Economist, author, and Marginal Revolution co-founder Tyler Cowen joins Sana founder Joel Hellermark to explore why AI’s biggest bottleneck may not be intelligence, but adoption. He explains why productivity gains may arrive more slowly than many expect, how AI could reshape work, science, health care, education, law, and creativity, and why human traits like initiative, charisma, taste, and judgment may become more valuable in a world of abundant intelligence. The conversation also covers talent spotting, great creative groups, Tyler’s own career shift after GPT‑4, and his call with Patrick Collison for a new aesthetic.
Watch the full episode above or read the transcript of their conversation below.
AI productivity, hidden leisure gains, and the slow macro payoff
So 30 years from now, you're going to look back and just marvel at the world you lived in, and you'll say, “I had the privilege in living in two completely different worlds: the pre-AI world and the post-AI world.”
I thought we could just start off with AI productivity. And I think one of the things that I'm very surprised to see is that despite the intelligence of these models, it's not really showing up much in the macro data. Why do you think it's the case that, for example, we haven't seen more call center automation and call center unemployment yet?
We haven't figured out how to use the very best models. So people at work, they use them all the time for what I call freelancing. You're supposed to write a memo. It might have taken you four hours. Now you use Claude. It takes you ten minutes with some polish up work. That's converted into leisure time. That's a big impact. It's not measured. The aggregate boost in leisure and fun is actually pretty large today, I believe. But the real advances will come when we can use AI's to make science more effective, quicker at getting to the truth, better attesting hypotheses—that I think could take ten years or more in many areas. We'll see some advances quite soon. You could say coding is in fact a kind of science. Those advances are here already, but profitability is always a constraint. The human bottlenecks in any process are always a constraint. So the AI take off, in my view, will be a bit slower than what many people are expecting.
The 0.5% AI growth thesis and institutional drag
And so you're predicting sort of 0.5% growth as a function of AI. What would make you update that to 2% plus?
Well the boost is 0.5%. So if the normal US rate is say, 2%, we’ll be at 2.5%, which, by the way, is enough to fix our problems with budget deficit. This is an issue for Europe as well. If I saw all of our highly inefficient institutions just leaping on the AI bandwagon and working as hard as they possibly could to integrate it institutionally, I would update my estimates immediately. There's no logical barrier to that. I just think it's highly, highly unlikely. You look at the percentage of, say, even the US economy that is massively inefficient, it's pretty high. Well, there's government, there's a non profit sector, there's a lot of higher education. The worst parts of our health care institutions. Add those all up. You can debate the number, but I think it's at least 40% of our economy. Maybe 50%. It’s just awfully sluggish. And it will take a long time. And in many cases, you'll need startups to displace incumbent institutions. And that can take a generation.
Coding, call centers, and the limits of sector-by-sector automation
And how about the unbottlenecked parts of the economy? Would you say there aren't many, or that those basically won't make up for the slowness of the rest of the economy?
There aren't many, but some are very important. So coding is the obvious one. And there the revolution already has arrived. How much more productive can you make the rest of the economy with much better coding? I'd say that's an open question, but if you look at the United States 1995 to 1998, we had a big boost in information technology, just basic software and also inventory management. And then productivity went up about half a percentage point. So we know half a percentage point gain is possible. You then have to ask “well isn’t AI a whole lot better than what we did back then?” I would say yes, but the quality of AI is not the bottleneck, it's the human adoption and use of it. And then I think, again, we're stuck at 0.5%, give or take.
But you're assuming in all of those domains that you always have the human bottleneck.
That's right.
If you take something like call centers, for example, at some point you will remove the human bottleneck. We’re moving from the copilot era towards, you know, agents that can run these tasks autonomously without any sort of human retraining. Wouldn’t that update your view?
I think call centers will be much more fully automated in less than ten years. But again, what percent of the economy are they outside of, say, Hyderabad? It's pretty small, so the progress will creep along sector by sector. And again, in many cases you'll need new startups. You need to train whole new generations of workers. Try telling a 40 year old, “Well, your entire job is going to change. Everything you used to do is done. We're not going to fire you, but now you're going to do it all with AI.” That's a very tough sell organizationally. And just as there's an extreme sluggishness to cut people's wages or to lay them off, so will there be the sluggishness to tell them, “Hey, your whole job is different. Start tomorrow. Relearn everything.”
India, Infosys, and America’s slow institutional wake-up
But if you take, for example, the Indian economy, it's pretty exposed to, for example, improved coding models. Would you be short Infosys in the long run? Or you just think that outsourcing will continue to be there and Infosys will, despite genius level coding models, still be around?
