The Past and Future of AI (with Dwarkesh Patel)

Dwarkesh Patel interviewed the most influential thinkers and leaders in the world of AI and chronicled the history of AI up to now in his book, The Scaling Era. Listen as he talks to EconTalk's Russ Roberts about the book, the dangers and potential of AI, and the role scale plays in AI progress. The conversation concludes with a discussion of the art of podcasting.

• 0:02Speaker 00

  Welcome to Econ Talk, Conversations for the Curious, part of the Library of Economics and Liberty. I'm your host, Russ Roberts of Shalem College in Jerusalem and Stanford University's Hoover Institution. Go to econTalk.org where you can subscribe, comment on this episode, and find links to other information related to today's conversation. You'll also find our archives with every episode we've done going back to 2006. Our email address is [email protected]. We'd love to hear from you. Today is March 25, 2025, and my guest is podcaster and author Dhrakesh Patel. You

• 0:43Speaker 00

  can find him on YouTube at Substack at Dhrakesh.com. He is the author with Gavin Leach of The Scaling Era: An Oral History of AI 2019 to 2025, which is our topic for today along with many other things, I suspect. Dhrakesh, welcome to Econ Talk.

• 1:01Speaker 01

  Thanks for having me on, Russ. I've been a fan, I was just telling you, for ever since, I think probably before I started my podcast, I've been a big fan. So it's actually really cool to get to talk to you.

• 1:11Speaker 00

  Well, I really appreciate it. I admire your work as well. We're going to talk about it some. You start off saying early in the book, and I should say this book is from Stripe Press, which produces beautiful books. Unfortunately, I saw it in PDF form, but it was pretty beautiful in PDF form, but it's, I'm sure, even nicer in its physical form. You say we need to see the last six years afresh, 2019 to the present. Why? What are we missing?

• 1:44Speaker 01

  I think there's this perspective in the popular conception of AI, maybe even when researchers talk about it, that the big thing that's happened is we've made these breakthroughs in algorithms, we've come up with these big new ideas, and that has happened. But the backdrop is just these big picture trends, these trends, most importantly, in the buildup of compute, in the buildup of data, and even these new Algorithms come up out as a result of this sort of evolutionary process where if you have more compute to experiment on, you can try out different ideas. You wouldn't have known beforehand why the transformer works better than the previous architectures if you didn't have more compute to play around with. And then when you look at then, why did we go from GPT two to GPT three to GPT four to the

• 2:29Speaker 01

  models working with now? Again, it's a story of dumping in more and more compute. And then that raises just a bunch of questions about like, well, what is the nature of intelligence such that you just like throw a big blob of compute at wide distribution of data and you get this agentic thing that can solve problems on the other end? It raises a bunch of other questions about what will happen in the future. But I think that that trend of this four Xing of compute every single year, increasing in investment to the level where at hundreds of billions of dollars now at something which was an academic hobby a decade ago is the mis-trend.

• 3:05Speaker 00

  And I didn't mention that you're a computer science major. So you know some things that that I really don't know at all. What is the transformer? Explain what that is. It's it's a key part of the technology here.

• 3:22Speaker 01

  So the transformer is this architecture that was invented in by some Google researchers in 2018. And it's a it's the fundamental sort of architectural breakthrough behind chat GPT and the kinds of models that you play around with when you think about an LLM. And what separates it from the kinds of architectures before is that it's much easier to train in parallel. So if you have these huge clusters of GPUs, a transformer is just much more practical to scale than other architectures. And that allowed us to just like keep throwing more compute at this problem of trying to get these things to be intelligent.

• 4:06Speaker 01

  And then the other big breakthrough was to combine this architecture with just this really naive training process of predict the next word. And you wouldn't have, I mean, it's now we just know that this is how it works. And so we're like, okay, of course, that's how you get intelligence. But it's actually really interesting, right? That you predict the next word in WikiText and as you make it bigger and bigger, it picks up these longer and longer patterns to the point where now it can just totally pass a Turing test, can even be helpful in certain kinds of tasks.

• 4:35Speaker 00

  Yeah, I think you said it gets intelligent. Obviously, that was a. You had quotes around it, but um, maybe not. We'll, we'll talk about that. At the end of the first chapter, you say, uh, this book's knowledge cutoff is November 2024. This means that any information or events occurring after the time will not be reflected. That's like two eons ago.

• 4:59Speaker 01

  That's right.

• 5:00Speaker 00

  So, how has that affect the book in the way you think about it and talk about it?

• 5:07Speaker 01

  Obviously, the big breakthrough since has been inference scaling models like O1 and O3, and even deep seeks, um, uh, deep seeks reasoning model. Um, so in an important way, it is a big break from the past. And previously, we had this idea that, uh, pre-training, which is just making the models bigger. So if you think like GPT 3.5 to GPT 4, that's where progress is going to come from. It does seem that that alone is slightly disappointing. Uh, GPT 4.5 was released and it's better, but not significantly better than GPT 4. So the next, the next sort of frontier now is this, um, how much juice can you get out of

• 5:48Speaker 01

  trying to make these smaller models, train them towards a specific objective? So not just predicting internet text, but solve this coding problem for me. Solve this math problem for me. And how much does that get you on? Because those are kinds of verifiable problems where you know the solution. You just see, get to see if the model can get that solution. Can we get some purchase on slightly harder tasks, which are more ambiguous? Probably the kind of research you do, or also the kinds of tasks which, um, are, uh, just require a lot of consecutive steps, like the model still can't use a computer reliably. Right? And that's where a lot of economic value lies. You gotta, you gotta actually be able to automate remote work. You actually gotta do remote work. So, um, that's a big, that's a big change.

• 6:29Speaker 00

  You know, I really appreciate you saying that's the kind of research you do. The kind of research I do at my age is, uh, what is wrong with my sense of self and ego that I still need to do X, Y, Z to feel good about myself? That's the kind of research I'm looking into, but I appreciate, I'm flattered by your presumption, uh, that I was doing something else. Now, uh, I have become enamored of Claude. Uh, there is a rumor that Claude is better with Hebrew, uh, than, than other, um, L L Ms, I don't know if that's true, obviously, because my Hebrew is not good enough to verify that. But I think if you ask me, why do you like Claude? It's an embarrassing answer. The typeface is really, the font's fantastic.

• 7:12Speaker 00

  The way it looks on my phone is beautifully arrayed. It's just, it's a lovely visual interface. There are some of these tools that are much better than others for certain tasks. Do we know that? Do the people in the business know that? And do they have even a vague idea as to why that is? So I assume, for example, some might be better coding, some might better, more deep research, some might better at thinking and taking, meaning taking time before answering it, it makes a difference. But for many things that normal people would want to do, are there any differences between them? Do we know of? And do we know why?

• 7:53Speaker 01

  I feel like a normal people are in a better position to answer that question than the A I researchers. The, I mean, one question I have is in the long run, what will be the trend here? So, right, it seems to me that the models are kind of similar. And it's, not only are they similar, but they're getting more similar over time, where now everybody's releasing a reasoning model. And they're, not only that, they're copying the, when they make a new product, they're like, not only do they copy the product, they copy the name of the product. So Google Gemini has deep research and OpenAI has deep research. You could think in the long run, maybe they get distinguished. And it does seem like the labs are pursuing sort of different objectives.

• 8:34Speaker 01

  It seems like a company like Anthropic may be much more optimizing for this fully autonomous software engineer, because that's where they think like a lot of the value is first unlocked. And then other labs maybe are optimizing more for consumer adoption or for just like enterprise use or something like that. So, but at least so far, tell me about your impression, but my sense is like they feel kind of similar.

