven's 10th Symphony is unfinished, but there were enough sketches for someone in 2021 to feed them into an AI model and come up with a piece of work. The following is a recording of what that model created, played in 2021 by the Orchestra of Bonn. The following.