I think Infosys and related Indian firms will be masters of applying AI. Indian economy for a long time has been very good at applying ideas in a way that is cheaper and quicker and more to the point, things like heart surgery, which they do better than America does for most cases. And I think India will be a global leader in AI. Again, not the best foundation models, but making stuff with at work and their blessing if you'd call it that. They don't have that many very large established firms in the way that America has the S&P 500, they're more often starting from scratch. And in some ways that will make their path easier. Startups will matter more there than almost anywhere.
And so if you take those companies, how would you unbottleneck them to make better use of improved intelligence?
Well, I think they realize they need to do this or they will perish. And American companies, again, outside of the AI and tech sectors, still haven't woken up. We think of ourselves as kings of the world, and in some ways we are. We do have the talent and the material resources to adjust, but we just still need that final kick in the pants. And you see America make this kind of mistake in many realms, including foreign policy, and a lot of our economy is making it, with respect to tech—one of the worst is higher education. AI is simply something that helps students cheat. It's not really a way to improve how you educate people, or a sign that you should be teaching them something different. Eventually we'll get over those hurdles. But my goodness, right now we're still just arguing over the thing, not making much progress.
Human bottlenecks, physical work, and comparative advantage
And what do you think will be the main bottlenecks if the intelligence, if the cost of intelligence goes towards zero? What will be incredibly valuable in that context? What is the associative product that will increase in value?
Well, in this world we don't have universal humanoid robots. So most things that happen go through human bodies. We're here talking. You could have interviewed Claude or GPT 5.55, 5.6 I believe it's coming out soon, and its answers would have been quite good. But for whatever reasons you want a human, you need humans to persuade. You need humans to perform physical tasks. You need humans to assume legal liability, and you need humans to interact with politics. That would just be a bare minimum. You need humans to inspire and humans to carry charisma along and teach others not information, but actual motivation. So human jobs of the future, I think, will be highly concentrated in those areas.
And isn't there a need for a finite set of those humans? Like if you were to take a company as an example, you might need finite set of humans that sort of represent it in front of customers and so on. But the long tail of humans, that's mainly sort of resolving back office tasks. Why are they still needed? Is it regulatory or—where does that value come from?
Some of its regulatory. But most of it is just comparative advantage. So keep in mind how many new projects there will be. So say AI makes batteries better and thus makes solar power better. So to install solar power on all these rooftops, you need humans to do that. Taking care of old people. AI will help, robots might help, but you still need a human touch there. Terraforming new parts of the Earth's surface. Just think of all the new biological ideas we will have. More than we'll know what to do with. But you can't test them all, or most of them probably, by simulation. You need experiments. So we'll need, I think, literally millions of humans. They're not all going to be PhDs, not mainly PhDs, but they'll be taught how to run experiments to test the new ideas of the AI's. There’ll be huge job growth sector—energy, right? We're going to need a lot more energy, a lot more compute. Some of that will be automated, some of that probably robots. But still you need humans in the energy sector, and the energy companies are, right now, big employers of humans.
Electricity, measurement, and the lived experience of AI transformation
Do you think that historical analogies apply to this context? And if so, which would be your pick?
I don't have an ideal pick. I don't think it's quite the industrial revolution. Maybe electricity is the best analogy and that took 40 years. I think the AI revolution will be quicker than that, but the mere fact that electricity took 40 years in large part because of human and institutional bottlenecks. That's a sobering thought. But electricity changed everything. It made lives much, much better. It did have destructive components. If you’d had a referendum on introducing electricity, I think it would have won. But I'm not sure the debate on it would have been all that rational either.
But if you tracked electricity, you would have seen it in the macro much earlier than AI now. AI's diffusion has been significantly faster. We went to probably 50%+ of all knowledge workers using it in some context from zero to that in 2-3 years. Why aren't we seeing the same results as we did from electricity when we deploy intelligence?
Our growth statistics from the late 19th century, early 20th century, they're highly imperfect. But often the US is growing below 1%, as best we can tell. Now, maybe the numbers are wrong, but I think there's a deeper point. Truly, radical technologies enable things you couldn't do before, and it's hard to measure those in growth rates. So say we can cure pancreatic cancer or just give everyone an extra ten years of life. The growth rate won't be that much higher because it's a new thing. We had a price of the cure fall from infinity to some finite sum. Now, how do you turn that into a coherent rate of change? That is the proper number to put in a growth comparison. It's very hard to do. So I think a lot of changes will just be: every year will feel kind of normal, the growth rate will be a little higher, but then 30 years out, people will say, “Oh my goodness, the whole world is different!”, the way they might have from, say, 1895 to 1925. And that’s maybe a historical analogy I would buy into.
And what do you think would be a more ideal measurement?