• 8:58Speaker 00

  Yeah, they do. And in fact, you know, I think in something like translation, a native speak, a truly bilingual person could, might have a preference or taste. You know, if I ask, you know, my uses, actually, I'll ask you what you use it for in your personal life, not your intellectual pursuits of understanding the field. For me, what I use it for now is brainstorming, you know, help me, help me come up with a way to think about a particular problem, tutoring. I wasn't sure what Transformer was, so I asked Claude what it was. And I've got another example I've given a little bit. I use it for translation a lot because I think Claude's much better. It feels better than Google Translate.

• 9:43Speaker 00

  I don't know if it's better than ChatGPT. And finally, I love asking it for advice on travel, which is bizarre that I do that. You know, there's a zillion sites that say the 12 best things to see in Rome. But for some reason, I want Claude's opinion and I have a, and give me three hotels near this place. And I have a trust in it that is totally irrational.

• 10:08Speaker 00

  But so that's what I'm using it for and we'll come back to what else is important. Because those things are nice, but they're not important particularly. What do you use it for in your personal life?

• 10:20Speaker 01

  Research because you know, my job as a podcaster is I spend a week or two prepping for each guest. And having something to interact with as I'm, because you know, you read stuff and it's like, you don't get a sense of why is this important? How does this connect to other ideas?

• 10:44Speaker 01

  Getting a constant engagement with your confusions is super helpful. The other thing is I've tried to experiment with putting these LLMs into my podcasting workflow to help me find clips and automating certain things like that. They've been like moderately useful, honestly, not that useful. But yeah, they are huge for research. And the big question I'm curious about is like, when we get the fully, like when they can actually use the computer, then is that like a huge unlock in the value they can provide to me or anybody else?

• 11:14Speaker 00

  Explain what you mean by that.

• 11:17Speaker 01

  So right now, they're, they're just, some labs have rolled out this feature called computer use, but they're just like not that good. It's they can't reliably do a thing like book you a flight or organize the logistics for a happy hour or. Countless other things like that, right? Which, you know, sometimes people use this frame of these models are at high school level. Now they're a college level, now they're PhD level. And obviously a PhD, I mean, a high schooler could help you book a flight. Maybe a high schooler especially, maybe not a PhD. So there's this question of what's going wrong? Why can't they be so smart in this? I mean, they can answer frontier math problems with these new reasoning models, but they

• 12:00Speaker 01

  can't help me organize, they can't like play a brand new video game. So what's going on there? I think that's probably the fundamental question that we'll learn over the next year or two is whether these like, these like common sense foibles that they have, is that sort of intrinsic problem where we're under? I mean, one analogy is, I'm sure you, you've heard this before, but like, the sense I get is that when Deep Blue beat Kasparov, there was a sense that like a fundamental aspect of intelligence had been cracked. And in retrospect, we realize that actually the chess engine is like quite narrow and is missing a lot of the fundamental components that are necessary to, say, automate a worker or something. I wonder if in retrospect, we'll look back at these models, if in the version where I'm

• 12:45Speaker 01

  totally wrong and these models aren't that useful, we'll just be, we'll just think to ourselves, there was something to this long-term agency and this coherence and this common sense that we were underestimating.

• 12:55Speaker 00

  Well, you know, I think until we understand them a little bit better, I don't know if we're going to solve that problem. You asked the head of Anthropic, something about whether they work or not. He said fundamentally, what is the explanation for why scaling works? Why is the universe organized such that if you throw big blobs of compute at a wide enough distribution of data, the thing becomes intelligent? And Dario Amodei of Anthropic, the CEO, said, the truth is, we still don't know. It's almost entirely just a contingent empirical fact. It's a fact that you could sense from the data, but we still don't have a satisfying explanation for it. It seems like a large barrier that, that.

• 13:42Speaker 00

  Unknowing seems like a large barrier to making them better at either actually being a virtual assistant, not just giving me advice on Rome, but booking the trip, booking the restaurant, and so on. Without that, how are we going to improve the quirky part of hallucinating part of these models?

• 14:03Speaker 01

  Yeah, yeah. And this is a question I feel like will get a lot of good evidence in the next year or two. I mean, another question I asked Dario in that interview, which I feel like I still don't have a good answer for, is, look, if you had a human who had as much stuff memorized as these LLMs have, right, they know basically everything that any human has ever written down, even a moderately intelligent person would be able to draw some pretty interesting connections, make some new discoveries, and we have examples of humans doing this. There's one guy who figured out that, look, if you look at what happens to the brain when there's a magnesium deficiency, it actually looks quite similar to what a migraine looks like. And so if you solve a bunch of migraines by giving people

• 14:45Speaker 01

  magnesium supplements or something, right? So why don't we have evidence of LLMs using this unique asymmetric advantage they have to some intelligent ends in this creative way? There are answers to all these things. People have given me interesting answers, but a lot of questions still remain.

• 15:04Speaker 00

  Yeah. Why did you call your book The Scaling Era? And that suggests there's another era coming soonerish, if not soon. Do you know what that's going to be? It'll be called something different. Do you know what it'll be called?

• 15:20Speaker 01

  The RL era? No, I think it'll still be. Scaling refers to the fact that we're just making these systems like hundreds, thousands of times bigger. If you look at a jump from something like GPT-3 to GPT-4, or GPT-2 to GPT-3, means that you have 100xed the amount of compute you're using on the system. It's not exactly like that, because there's like some, over time, you find out ways to make the model more efficient as well. But basically, if you use the same architecture to get the same amount of performance, you would have to 100x the compute to go from one generation to the next. And so that's what I'm referring to, that there is this exponential buildup in compute to go from one level to the next.

• 16:03Speaker 01

  The big question going forward is whether we'll see this, I mean, we will see this pattern because people will still want to spend a bunch of compute on training the systems. And we're going to get, we're on schedule to get big ramp ups in compute as the clusters that companies ordered in the aftermath of ChatGPT blowing up are now coming online. And the question is, how, I mean, then there's questions about like, well, how much compute will it take to make these big breakthroughs in reasoning or agency or so forth? But stepping back and just like seeing a little forward to AGI.

• 16:40Speaker 00

  Artificial general intelligence.

• 16:43Speaker 01

  That's right. There will become a time when an AGI is, can run as efficiently as a human brain, at least as efficiently, right? So, human brain runs on 20 watts. An H100, for example, takes on the order of 1,000 watts. And that can store maybe the weights for one model or something like that. We know it's physically possible for the amount of energy the human brain uses to power a human level of intelligence. And maybe it's going to get even more efficient than that. But before we get to that level, we will build an AGI which costs like a Montana's worth of infrastructure and $100 billion of CapEx and is clunky in all kinds of weird ways.

• 17:27Speaker 01

  Maybe you have to use some sort of inference scaling hack. By that, what I mean to refer to is this idea that often you can crack puzzles by having the model think for longer. And in fact, it like, it weirdly keeps scaling as you add not just like one page of thinking, but 100 pages of thinking, 1,000 pages of thinking. And I often wonder, so, there was this challenge that OpenAI solved with these visual processing puzzles called RAGI. And it kept improving up to like 5,000 pages of thinking about these very simple visual challenges. And I kind of want to see, what was on page 300? What big breakthrough did it have that like made that? But anyways, so there is this hack where you keep spending more compute thinking and that gives you better output. So, that'll be the first AGI.

• 18:10Speaker 01

  But, and it'll get more, and we will build it because it's still valuable to have, it's so valuable to have an AGI that we'll build it the most inefficient way. We don't, we won't, the first one we will build, won't be the most physically efficient one possible, but, yeah.