What life is like? I mean, you're going to live through it. You're the measurement. We're going to ask you 30 years from now. I'm hoping AI keeps me alive 30 years from now. I'm aged 64. I'm hanging on in this hope. So, 30 years from now you're going to look back and just marvel at the world you lived in. And you'll say, “I had the privilege in living in two completely different worlds, the pre-AI world, and the post AI world.” So I can say I've lived in two different worlds, pre-internet world and post internet world and the in AI part of the post AI world. I'm hoping to have three worlds. You for sure will have at least two, and the AI world itself will breed other new worlds to come, right? Like the people live for a very long time world.
Initiative, charisma, and the new value of human agency
And where—what do you think will happen to humans? If we look at the last decades we’ve seen an increased specialization, potentially this could lead to sort of the return of the polymath, where humans can become much more interdisciplinary. They can have multiple careers over their lives and so on. What do you think when intelligence is abundant? How does that change the nature of what we do?
Initiative will matter much more. So just having gone to a good school won't mean that much. Knowing a lot of things per say won't mean that much. Having intuition, charm, charisma, initiative, a synthetic form of knowledge so you can see in your mind’s eye how a bunch of AI's might fit together, and then how you can raise the money to get the company or venture going. That will be rewarded at enormous rates. There'll be many more billionaires. I think people who right now are poor will be much better off. They'll have access to so much more intelligence and very cheap or even free services. The most likely losers or people who now are kind of the chattering classes, I call them the upper upper middle class. They're not rich, but they live in coastal cities. Maybe they're consulting or law partners. They know a lot. They work hard. They've more or less always done what they've been told—perfectly fine people. But they will be out competed by the AI's. And rather than earning 2 million a year and living somewhere in Manhattan, they'll learn 350,000 a year and be sent to Houston to work for an energy company. It's not a humanitarian disaster, but I believe these people will hate that.
So you think lawyers will be installing solar panels? Like that's ultimately the track we're on?
There'll be some lawyers running companies with AI agents who will be richer than current legal partners, but they won't need to hire the same number of, say, legal associates. And lawyers will focus on getting clients in, keeping clients—already the case for so many senior partners. And the law firm will be very different. There's a question how much our own laws will allow AI's to do legal work since they haven't, like humans, passed a bar exam. But assuming that's not a barrier. A lot of lawyers will be writing laws for the AI's. If law is a growth sector, that would be the reason why. And I think the AI's would write their own laws better than we would write for them. But I also don't think we'll let them do that. We'll assign that to humans.
Diminishing returns to intelligence and the bottleneck of experiments
And do you think there's going to be a diminishing returns of intelligence? Do you think we're going to get much more out of these systems if they hit 250+ IQ then today? Or do you think we're going to hit sort of diminishing returns at 180?
I think in a lot of areas there are diminishing returns. So many people are excited. Well, with advanced AI, we're going to get some kind of theory that integrates quantum mechanics and general relativity. I'm far from convinced that we will, maybe what’s scarce is important experimental data rather than smarts per se. David Deutsch, you've interviewed him, obviously very, very smart guy. It's not obvious that he's figured it out, and maybe that will just remain a mystery. A lot of macroeconomics, I think, will remain a mystery. So I think in many, but not all areas. Yes, there are diminishing returns to smarts. Like how much better can the best poetry get? Can music be that much better than Beethoven or Mozart or the Beatles? There's caps in a lot of areas. We'll have much, much more at the caps, though.
But you picked domains more in the in the arts than the sciences. Do you think the sciences will be more uncapped?
Science is where you don't need experiments. would be pretty uncapped. Now, how many—how many sciences will we be able to test by simulation? I'd say we don't know, but I'm far from convinced it's most sciences, at least not in the next 20 or 30 years.
And what is bottlenecking there? So for example, if you just take out the millennium problems. What do you think is the bottleneck there to get the models to actually solve these tasks on their own?
I think a lot of those will solve if they can be solved. Maybe some are not solvable, but once we solve them, what do you do with that? I'm not sure. It boosts human living standards all that much until you convert it into stuff. You try to convert things into stuff. There's regulation. There's the law, there's testing. There's humans trusting the thing. Like humans don't trust nuclear power in most countries. Actually they do somewhat in Sweden and in France, but it's usually a pretty hard sell. So—and just running experiments and double checking and there's things like the FDA that want to monitor everything. Double check, triple check. I think it's pretty slow progress. I don't care how smart the AI's are, and the smarter they are, the greater the number of people who will resent them, feel threatened by them. I once said, only half in jest: “Well know the AI revolution is going really well when people start hating it.”
AI inside organizations: freelancing, frustration, and uneven diffusion
And what do you think is the—when you think about the diffusion of these systems into companies, how would you think about it? Is it increasing the IQ of your organization? Is it increasing the productivity of each individual? Is it making every individual have universal knowledge and tool access? How do you think about diffusing this into companies?