• 18:25Speaker 00

  Could you think of another technology where trial and error turned out to be so triumphant? I did a wonderful interview with Matt Ridley while back on innovation and technology. And, you know, one of his insights, and I don't know if it's his, but one of the things he writes about, I think it says, is that a lot of times the experts are behind the people who are just fiddling around. He talks about the Wright brothers are just bicycle guys. They weren't, they know anything about aerodynamics particularly. They just tried a bunch of stuff until finally they lifted off the ground. It's the implication of, I don't know if, I think that's close to actually true.

• 19:07Speaker 00

  Here we have this world where these unbelievably intellectually sophisticated computer scientists are building these extraordinarily complex transformer architectures, and they don't know how they work. That's really weird. And if you don't know how they work, the easiest thing to make them better is just do more of what worked so far and expect it to eventually cross some line that you might be hoping it will. But can you think of another technology where the trial and error is such an important part of it alongside the intense intellectual depth of it? It's really quite unusual, I would guess.

• 19:46Speaker 01

  I feel like most technologies, I mean, I would actually be curious to get your takes on economic history and so forth, but I feel like most technologies probably have this element of individual genius is like sort of overrated and building up continuously on the slight improvements. And often it's not like one big breakthrough in the transformer or something. It's like you figured out a better optimizer. You figured out better hardware, right? So a lot of these breakthroughs are contingent on the fact that we couldn't have been doing the same thing in the 90s. In fact, people had similar ideas. They just weren't scaled to a level which helped you see the potential of AI back then. But I do think this is actually a really important question, Russ, because the biggest, I mean,

• 20:31Speaker 01

  the sort of big question here is not like what model do we want to use this year or something. The big question is, will intelligence feed back on itself? And to what extent will it feed back on itself? And if it does, do we get some sort of superhuman intelligence on the other end? Because the things are making better models, or something like that. And there, the question is, okay, can you just have a million super intelligent AI researchers, a million automated Ilya Sutskevers or Alec Bradfords? And they think about like, what is the, what is the architecture of the human brain? And how do we replicate that in machines? Or do you need this sort of evolutionary process, which requires a ton of compute for experiments, which maybe even requires hardware breakthroughs? And that will still be transformative.

• 21:14Speaker 01

  And, you know, hopefully, at some point, we can talk about this. I am keen to get more economists' takes on the potential of explosive growth and so forth. That's still compatible with a world where it takes, you know, more than a year to get intelligence explosion. But that is a fundamental question of like, does intelligence feed back on itself or not?

• 21:31Speaker 00

  Yeah, I don't, I think intelligence is a little bit overrated, you know? So I'm a little bit, I'm kind of a skeptic on this. And I also believe that most of the really tough human problems aren't insoluble because we don't, we're not smart enough. It's because the world's complicated and has nothing to do with intelligence. Because there's trade-offs and the definition of good is not well-defined, or best, or better even. But I know that puts me in a small pessimistic camp. And I really don't think it matters, actually, because we're going to see a lot of these changes, we'll see in real time, we'll see if they work or not. You know, thinking back to the trial and error, I realized while you were answering

• 22:16Speaker 00

  the question, like, isn't this common? Well, the pharmaceutical industry has a lot of hit-or-miss, wild guesses, and then something works. So we imagine a day where, because of our genetic or biotechnology, we could custom design pharmaceuticals more effectively. But I think most of, we're not at that day yet. And so there is a lot of that in that industry, in that world. So maybe it is more common than I think. I don't know.

• 22:45Speaker 01

  Yeah. I mean, there was, who is that economist or Alan Bloom or Robert?

• 22:52Speaker 01

  Robert Allen, who had the theory of the industrial revolution where it happened in Britain first because coal was cheap enough there that you could make these initial first machines that were actually super inefficient, where I think the first steam engine used the pressure from the steam condensing to move the piston back, whereas future more efficient machines would push it directly with a steam. And so anyways, the other countries just didn't get on this sort of, at least beforehand, didn't get on this evolutionary step ladder where you make the inefficient machines, the coal is cheap enough, you can just like throw coal at it, see if you can get it to work, and then later on you get up the cost curve and you find these improvements and so forth.

• 23:34Speaker 01

  Similar thing has happened with hardware with AI, where in the 90s and the 80s, people had these ideas about, well, you could do deep learning and you'd have these different architectures, and now we have the compute to actually try out these ideas, and we'll get more and more compute over the next few years, and maybe we should expect an acceleration, but similar trends.

• 23:52Speaker 00

  I'm thinking about Fred Smith, I've told this story before, but the first night of Federal Express, the story goes that there were two packages that were carried by Federal Express, and one was Fred Smith sending a birthday present to his mom. So it didn't start off going gangbusters, and after quite some time, they couldn't make payroll and they had to close the doors, except that Fred, on his way back from the bankers in Chicago saying no, saw a flight to Reno, went to the roulette wheel and made enough money with his sister's trust fund that he kept that company afloat. And it would be nice to say that Fred understood that this was an inevitable idea, but we

• 24:35Speaker 00

  know the real reason that he kept going was that he's a human being and he had put his heart and soul into this, and he was not going to give up until every avenue had been pursued. And it wasn't a theoretical understanding of the world that allowed him to do that, but something more human, I think.

• 24:57Speaker 01

  Can I try out an idea on you? Yeah, sure. So, part of me thinks, well, like, look, this idea of a single superintelligence because of its ability to think in an armchair and divine how the world works, you know, some people have this idea that like, oh, it's gonna figure out how to get rid of its political opponents and build a nanobots and whether it's like persuasion or some sort of other manipulation, coming up with future technologies all by itself, it'll have this omnipotent effect on the world. Yeah. Now, I don't buy that, but here's an idea I do buy. So if you look at what changed between primates and humans, it was our ability to, it wasn't just individual, Joseph Fenner has this wonderful book about this, it wasn't just, it wasn't

• 25:41Speaker 01

  just intelligence, it wasn't even mainly intelligence, it was our ability to coordinate with each other, to share knowledge, to accumulate knowledge over generations. Now, when we think about AIs, instead of thinking of a single AI, superintelligent AI, what if we thought about billions of beings who are thinking at superhuman speeds, who are able to communicate with each other in ways that humans simply cannot? They can literally share their mind states with each other, they can distill their insights, they can merge, like, literally merge, they can copy themselves arbitrary amounts of time, so if you have a super skilled worker or who has some tacit knowledge, you can just make arbitrary copies of them for pennies per copy, and fundamentally you have this much larger population size, and if you, you know, believe, a lot of theories of economic

• 26:24Speaker 01

  growth say that more people means more ideas, and that feeds back on itself. So then, you know, for the same sort of like, old school economic growth, cultural revolution reasons, I'm like, this hive mind of AIs actually will be, not as an individual AI, but as a sort of like, whole economy-wide phenomenon, will just be a total shock and a change from the current system.

• 26:50Speaker 00

  Could be. Let me ask you a question, since I'm going to duck yours. So I really like my smartphone, an embarrassingly large amount, and I recognize that my affection for it is somewhat akin to an addiction, as best as we can define that. And I sometimes wonder, and I've probably asked this question before on the program, if Steve Jobs were alive, and let's say he thought Jonathan Haidt was right, I mean, I'm agnostic on the Jonathan Haidt work, but I actually, I do think he's on to something. I do think phones are not so great for kids, and I'm pretty sure they're not so great for adults.