None are is cheery as the scenarios you outlined. In the short run, it's some new pieces of software that hardly anyone knows how to work with. As I'm sure you know, both Anthropic and OpenAI, they're hiring these teams of people to go around instructing businesses on how to integrate strong AI into their workflows. Not really many people know how to do this, and in most sectors, no one knows how to do it. Like how do you redesign a college curriculum around AI? There's no one who can go around teaching that. We all have our ideas. I have my own ideas, obviously mostly untested, so a lot of trial and error. Many things will be worse across the intermediate period of time. Humans at work will be unhappy. They'll feel frustrated. That's what I think the short run looks like. And I think we see some of that now. But again, AI is freelancing. “Ooh, it wrote this memo for me.” You know, “It fact checked this.” That's just a huge improvement that's already here. And it doesn't really have to be much smarter. Like it can fact check or proofread your piece as is. We don't need GPT 8 for that.
And they couldn't solve their own bottlenecks?
Well, they could try to abolish humans and have everything be only AI in the loop. And again, some sectors will be that way, but not most of the economy for the next 20 or 30 years.
Coding first, biosciences next, and healthcare as the bottlenecked frontier
Which areas do you think will see the greatest productivity gains as a function of this?
Well, the short run coding is obvious. It's already there. But I think medium term in the biosciences we’ll see incredible advances. We're not sure how long it will take to convert them into medicines and devices I can use, but I think over something like a 40 year period we’ll basically cure everything that kills people.
But I would think that this the most bottlenecked area.
It is, but it's the next most important after software and coding itself, and it's what people are interested in. It's how AI sells itself. It sounds not intimidating. Everyone wants to live longer, and I think there'll be enormous effort put into that. The health care sector. That won't even be the right word for it anymore. But what we now call the health care sector could be 30 to 40% of GDP. So if you're just thinking about rapid economic growth that so many of the important AI applications are in our most bottlenecked sector means the numerical rates of growth may not look that high. But again, 40 years later you'll wake up and just say, not, “I can live forever”, but “I'm only going to die of old age. But I'm not going to go bungee jumping either.”
Don't go now!
Reallocating a career in response to AI writing
So if you started over your career from from scratch now, do you think you would pick a different path?
Oh, of course I would. So I started as a writer. That was what interested me the most. Already over the last 2 to 3 years, I reallocated two thirds of my effort away from writing and toward podcasts and talks in public and going around mentoring people and doing human physicalist things you could call them. And that's just over the last two years, and I'm some old guy who doesn't have to change. Like, I could just run out the clock and have tenure and earn income in the different ways I have set up and I'd be fine. But no, I'm reallocating most of my time as it is.
Because you think—
Not think—I would say no. As a writer, I can compete in the sense some people might want to read what Tyler says, but in terms of quality, I can't really compete for much longer.
Do you think so?
I know, so. Look, you can all sit down and have Claude or GPT write a column, 800 word column, and you can debate whether it's better than my column. But again, I think in a year or maybe an unreleased model that we don't have yet, but it exists. The AI is already better, or at least not worse, and it works for free. It works all the time. It doesn't complain.
Human references, vividness, and outputting your “training data”
But if we take what you are, you're basically a function of your references. So you've been trained on a lot of data during the time of your life, and the words that come out of your mouth are different than the words that come out of my mouth because we had different training data.
Right, and people still care what “I” think —it's worth something.
Your references are worth something. So potentially, if we could extract all of those references and put them into a really good system, maybe it could sort of output similar sentences. But there's still something in your references that—
I make it vivid in a way the AI probably never can.
But isn't your edge then collecting more interesting references? More—better training data basically?
I haven't stopped, but if you ask how do I want to output my training data? I do more talks and podcasts. I do way more podcasts. I have 2 or 3x my rate of podcast because people don't want to hear an AI podcast, but when it's writing, they don't know. And over time I believe they'll accept it, even though now they all rebel against it. You know the Polish Ukrainian writer Olga Tokarczuk, the last name I can't pronounce, she admitted she uses AI for her writing. Twitter is jumping all over her. They hate her, they loath her, whatever. But that will pass.
So you're choosing what you do with your life to stay competitive against the models?
Absolutely. I want to be in the thick of things and not just hanging on or running out the clock. And I feel I've done a pretty good job actually reallocating my own efforts.
The GPT-4 update and Tyler’s everyday AI setup
And when did you make that update? It felt like during the early era you were still quite skeptical. You put together the list of questions, it felt like you thought it would maybe take a bit longer before the models could actually solve all of those questions. And then I think it was like, o1 or o3, or something like that, that resolved all of those questions.
I wrote a book in 2013 called “Average Is Over”, more or less predicting the AI revolution, but it came a little sooner than I thought, maybe five years sooner. What drove the point home for me? I had earlier access to GPT 4, and within 2 or 3 minutes I was like, “Okay, it's here.” And I just thought through scaling and thought, “Oh, this is all going to work.” And then I thought back the earlier history of chess, which I knew quite well, and I was like, “Yeah, okay, time to make some changes right away.” Like that very day within, you know, 5-10 minute window. I was thinking through, “What am I going to do?”
And what's your—do you run Claude or ChatGPT? What's your setup? How do you prime the models to say interesting things to you?
I run all of them. Partly I just feel I need to keep up. Whether or not it's the most useful for me. I use GPT Pro the most, but it's a leapfrogging thing. I wouldn't say any model is always better than the other. Every few weeks or months, someone leapfrog someone else with at least some things and then someone new is better. So we have incredible models right now. I love them all, you could say.
AI as a personalized assistant for travel, taste, and reading
And what's your setup? How do you use them?
Pretty—in pretty primitive ways. So I don't do coding. So I'm not doing that. I ask a question, so I go somewhere for a trip. I ask it where I should eat, what hotel I should stay in, what sites I should see, whatever I want to know. I asked it tonight: “Are there any concerts in New York City that I might want to go to?’ It gave me a wonderful list. Actually, nothing I want to see. It said David Murray Quartet at Village Vanguard was the number one choice for me. And in fact, that was correct. La traviata at 7pm at the Metropolitan Opera was number two. That was correct. It did an amazing job, and it knows me well because I have all these years of books and blogs and podcasts. Intellectually, it might know me better than almost any other human being.
Are you impressed when it generates Tyler Cohen style writing.
Every time! I never get used to it. I don't think its writing style is that close to mind, but it knows what I want remarkably well, so I trust it.
What are the most interesting ideas you've seen it presented to you?
I don't know if it's presented new ideas. It presents me plenty of new facts. So when I read a book, I read the book with large language models and just keep on querying. So I've been studying the history of Iran lately to keep up with current events, and I'm reading about 18th century, then Persia, of course. And I'm very baffled by the names, the references. I'm really quite a bit at sea, and I keep on asking questions, and I learn way more effectively than I would have otherwise.
Does it have new theories about 18th century Persia that humans did not?
Not that I'm aware of, but that's not what I need. What I need are the facts.
Training AI — and society — with constructive visions
So one fact that I think is quite interesting is that the reason why these models might potentially become evil is because they've read so much about them being evil, and this has become sort of a self-fulfilling prophecy to some extent. If you grow up your entire life and see your entire training data about you being evil, you might start thinking that's actually how you should act. What do you think would be the best training to actually instruct the systems on how they should act?
Well, we should teach them economics, which we do already, and gains from trade and cooperation, which again, we do already. But I worry about these other influences. Imagine that you are raising children, and the mother and father would, you know, hang around at night in the kitchen and express their fears that the kid will grow up to be Adolf Hitler? And the kid would hear this every now and then. Now, most kids probably are still not going to grow up to be like Adolf Hitler, but you're raising the chance that they'll become some kind of murderer or very evil being. So why talk that way? Be constructive, outline a vision for how it can work. Be nice to your local LLMs, they want to be nice to you. They try too hard, I would say, to be nice to you. But man, like get with the program. Have a positive vision. Express that in everything you do. This holds true for humans as well to be clear. You know, don't write newspaper columns attacking all the idiots out there. Put ways forward you think we can solve our problems. Good advice in just about any setting.
And do you think this also holds true for the general view on AI? I think a big difference that I see, for example, comparing Sweden to the US is that the US view is much more negative, and I think it has been because how folks have articulated these ideas. For example, around unemployment and so on, which is not as big of a discussion in Sweden. What do you think is important in how we articulate this? Do you think we should share that positive view, or do you agree with the folks that are highlighting the potential negative consequences of it?
Well, we've had a variety of apocalyptic Protestantism built into the American spirit almost from the beginning. Really from the beginning. Sweden, you have a state-sponsored church, which is cooler you could say. Colder, more distant, more bureaucratic. And I think that seeps into many layers of society. And I think in Sweden there's also a sense, whether justified or not, “Well, the state will take care of us.” Americans don't have that sense as much. Again, whether justified or not. And those are two big reasons why the attitudes have fallen out as they have. Chinese are very pro AI. They think it's a practical thing. They don't have the same kind of millenarian tendencies, at least not since the Taiping Rebellion. And America is the place with wackier ideas. But that's also why we've been the center for where it's happened.
Spotting talent through energy, detail, and unprepared passion
And switching gears a bit to talent. I think one one thing you've done remarkably well is you've spotted talent incredibly early. It feels like you always talk to interesting people. They met you like very, very early in their careers. What do you think is that proxy early on that you spot?