• 27:31Speaker 00

  And I'd ask Steve Jobs, would you like to turn the clock back and not think of this product? It's just a phone. It's not a computer in your pocket. Would you do that? And I think human beings have a lot of trouble with saying, oh, yeah, I guess it's been a, maybe not a net benefit. It's, I'm uneasy with it. I think most humans would have trouble doing that. I think they would say onward and upward. Similarly, if we had the billion AI agents, or whatever we call them, one of the first questions you might ask them is, do you think it's good for humanity, for you to be unleashed?

• 28:13Speaker 00

  Forget the incentive, the problem there with self-interest, if that's definable in that world. Could they answer that question? I mean, is that just an example to me of something that intelligence is not really good at? And the fact that they could all be linked together and share their thoughts, you'd have to believe, and this is a very Hayekian point, we're bringing in a little bit of economics here, you'd have to believe that they not only could coordinate their thinking and not only pull together all the knowledge that they, which they all share, by the way, so it's not so different, unlike humans, they will probably have a very, a similar database. Are they actually going to be able to forecast?

• 28:59Speaker 00

  The N futures that would exist in different worlds under different policy outcomes. And of course, then they'd have to then aggregate the well being, which is literally, to me, a meaningless question. But you could teach them to do it. They'd be really enthusiastic about it. The people who are in charge of it would be enthusiastic about it. But I think about, if you go back to Hayek's Nobel Prize address, he basically says macroeconomics is unknowable. Just like you can't know which sports team's going to win tomorrow. On average, one of them might be more likely to win. But if the quarterback of the other team, or the striker of the other team, or the pitcher of the other team got into a fight with his wife the night before. But in theory, if you really believe the world's deterministic,

• 29:42Speaker 00

  which I'm agnostic about, but skeptical, you just know all that, too. Because you'd have the enzymes and the chemical imbalances and all the data on all these people. And they could actually forecast all these outcomes. So that's what I'm skeptical about. Do you think that's, we're going to go in that direction?

• 30:03Speaker 01

  Ultimately, I think not. But let me just lay out the case for somebody who might think that something like this is plausible. Right now, central planning, or this kind of central coordination and forecasting, there's many reasons that it doesn't work. But one of them is Xi Jinping has the same amount of compute in his brain as every other person in China, right? 10 e, 15 flops. In the future, there could be a scenario where the central power, or the coordinator or whatever, has much more raw compute than the periphery and also has much greater bandwidth with the periphery,

• 30:45Speaker 01

  where right now Xi Jinping can only monitor so much as one human. The future versions of him could be like, you can run a copy of Xi Jinping for pennies per hour, but you can run mega Xi Jinping, which is like 1,000 times bigger and is running 1,000 times faster speed and has copies of it monitoring every single communication coming out of the CCP and every single time that you interact with a CCP website or something. And then they just distill all that back into the central blob or something. So that's the case for why it's plausible. The reason I still think that this kind of central planning will still not work is

• 31:27Speaker 01

  The whole, I think this perspective underwrites the extent to which the whole world is getting more complicated as a result of AI. So if you took Apple as it exists today, or just like a computer, you're like, could, could Apple as it exists today coordinate the, the, I don't know, maybe Roman economy is pretty complicated, but like, the economy of Uruk? Maybe it could, like, maybe the economy of Uruk can be coordinated in this way. But could Apple coordinate the world economy as it exists today? I doubt it, right? And so maybe ASI might be able to coordinate the economy as it exists today, but not the one where there's also other AGIs were deployed through the entire economy, or doing their different things. I think the big worry I have is, before we get to this decentralized world where every

• 32:08Speaker 01

  sector has, you know, drastically grown and become more complicated, and then because higher population size, there's much more specialization, and all these new institutions for AIs to coordinate with each other that we figured out for humans, like joint stock corporations, and state capacity and whatever, whatever that looks like for AIs. There will be a sort of like decentralized state where this has been sort of like, there's an equilibrium. But before we get to that, should we worry about a, somebody having a big head start? They get the first hive mind, and this hive mind, because the rest of the world hasn't adopted AI, is more powerful than the rest of the world combined. And that question really comes down to the speed of deployment versus the speed of AI

• 32:53Speaker 01

  feeding back on itself. And it's a big debate, and we sort of touched on it before, but it's probably the most important debate in the world today. And it's funny that we don't pay enough attention to it, but...

• 33:05Speaker 00

  Well, I think one of the reasons we don't, we don't really understand, ASI, by the way, is artificial superintelligence. I think the reason we don't pay enough attention to it is we don't, it's far away in emotion. It's not far away in time, perhaps. But I want to go back to something you said earlier, just to give you a hard time. Talking about a leader who's got really high access to a lot of intelligence. I doubt there's a very good correlation between the IQ of, say, an American president and the success of their administration. Now, I think you'd have to argue, "Well, I know, but that's just human intelligence. This is going to be so far beyond 180 IQ. It'll be like 10,000 IQ.

• 33:50Speaker 00

  And so it's not even the same thing. We can't even, again, it's very, I like to tease people in this field. It's very related to religious thought. You know, it's like, we can't imagine the mind of God, who can see every person, know every thought, not just here, but in Alpha Centauri as well, if they're, you know, in that galaxy, far, far away. And I just, I wonder if that's just wishful thinking, but, you know, it's hard to know. Hard to know. I want to cue you a quote from a recent interview I saw. This is from, I saw this on Sve Maucheritz's page today. He quotes an interview between Daniel Kokotajlo and Dean Ball.

• 34:34Speaker 00

  I don't know where, I couldn't see a source where he got this, but we'll link to the Sve Maucheritz page if we can. He says, can you, he asked Dean Ball, who's a thinker on AI, he says, "Could you give some examples of things that AIs will remain worse than the best humans 20 years from now? That they'll be worse at?" Dean Ball's answer is fabulous. "Giving massages." I'm not sure he's right there, but, okay. "Giving massages, running for president, knowing information about the world that isn't on the internet, performing Shakespeare, tasting food, saying sorry." Now, that's a nice list.

• 35:16Speaker 00

  It covers a very wide range of the more glorious, not all of it, but some of the most glorious parts of the human experience. What do you think? Are there going to be things that human beings, 20 years from now, human beings are still, besides those, unless you want to disagree? Can you think of some things that AI is going to struggle to dominate us, or they can pretty much dominate us all over the, all over the board?

• 35:42Speaker 01

  I think 20 years is a long time.

• 35:43Speaker 00

  Yeah, it is.

• 35:44Speaker 01

  I like the sentiment behind that list, but I'd be on board if it was more to like two years instead of 20. The, so here's, here's something that's super paradoxical and surprising about the way in which these AIs are getting better. In the 90s, Hans Moravec, which is actually the sort of the first guy who came up with this idea of scaling, you can predict when human intelligence comes about as a result of how much compute you're using. And in fact, he predicted 20, in the 90s or 80s, he predicted 2028 as a year in which we'd get the compute to build a human-level AI. And it's a very simple projection, you know, like not that far off. We'll see, we'll see, I guess. But he had this really interesting paradox, which is that when we think about what is

• 36:29Speaker 01

  hard, what is, what it will take a long time to sort of like develop as a skill in AIs, we think about what is hard for humans. But we don't think about the amount of time that evolution has had to spend optimizing a skill. And the kinds of skills that evolution has to spend a bunch of time optimizing are just so commonplace that in the humans, in the human spectrum, they just don't vary that much. Because, you know, for a billion years, you've been learning how to move around and have control over your limbs and understand your environment and have this long-term coherence. Only for the last couple hundred thousand years have you maybe like learned anything relevant to doing math or developed the relevant cognitive skills. And so these things have cracked, like are about to crack frontier mathematics, coding,

• 37:14Speaker 01

  so forth. So, but there is this old-school perspective on human intelligence, like the Aristotelian perspective, that what makes us human is our ability to do these kinds of reasoning, reasoning tasks and to have these higher-level abstract thoughts. In fact, that's probably the least human thing about us, or the perspective of what will AI get first? It will get this sort of Aristotelian understanding of humans, that will be the first skill it gets. And the sort of like very reptile brain stuff will be the, actually the last skills that it automates.