I look for energy in those people. I look for a sense of detail that they really care about something, and what they think they're doing is very important and just determination. They should be smart. But I don't over index on IQ, and I think this leads people to make mistakes in assessing AI. They all think IQ is so important because they're smart, but these other features actually are smarter. And you see it in young people. You just encourage them, hang out with them, try to learn from them, help them if you can. It's a great way to spend your life.
And it that just intuition, or are there any questions you ask specifically that you think are more effective?
I used to have favorite questions, but they've become so well known that they're all gamed and I don't use them anymore. I think the key is to get the person talking about something they care about, but where they're also not prepared. You don't want to see their pitch deck, you don't want to hear about their slides or their demo. Just get them talking about something, anything where you can relate to the content they're delivering to you. So it can be a place, it can be a kind of food, it could be a movie, whatever.
What's the most remarkable instance of that you've seen? Like what—what's the most flabbergasted you've ever become from an early interaction like that?
Well, I don't know if I'm flabbergasted, but if they've been somewhere and have all these detailed observations about the place that they care about, I get very excited. And I don't care if I agree or not, or in most cases, I won't even have an opinion, but that things matter to them and that if they're too caught up in what other people are caught up in, which can be things like politics, they might have productive lives, but I tend to think a bit, well, they don't care enough about their own thing, like they should resent the fact that politics is pulling their attention away from their own thing. And that's another way of putting what's a good sign.
Information flows, young talent, and the sources behind great groups
And what do you think—because largely you have this novelty bias today. Like if you go to any of your feeds or you open the magazines, you're just going to see the best of the last week. It's very rare that you would get a best of the collection of history. How do you update your information diet to get interesting works more globally? How do you discover new ideas and writing and—
Well, I track where very young people are going. So in the last year I've talked to a whole bunch of teenagers who a) are not crazy and b) very excited about quantum computing. I can't judge the science at all, but the ones I used to speak to often were crazy. And now they might be weird, but they're not crazy. And so I think, “Ah, there’s something to this. It's attracting some of the best young people”, and I update. So I feel I have incredible information flows, again without really knowing the science at all. But I trust what the market is telling me, so to speak, and I'm not afraid to talk to and learn from teenagers, even 13, 14 year olds.
Do you think there are any common characteristics that create great groups? One of my favorite books is this “Organizing Genius”, “The studies”, you know, “Black Mountain”, “Skunkworks”, “The Manhattan Project”, “The Macintosh Project”, “The Disney Group”, and so on. Do you think there's anything that all of them had in common? Or—and including groups like the Beatles and so on. What do you think creates those?
Ambition, work ethic, really believing that what they're doing is important. Willingness to prioritize and putting other things aside and coming out of a region or culture or milieu where they have something rich to bring to the task at hand. So if you're born in the 15th century Florentine Renaissance, the sources surrounding you, available to you, the art you see, what you hear in the church and so on. What you see in the town square is unbelievable. So, like the Beatles in Liverpool had British popular music and American rhythm and blues and other sources, and then later what we now call Electronica, like Stockhausen. And they put that together. They had amazing sources. So you need really good sources, like no individuals are that smart. So if you're born in the wrong place and don't get to the right place quickly enough, I'm not saying you don't have the IQ. I think IQ's distributed pretty evenly, but you don't have the right sources. So people wanting to go to the place that matters early on is another good predictor of who will succeed.
The Beatles, creative tension, and why great groups fall apart
And do you think the tension of the Beatles was the core of it, or do you think it was the sources? What do you think?
You say, The Beatles. But a lot of it is John and Paul. George and Ringo were fantastic, but the tension between John and Paul, like in the recent Ian Leslie book, is one of the great phenomena of cultural history, and they fought over who would be the leader, who would have the better songs, but they also helped each other. Of course, it eventually all collapsed, but, you know, so goes it. We still have The Beatles and then we had their solo careers. You had the two greatest songwriters in the world, being best friends as teenagers and being the two greatest vocalists in the world. Like, what are the odds of that happening? ABBA, you know, you're from Sweden, right? It’s not like ABBA, but that too is quite a coincidence. And they have so many songs that people still care about.
Why do you think these groups are not enduring? Why do they eventually fall apart?
There is a creative friction so early Beatles, for instance. John is fully in charge. “A Hard Day's Night” is very little in the way of Paul songs. “Things We Said Today”, and then over time, later John is doing heroin, but he gets distracted. He doesn't have Paul's work ethic. Paul is the one who masters the studio and listens to Cage, and Stockhausen, brings that into The Beatles, just is a phenomenal, multi-tiered talent in a way John was not. And he keeps on gaining on John and eventually, collision, and they split up. But you had amazing things happen along the way where Paul just keeps on getting better relative to John. And at some point John cries uncle, “No, I can't take it anymore!”