• 37:48Speaker 00

  Yeah, the...

• 37:51Speaker 00

  I think it's the amygdala, I can't remember, but the part that's working behind the scenes to figure out stuff that you can't even know you're figuring out.

• 37:59Speaker 01

  Yeah.

• 38:02Speaker 00

  So I haven't described your book, but what your book is, you and I have something in common. You've interviewed a lot of people about AI, and I have too. You've interviewed, I asked you before we started, you said maybe about 20. I think I've done 15 or so.

• 38:18Speaker 00

  In one sense, I'm way ahead of you, Dhrivesh. I interviewed Nick Bostrom, I think in 2014, when he was worrying about it. But you're ahead of me. And more than that, what this book is, is it's a compendium, it gives me ideas. It's a compendium of excerpts from the interviews you've done, grouped around these important issues in AI. And it's very, as a result, it's very interesting. And as I'm reading the book, even though I know some of the insights from my pitiful 14 interviews, but you actually not only, probably, you only did more, you probably remember them better because you wrote the book, pulled the book together. So that's an interesting cognitive thing. But one thing that struck

• 39:01Speaker 00

  me is despite having interviewed all these people, and I've interviewed them across the spectrum, as you have, worriers, optimists, practitioners, philosopher types, your book, in a way, was kind of the first time I got a little bit scared. And so I want to talk about why. And I don't want you to react to it. And it's going to be a long question. It's already too long, I apologize. But you wrote an essay called "What Fully Automated Firms Will Look Like." And you've been alluding to it in some of the conversations we've had already. Here's what you write. "Even people who expect human-level AI soon are still seriously underestimating how different the world will look when we have it. Most

• 39:44Speaker 00

  people are anchoring on how smart they expect individual models to be, i.e. they're asking themselves, what would the world be like if everyone had a very smart assistant who could work 24/7?" Everyone is sleeping on the collective advantages AIs will have, which has nothing to do with raw IQ, but rather with the fact that they're digital. They can be copied, distilled, merged, scaled, and evolved in ways humans simply can't. What would a fully automated company look like with all the workers, all the managers as AIs? I claim that such AI firms will grow, coordinate, improve, and be selected for it unprecedented speed. Now, let's not debate whether that's really feasible and whether it's soon, or tomorrow, or 40 years from now, or 100 years from now. Let's take it as true. And I'm reading that, I realized,

• 40:28Speaker 00

  along with some of the really bullish optimism of some of these practitioners, which you talk about at the beginning of the book, some of it, they're self-interested, but a lot of it's just they've drunk the Kool-Aid. It's not a... It's just the way they see the world. And it's... There's something kind of beautiful about it. But I realized when I finished some of those excerpts and some of the remarks on those people in the field, it's not that I got scared, I just got sad. And I want to lay out the sad thing, and you can talk about scared and sad. And I worry, and maybe it's a silly worry, that so many of the joys of my life will not be available to my granddaughter, who's two and a half. I remember the time I was taking linear algebra, and I realized you could prove

• 41:11Speaker 00

  the central limit theorem in multiple ways. Not so... Sorry, not linear algebra, it's an advanced statistics class. Those are the two hardest classes I took in college, by the way. So that's why I got confused. So it turns out you could prove it with something called characteristic functions, and then you could prove it with something called eigenvision, I think. I don't even remember. I was 21 years old, so almost 50 years ago. And when I saw... I just remember the feeling I got when I saw that these two mathematical techniques that were nothing alike could get to the same result, and that there was some underlying coherence there. And it was so exhilarating. And so I asked Claude, just before we started talking, to do it, give me those proofs. And I can't follow them anymore. But, you know, I could ask Claude to help me understand them. And so my granddaughter will probably

• 41:56Speaker 00

  never have that thrill. At some point, pretty late in my life, my father told me I was a good writer. I think I was in my 50s. It was thrilling. He spent his life criticizing my writing. I think he thought that was good for me, and maybe it was, but that made it all the sweeter when he praised something I'd written. I don't know if my granddaughter will ever find her writing thrilling someone, given the way she's gonna grow up. Buying my first car, being able to afford it, having a sense of agency and success, the thrill of discovering a used book that you've been looking for, that's been gone for a while, the thrill of discovering a novel thinker or a new podcast with Dhrivesh Patel, it's like, who is this guy? Yeah, you reached out to me. I didn't know, I apologize. I thought I knew your name, and I saw what

• 42:41Speaker 00

  you've been doing and who you've been talking to, it's like, wow, this guy's doing great. Then I found out you're how old?

• 42:47Speaker 01

  Twenty-four.

• 42:48Speaker 00

  Yeah, twenty-four. Good grief. So that's really fun. I'm gonna start reading your stuff. And this is just riffing off the top of my head. Now, I understand my granddaughter will have different thrills, maybe better ones, but reading your essay in your book made me realize for the first time that AI isn't just a better tool that will save me time, or to say the way Google Maps does, or Google Translate, it's gonna remake the world in ways that will change the fabric of daily life, not just making it faster or cheaper, but really, really different. And the world might be a better place, it will be more materially prosperous, for the reasons you write about in that essay. It's not a small thing, but will there be more human flourishing? I'm not so sure. For some

• 43:33Speaker 00

  people, absolutely. Many of them, you interviewed for your book, and they'll have the extraordinary experience of knowing they haven't just put a dent in the universe, but a big dent. And it's this thing, this thing that I can interact with, that is mysterious, is really exciting, but it's gonna change everything. And it's scary. And not in the way that it's gonna turn me into a paperclip, or run roughshod over me, or ignore me as a mere mortal. It's just gonna, the world's gonna be really different. And I wonder if even, what's it gonna be like to read a novel written in 1950 for my granddaughter? I don't know. Your thoughts? Sorry about that long-winded thing.

• 44:17Speaker 01

  No, no, I mean, I'm actually glad, because it is such an important question, and it's something we kind of gloss over when we're having this big picture discussion about what are the trends in AI or something.

• 44:32Speaker 01

  I wouldn't even dismiss the paperclip or concerns because I have so much uncertainty. I've changed my mind so many times. I've changed my mind in the last week on some of the most important questions about AI. And so I'm like, who knows? Like, this has never happened before. We'll see. But, okay, so putting that aside for a second.

• 44:49Speaker 01

  Yeah, so it is the case, and I think it's worth being honest about this instead of having cope, that sooner or later, there will be a point in which a human is just unlikely to add anything sort of like counterfactually valuable, at least in a sort of like count the GDP number sense, right? Where horses right now, you can ride them for fun, but you're not, they're not there because they are needed to move you around. And it is obviously a big change. I have this intuition that humans have adapted to such diff, like, think about the world today and how many different transitions there have been,

• 45:34Speaker 01

  not just from the forager past. And even then, people underestimate how big a deal fire was or how big a deal tools were. The agricultural revolution, the industrial revolution, secularization of society, the buildup of states instead of having just like your local kin or whatever. If we've dealt with all of that, I feel like we'll do okay with material superabundance, but you're not that, sort of in the grand scheme of things, economically productive. And the other thing I think people gloss over, which I think is important to the picture of what happens after, but people are maybe uncomfortable about, is transhumanism.