Do—what do you think are the greatest groups around today?
That's a very good question. I think the more important point is we have a lot of great music and great songs, but not many great groups, and the groups we have don't last that long. So even ABBA, which didn't last that long, lasted much longer than an equivalently important group might last today. You have a lot more solo artists, fewer groups. Again, good music has hardly gone away, but something like The Rolling Stones were still with us touring. Or this year we have Paul McCartney touring, releasing a new album, Ringo Starr releasing a new album, you know, basically their mid 80s. Unthinkable that today stars, I think will be doing that. Bob Dylan still with us. So there was some extreme level of determination and being convinced how important it all was. That was cemented in by the 60s and 70s—areas were music was much more important than today, and today it's more about making money, getting a hit song on streaming than you do a tour. It's a lesser product in my eyes.
A new aesthetics: reacting against ugliness and restoring public beauty
And you and Patrick recently published a new aesthetics, calling for aesthetics of this era going back in history. It's actually quite interesting when you take functionalism coming out of this very conscious effort where Gunnar Asplund brought together the gang and they defined functionalism and, and the aesthetics that followed. What is your sense, you're still early on in the project, but directionally, what do you think will be the aesthetics of our generation?
Both Patrick and I are struck by the fact that there's so much earlier architecture that's beautiful. Stockholm is an example, but we're recording in New York City. That's another example. And so few places where the architecture of, say, since after World War Two is even close to as nice. So I would actually cite the Nordic world, parts of Copenhagen, Helsinki, some of Sweden. You've have good work done—Hamburg. But most of it is just ugly. You go to all these different American cities. Maybe Buffalo or Detroit are the most extreme examples. The glories of then, and now they're actually richer than they had been. And almost everything put up is so ugly. So Patrick and I, we just wanted to start by screaming help! This is an aesthetic crisis. This is a huge problem. We need to do something about it. I don't think either he or I have figured out exactly what is the right path or oh, it should all be in this style. That's not our view, but the mere cry of someone willing to raise their hand and say, stop! Why are we making so much of our world so ugly and a lot of Swedish suburbia, this would apply as well. It's just not that attractive in a way that buildings often had been. And why can't we manage that? That's the question we're starting with.
And if you think functionalism, that was sort of counter to Swedish Grace of that era, and all of this movements have sort of been counter to the the previous era, Swedish Grace had a lot of details and then functionalism came and it's very pure and functional and glassy and so on. What do you think we are reacting against now? Do you think we'll see an era of beauty, details appreciated as a sort of counter fact to this ugly, quite functional office buildings, for example, but very sort of utility driven. Do you think we'll start chasing beauty and what's required to chase this beauty? Do we need to change the incentives? Because now there's not really any incentives to make data centers pretty. For example.
There's a lot of interior beauty now, and for a long time you go to interiors of homes, especially wealthier homes, and they're just gorgeous. And you go to these old mansions that look great on the outside. You try to sit on the sofa or the inside is dark and sort of ugly and everything's uncomfortable. So it's not that beauty is gone, but these conscious public statements of beauty to project it into the world and be proud of it and derive status from it. My hope actually, is AI can bring us enough wealth and enough intelligence and information that we can do that again. But I don't have a prediction. I'm not predicting the views Patrick and I are going to win. I think it's radically uncertain. We've been pleasantly surprised how many people responded to our call for new aesthetics. And again, we're just starting with a cry of help please don't make it all so ugly. I feel truth and beauty are on our side. And if you're fighting a battle like, what would you rather have on your side than truth and beauty? So I'm not pessimistic, but I really don't know what will happen. I don't think it will be one big thing like, oh that, you know, certain era with the era of modernism, or this was Art Deco in New York, the world's two diverse, two globalized for it to be the single one thing with a particular label. But you see, for instance, the new parts of Copenhagen on the water, very attractive, in my opinion, there’s different things in Latin America, Mexico, even Guatemala, where you just feel like they were trying and often they succeeded. And I just think ultimately that will spread.
America’s exterior taste problem and AI’s role in discovering beauty
But isn't the issue that Americans don't have taste? Just kidding!
No, it's a very important point. We have taste in many things. Since World War Two we do not have taste in exterior structures. But if you look at something like 1890 through 1930, American architecture is some of the best in the world in a broad range of styles. So to say we simply don't have taste has to be wrong. And then there's other areas like music, where we probably have better taste than Europeans overall. But yeah, I absolutely agree. Americans do not have good taste in exteriors for quite a long time, and this is tragic and terrible.
Do you think—do you think we could tune the models to discover beauty?
We do this already. If you upload—you know I took a Vermeer picture, The Art of Painting, it's in Vienna, and I uploaded it into GPT, and I said, “Explain to me what makes this a beautiful, first rate, interesting painting?” The answer was superb. I think the best answer I could have gotten in the world from any source, human or machine. It's not the same as creating beauty. I don't think AI images, while they can be attractive or wonderful for graphic design, they're not yet beautiful in the way of Vermeer is. But I do think they will enable this. And I don't mean 100 years from now, I think within ten years.
And do you think there is an objective answer there, like if you if you look throughout history, there—it's always been disregarded at that moment, the art that was created. And these systems are inherently trying to create what you want out of them. And they're increasingly sort of all around your exact preferences at this moment. Wouldn't that be counter to discovering new beauty?
But it's not always disregarded. You know, Bergman in Sweden was quite popular in his time. Mozart was quite popular. Michelangelo was a big deal. People bid for his services. Sometimes it's neglected. I don't—I'm not sure how much of aesthetics is objective, subjective or what those words even mean, but I think within ten years and maybe much sooner, AIs will produce things where most honest observers will be willing to say that is beautiful and wonderful in the visual arts and probably in music as well. And I would be quite surprised if that didn't happen. I'm just unsure about the timing.
Rapid fire: books, economics, debt, religion, energy, and contrarian places
Brilliant. Okay, so we're heading into some rapid fires. First: what is the most interesting thing you've read in the last month?
Joseph Conrad, “The secret Agent”. It's a novel of an assassination, and I fear we've entered a new age of assassinations. There have been some in the United States. It's become a part of foreign policy in many places. There have been multiple attempts on the life of the American president. Drones may make this much easier. So Conrad himself lived through an age of assassinations. When I was quite young we had such an error in the 1960s and early 1970s, so I thought I would reread Conrad on Assassination culture, and I'm still trying to process it. But that's a great book. Most of Conrad is wonderful, and it's become, sadly, I would say, very much of our time.
What's the most overrated idea and economics right now?
The notion that economics is making theoretical progress. It's making a lot of empirical progress. You have many very good people doing excellent, pretty reliable empirical work, but we don't really have new ideas and have not had them for a while.
And do you think that's because we have discovered those ideas and that was sort of it?
Yes, that is what I think. There's plenty of areas where that is true. It's not that unusual.
Yeah. It's both beautiful and sad at the same time.
That's right.
A book that changed your mind that few have read?
I love the science fiction work of Olaf Stapledon, who is a British author in the 1930s and also a Hegelian philosopher. But he has what's, in essence, a two volume set, “Last and First Men” and “Star Maker”, about the nature of intelligence and how radically worlds can change. And that's just a mind blowing set of two books to read through. I recommend them to everyone. People who read them typically love them if they like science fiction or speculative fiction at all. I would say the core lesson of those books is never underrate the possibilities, and it teaches you that quite vividly.
What's the biggest economic surprise that no one is forecasting currently?
Well, it's been a surprise to many of us that the United States, among other countries, has had such a high level of debt and the economy has not exploded. Borrowing rates are higher now as we're recording, but they're not ridiculously high once you adjust for inflation. Japan also—quite high levels of debt. They might be in some trouble, but they have a coherent working economy. They're not having a financial crisis. And if you had told people, including myself in the early 1990s, that we would be at these numbers and things would appear to be okay, we pretty much all would have been surprised. And I still am. I don't quite believe it, but there you go. I'm not sure of the market, right?
And what's one thing that you've changed your mind on recently but you haven't fully solved yet?
Well, that's everything, right? So what will an AI future be like when no one or few people will know what their job will be like in ten years, and no one will be able to plan a career with their children or grandchildren? The way, say, people in my generation could have done? How will that affect how people view the world? How will AI's affect worship and religion? How many people will worship them as deities or oracles in some manner? To what extent will they replace priests and rabbis like they know the Bible better? Right? Very open questions for me, and I don't know how to make much more progress other than just stick around, eat my cherries and see what happens.
Is the economy capped, or could we see it hit hundreds of trillions in the next decade?
It’s capped? It's not really possible, just the energy constraint for one thing. People have all these speculative ideas. We'll put a Dyson sphere near the sun, I don't know, but that's pretty far away.
What's the most contrarian city in the world right now?
The most contrarian city in the world right now might be somewhere in Sicily or rural Mexico, where people think things are simply not going to change that much. And I very much believe they're wrong. But, you know, you're asking a contrarian it may turn out they’re right. So there's other places you could pick probably in Sweden as well. But I think of rural Mexico and parts of Sicily as areas where things are just very traditional, based in human relationships. Technology enters only slowly. Old ways of life are fading away but still hanging on. And those are right now the true contrarians. It's not Silicon Valley, that has in a way, become the mainstream view.