• 46:17Speaker 01

  I do think it's important that we give humans a freedom to have their minds, their bodies, their spirits.

• 46:28Speaker 01

  Amplify, change, whatever they want. And the intuition there is just, people think of this in a very sort of dystopian way of cyberpunk or something. But just, but there is a sort of like Schumpeterian or Hayekian thing of like, we just like don't know what we don't know about what the future, like instead of just like thinking about new technology as dystopian, which is like, try to imagine how the future might view the fact that they have minds who can experience greater beauty, greater creativity, greater connection with other minds, aren't debilitated by the kinds of things that we find cumbersome. And I mean, another intuition here is, if you, if, I'm sure you've heard this thought experiment, but I think it comes from Phil Tramwell, where if I said, I'll send you back

• 47:12Speaker 01

  to the year 1000, but how much money would I have to pay you in the year 1000 such that you get this much wealth in the year 1000 and I send you back to that year? I think it's like quite plausible, the answer is there's no amount of wealth that I would rather have in the year 1000 than just be alive right now. And so this, especially in the future, when we're talking about the kinds of things AI, high minds will invent, whether it's on health, whether it's on just this well-being and flourishing and connection and it's, I think it's like hard to estimate. I think it's very likely your granddaughter or your granddaughter's granddaughter will have a similar perspective to us that we have to the people in year 1000.

• 47:52Speaker 00

  Yeah, I've never heard, I don't know who Phil Tramwell is, but that thought experiment is kind of, I spent way too much time writing about in my work.

• 48:02Speaker 01

  Oh, sorry. Maybe actually, maybe you're the source.

• 48:05Speaker 00

  No, no, no.

• 48:06Speaker 01

  I got secondhand.

• 48:07Speaker 00

  No, it's not me, because I never go back to 1000, I go back to 1950. And there's almost nobody alive today who would literally, if they could experience 1950 in day-to-day moments, having experienced 2025, would go back. And the strange part of it, of course, is that people in 1950 were plenty happy, some of them, some of them weren't, obviously. We could debate how many and what their characteristics were that was, and they were treated badly and so on. But I, so I don't, I agree with that thought experiment. I think it's powerful and profound. I think.

• 48:47Speaker 00

  It's not just that I'm used to this and going back to 1950 would be painful. I live longer. I I live higher quality life. My day to day life is more interesting because I'm not working as a farmer in 1900, say. So I don't. I used to be. I used to be you, Dhrakesh. I used to have that boundless optimism about the power of human creativity to make the world better. But I'm older than you. I'm not wiser. But when you're older, I think you get a certain. You lose some of your optimism.

• 49:29Speaker 00

  I don't think it's necessarily rational to lose it, but it comes from a variety of things outside of of analytical thinking. So it's nice to be reminded of it. I hope you're right.

• 49:40Speaker 00

  I hope. You know. Having an artificial knee does not ruin the human experience.

• 49:48Speaker 00

  So so it's nice to think that artificial intelligence will enhance our experiences together and individually in ways that will be meaningful to us. So I like the idea of it.

• 50:01Speaker 01

  Yeah, and you and I, Russell, will be in a especially good position because what we're doing. I mean, we've got our. We've got our post AGI career already.

• 50:09Speaker 00

  Correct. Are you sure?

• 50:13Speaker 01

  I, yeah, I think.

• 50:15Speaker 00

  Isn't there?

• 50:15Speaker 01

  Maybe the.

• 50:16Speaker 00

  Isn't that new technology? Is it some chat GPT, I think, that that creates podcasts between two interesting, thoughtful people? We're like obsolete almost already.

• 50:25Speaker 01

  And it'll it'll get better, but I think the sort of

• 50:29Speaker 01

  vicarious experience of knowing. I mean, it's already the case that on any given subject I might do a podcast on, there might be a better resource than that particular podcast. But people still want to tune in because there's some consistency. There's some feeling of continuity and intellectual camaraderie. Maybe that maybe you, once you have personal AI friends and you're the your actual friend is the one interviewing your virtual. Russ Roberts about his new book or something that changes, but I yeah, at least I prefer this over my previous. My the career I was probably going to end up in as a computer science major as I'm thinking about what will survive AGI.

• 51:11Speaker 00

  I feel that too. We'll talk about that in a minute, I hope, why that is the case. I referenced recently in the conversation with Ian Leslie, it hasn't aired yet, but shortly. That in my conversation with Chuck Klosterman, people, he wonders about the question of who will be remembered as the. As the quintessential rock and roll band from this short era in the 60s and 70s when rock and roll was a was a thing. And he points out, Klosterman does, that usually it's just one. And 100, 200 years from now, will anyone be remembered as the creator of AI? Is there, will there be a name when someone will say, wow, you know, Steve Jobs really changed the world of computing.

• 51:56Speaker 00

  Steve Jobs will, I think, beat out Bill Gates, but we'll see. But on this, it's interesting to me. You said, you said, oh, you said, oh, yeah, that was figured out. The transfers, a couple of researchers from Google, you didn't mention their names. Part of it's because there's more than one, so that's makes it challenging. Is anybody you think going to be immortalized as quote the creator, even if it's not true?

• 52:19Speaker 01

  Only in retrospect, maybe Jeffrey Hinton, maybe Ilya Sutskever. Though it is interesting that obviously I'm not any, I'm not a practitioner or anything. I just happen to know about it from reading about it. But even just having some familiarity with what kinds of breakthroughs have happened in AI. This near mode where you're like, oh, there isn't a one big breakthrough that is the locus of progress. It is just the industry and the fact that you at all, from all layers of the stack, from hardware to software to everything in between. You've been making these small improvements. And this is probably related to the idea of like gelman amnesia.

• 53:03Speaker 01

  It's not the same concept or where the more in this case, maybe if the industry you understand best, you understand it's a result of all this combinatorial human creativity where people are making these small discoveries and it's the hive mind of humans talking to each other and so forth. Maybe the industries you don't understand also work this way. And whenever you think about like, oh, this is the genius behind flight or cars or whatever break through steam engines, it's also a similar story.

• 53:35Speaker 00

  Yeah, I think that's a great insight. I love that. I think that might be true. You know, when you're close to it, you understand how complicated it is, but it gets distilled into this actually quite misleading idea of the great innovator when it's actually a hive mind anyway. It's just not obvious unless you're really close to it.

• 53:55Speaker 01

  Yeah.

• 53:56Speaker 00

  Do you have any generalizations after spending a lot of time with these folks? So not only have you interviewed more people, but your podcasts are longer than mine. So you're way ahead of me on the contact. One thing that strikes me is they're awfully optimistic. On average, there's some warriors, but a lot of them are just cheerfully optimistic about technology and the future. Any thoughts you have about them as people or their perspective?

• 54:25Speaker 01

  I think maybe I'm a biased reporter here because I'm friends with many of them and so forth and I've interviewed them. But I think we live in a good, one of the good universes. I think there's a lot of adjacent universes where the kinds of problems that we should be thinking about, I think we should be thinking about the paper clippers because again, it's so hard to think about. Maybe we'll look back, I guess we won't be looking back, but maybe somebody will look back and say, Eliezer had it right all along. But the fact that the kinds of people who are building these labs.

• 55:06Speaker 01

  Think about this. I mean, you may say it's lip service, but lip service is better than nothing. And often, these solutions are, maybe a lot of the key solutions will just end up being like, you do the obvious thing, that's not going to cost you that much, but you have to make it somewhat of a priority. And there's a lot of adjacent universes where the idea of a misaligned AI, or the idea of this hive mind getting out of your control, is just something that doesn't occur to people. And you just think about them as tools until you get to this point. And so I do appreciate the fact that these people do seem to be, and I take them at their word, or at least give them benefit of the doubt unless shown

• 55:49Speaker 01

  otherwise, that they got into this, or as they got into it and they thought more seriously about the prospect of AGI, they do take this like, what happens in the world after question seriously. And it's better than what it could have been.

• 56:04Speaker 00

  The eleazar you mentioned is eleazar Yudkowsky, who is one of the warriors. I'll link to my interview with him. And do you have one also? We'll link to...

• 56:16Speaker 01

  Yep.

• 56:17Speaker 00

  Okay, we'll link to that as well. How has immersing yourself in this world changed you, if at all? Besides keeping up at night versus not worrying versus keeping up at night, which you alluded to a minute ago?

• 56:34Speaker 01

  One thing that has happened is, whatever interview I do about whatever topic, it just, I can't help but ask, how does this relate to AI? And it can be sort of, I'll be interviewing this geneticist of ancient DNA, David Reich, who I'm sure you must have interviewed at some point.

• 56:54Speaker 00

  No, I haven't.

• 56:57Speaker 01

  And we're talking about what happened 60,000 years ago, and why did this one tribe wipe out this other tribe? And I'm like, oh, or you know, why did humans, what were the genetic changes between humans and primates? And then you're like, does this provide evidence for the scaling hypothesis? Are the structural similarities between monkeys and humans' brains similar? That it was only scaling that changed. Or when I have Joseph Henrick on, author of Secrets of Our Success. I'm like, well, why don't you think the hive mind of AIs is going to be even more powerful because of the population size and the bandwidth? So maybe it has made me more of a one-track thinker, but it just because if you think this is happening soon and if you think this is it's also such an intellectually fascinating question because you might think that AI is just about answering these

• 57:41Speaker 01

  technical questions about what will happen in scaling, what will happen with these new training techniques. But fundamentally, you're trying to figure out what will a future society of these different kinds of beings look like? And there's no field of human knowledge, which is irrelevant to answering that question.

• 57:58Speaker 01

  You need to know the economics of growth theory or you know, how do we model billions of extra agents?

• 58:08Speaker 00

  Labor markets. Labor markets, we'll call it that. It's a it's a it's a form of hand waving that sounds somewhat systematic. Yeah.

• 58:17Speaker 01

  And you mean, even the primatology and cultural anthropologists, geopolitical thinkers about how will states think about this kind of competition for this kind of power? It's all relevant. And so the great thing about the book was the fact that you could just put these different kinds of thinkers side by side and see how they use their conceptual tools to answer these rather gnarly questions.

• 58:40Speaker 00

  That's very cool. While we're talking, I think I think this is true. I think if you believe that AI will eventually be able to anticipate the future perfectly because it'll be so smart. Of you, again, I'm highly skeptical of, but it's interesting to contemplate it. I think it'll also be able to backcast. Want to be able to totally recreate the past for the world's current conditions and I mean, if you're determinist, I'm not, but but I might, maybe I should be. So that'll be great. We're going to go back, Joseph Henrick, all those those folks who've done all this this incredibly imaginative speculation based on four fossils and you know, two carcasses found preserved in tar. They're going to find out whether they're right or not.

• 59:19Speaker 01

  There's actually a really, um, uh, uh, there's this interesting speculation that, look, if they made, I mean, I'm sure you heard the bottom argument about there'll be more simulations than real world and so we're probably in a simulation. There's an interesting wrinkle to that.

• 59:34Speaker 00

  Yeah.

• 59:35Speaker 01

  Where if you are a future civilization, um, what is the period you'd really want to simulate? Um, maybe you want to, I think you just like want to run a lot of experiments on developing smarter intelligences and see, you know, how does that go? Cause it's like the most important thing that will ever happen. Maybe it's relevant to a separate question that they are asking themselves about, like maybe they're building like a level above and they're like, will this go well? Let's run the experiment. Um, and so if you were going to simulate the past, this seems like the point of time which you'd want to simulate. Um, and maybe that's, um, adds additional credence to the idea that we're part of this forecasting, uh, backtesting that the, uh, the AIs will want to do, uh, already. Um, but yeah, I think counterfactual history will just be, yeah, such

• 1:00:20Speaker 01

  a improved field as a result of this. Um, we already have, and it's not even AI. Have you seen the, I'm sure you saw Nat Friedman's, uh, and Luke Ferrator's, uh, Herculeanum squirrel thing?

• 1:00:30Speaker 00

  No.

• 1:00:31Speaker 01

  Oh, oh, I think you'll love this. So 79 AD, um, the mountain of Subius erupts and there's this library in Herculeanum.

• 1:00:39Speaker 00

  Oh, they just found it, right?

• 1:00:40Speaker 01

  In Rome.

• 1:00:41Speaker 00

  Yeah.

• 1:00:42Speaker 01

  And so, I mean, we've known about it for hundreds of years. We've had the burnt squirrels and they CT scanned it and then they launched this competition of using modern, modern AI techniques to try to visually, try to process the CT scans and like read the burnt squirrels. And actually, one of the guys who's running Doge, this, you know, 21 year old, uh, super genius, um, he listened to my podcast with Nat where we announced the prize and he then got into, got interested in working on the prize and he solved it. Um, and he, and so he figured out how to, um, decode these scrolls. And so we're going to learn a lot more about the past, just from like discovering knowledge like this with the new techniques.

• 1:01:19Speaker 00

  It's very cool. You know, you said he's super smart or the phrase is, he could be an AI already. We don't know about, it could be developed in a basement in China somewhere and he's been rolled out his, uh, never mind. Um, I'm not even gonna go there. Uh, you mentioned that you take a week or two to prep for an interview. Um, how would you describe your preparation? And I'm sure it varies by different guests, but describe what you do. What's your goal?

• 1:01:46Speaker 01

  The first house I sort of I go into this by picking a guest which merits that that depth and if I'm going to be spending a lot of sort of this is my life, right? I spend my life just reading and researching in advance of interviews. I just want to pick the guest who will is interested like I want I'm excited to spend two weeks reading their books and papers and everything. And often that means turning down guests who are much more bigger names or more influential or something just because maybe talking to you for two hours, not you, but talking to somebody in general and for two hours is might be interesting, but is it are there two weeks worth of prep? And then the prep itself, I'd be curious about your process, but for me, it's like

• 1:02:29Speaker 01

  just the obvious thing of read their books, read their papers, talk to their colleagues, read the adjacent literature in their field. And it depends on the field, right? So AI is different than some more codified field like economics, but more or less do the obvious thing, but do it long enough. Yeah. What's your process?

• 1:02:49Speaker 00

  Well, we're the opposite. I want as many interviews as I can find that I don't have to read for two weeks because I have a day job running a college or

• 1:02:58Speaker 01

  That's right.

• 1:02:59Speaker 00

  Being the president of a college, I don't run it. It's impossible, but I try to help it. So I tried when I have a guest with a book, I try to read the whole book and I say, I try to read every word, but if I can't read every word, I'm going to read every page. So I'm going to skim a page, which I've gotten very good at and figure out what are the key ideas on this page to make sure I don't ask a stupid question or treat a guest unfairly because, oh yeah, they talked about that and jumped six and what I didn't when I missed it.

• 1:03:28Speaker 01

  Yeah.

• 1:03:29Speaker 00

  And because I'm busy, well, two things have happened. One, I'm much better at that than I was 15, 20 years ago. And when the podcast started and I'm also looking for guests, I don't have to read a book because it's hard. But I think the more interesting prep time that I don't spend, I'm going to tell you this, I'll give you this example and you can think about your own practice.

• 1:03:55Speaker 00

  I've gotten very good at reading that book and I can develop the questions to fill an hour plus in really about 10 minutes. You know, I don't have to spend a lot of time thinking about what the key questions are. And it's more than just what are interesting questions. I try to have something of a narrative arc over the course of the of the episode so that the listener can hold it together, right? It's not just, oh, here's some entertainment. Here's some stimulus, mental stimulus, and this is interesting and that's interesting. I try to help them understand the the argument of the book. And sometimes I'm doing that in a way that the author didn't do. I might have a different conception of it. And so, but I can do that very quickly. What I don't spend enough time doing is thinking, how

• 1:04:37Speaker 00

  do I get better at this? What we can talk about this in a minute. But how do I I mean, I've got really used to this because I've done almost 1,000 now. And they're better than they were. But they could be a lot better, maybe. And how do I think about getting better? And I don't do enough of that. So it's not a big part of my prep. What's the key thing in this interview coming up that I have to nail to take it up a notch?

• 1:05:04Speaker 01

  Hmm. And you mean more from the raw content or the distribution or or both or something else?

• 1:05:15Speaker 00

  It's not the raw content. It's not the just you say the distribution?

• 1:05:19Speaker 01

  Yeah, as in, you know, reach on platforms and things like that.

• 1:05:23Speaker 00

  No, what I'm talking about is our rapport. So

• 1:05:26Speaker 01

  That's right.

• 1:05:27Speaker 00

  You and I have been having a good time. I've teased you a few times. You took it well. You didn't get offended. And it's conceivable that because of that, either in the next 10 or 15 minutes as we continue the conversation, or if I invite you back, we're going to have a more transc literally a more transcendent conversation than we could otherwise have. One of the challenges is, is we're doing this over Zoom. And people say, oh, that must be hard. Uh, or I used to do more of the phone. And they'd say, isn't that, and that's really hard. You can't see the other person. But a lot of guests are shy. And they don't want to look at me in the eye. And it makes them uncomfortable. And so face to face can actually be more awkward. There can be less rapport.

• 1:06:07Speaker 00

  So, you know, what I want to get out of you is an unsafe, ideally, an unsafe response. I don't want to fool you. I don't want to surprise you. I don't want to hurt your feelings. I don't want to cause you to regret what you said. But I want to unlock something that

• 1:06:23Speaker 01

  That's right.

• 1:06:24Speaker 00

  You don't, haven't written before. And you haven't revealed to anyone else in a public podcast, say. And we've had a couple of those, I think, for me. I mean, I haven't read everything you've written, but I thought this had a spontaneity and a chemistry to it that is, sometimes doesn't happen. So I don't think enough about how to make sure that happens. Maybe you can't. Maybe it's just luck. Maybe it's, I don't know.

• 1:06:45Speaker 01

  I mean, one of the things I love about your podcasts is that you do this. Um, and so maybe, I, I think that's sort of like what makes Econ Talk special. I've been thinking about, um, you know, you have this intuition, at least I have this intuition when I do interviews, like, oh, maybe they'll find this sensitive or whatever. But, um, especially if they're big names, like whenever you have, go to like these dinner parties with, whether it's influential people or scholars or whatever, between themselves, they're like constantly teasing each other. And they're like, um, didn't you totally mess up that, uh, business thing? Or, well, I think I did, like, this idea is totally wrong. And here's why I think that, right? They're constantly, um, uh, doing this to each other.

• 1:07:28Speaker 01

  And this is the sort of relationship that they expect, that they find fun, right? That's why they go to, that's why they have this rapport with their friends and colleagues and business partners or, um, uh, so forth. And so what I've been trying to do is like, if I was at dinner with them and I had this relationship with them, how would I have this conversation with them? And part of that is, um, maybe making it less salient that this is an interview. And I think you do a wonderful job of this of like, if we were actually at dinner, um, it was just you and me, what would the conversation be like? And it would just be, in many, so there is this thing that happens where people, um, Interviewers will have the sense of like, I don't want to know, or I don't care about the answer to this, or it's like, you know, the answer is obvious, but I have to ask this for the audience. So the audience can keep up or something.

• 1:08:13Speaker 01

  And whenever they do that, it's quite obvious, and at least me as part of the audience, I'm like, I just like, if you didn't find that interesting, why would you think that I would find it interesting? Right? Why don't you just ask the thing that you actually want to follow up on? Or, and you can, even if it's like, even if I don't personally find it interesting, the fact that you find it interesting and you're like, more active and you're like, this is the thing, this is the crux, I want to know more about this. It just would make the whole thing more lively and more fun to follow along with.

• 1:08:40Speaker 00

  You know, great journalists, and I know a couple, not, I'm not lucky to know more, a lot of them, but I know a few great journalists, and this is a skill, right? It's a, first, it's putting people at ease. It's getting them to enter into a space of, it's just you and me having this conversation, no one else is listening.

• 1:09:02Speaker 00

  And it's asking a, the real art, and this happens very rarely for me, but I think a good journalist or a spy can do this very well, is asking, knowing when to ask the revealing question, that if you started the conversation with that, the person would totally freeze up, put up walls that would never, would be guarded for the whole rest of the time. But after enough time has passed, and again, this could be the third interview, you know, three years from now, or the fifth, you know, that matters, but ideally, within this interview, you've gotten them to trust you, and you've put them enough at ease that they feel safe telling you something,

• 1:09:44Speaker 00

  either because they've forgotten that there's a few hundred thousand people listening, or they don't care anymore. And I, and for me,

• 1:09:57Speaker 00

  It's the "don't care anymore" part that's really rewarding. When you realize someone just wants to share something with you, or vice versa. "I want to share something." By the way, the other thing I love, for me, a successful interview is, "I have a new thought." That's very rare. We had a couple here, for me, already. That's all gravy.

• 1:10:15Speaker 01

  Yeah, yeah.

• 1:10:19Speaker 01

  I mean, it's a thing that I probably don't spend enough time thinking about. Like, because I mean, I've spent a week prepping for the interview. And then, this, maybe I should game plan more of, "What is the first sort of joke I'll make to put them at ease before the cameras start rolling? What's this, how will we transition or something?" But I do, just like, just the general vibe of, "A colleague, if you're at a dinner together, would kind of hassle you in a very teasy way." I mean, the key thing is to make sure that it's a vibe of, "Not a reporter who's like, trying to get you for some naughty thing that you might have done or something."

• 1:10:56Speaker 00

  Exactly, not sandbagging.

• 1:10:58Speaker 01

  Exactly, but, somebody you respect, a colleague, will just have like, you know, they won't just like, ask you to ramble, they'll be like, "Wait, this doesn't make sense to me," or, "Yeah, that sounds wrong." So just having that sort of like, teasy, friendly attitude, I think. And often, these people are expecting this, and often they, because this is their life, right? They're not expecting to go unchallenged, and they enjoy it more. It's like, it's more, "I had to be away for that." Makes for a better conversation, too.

• 1:11:32Speaker 00

  Yeah, it also has implications for who you choose as your guests. I'm sure you've experienced it, sometimes you invite someone on the program, they don't like to talk. Or they're not comfortable talking, and they're not good at talking. They're really good at writing, and you made a mistake, you invited them, and you find yourself talking more than you wanted to. I've had guests, I won't name them, but, they didn't know I was an economist. They didn't do any homework, they didn't do any prep. They just got invited, they don't get invited very much, so they said yes, and then, "They're treating me like I'm a reporter, doing a, you know, podcast," which is, actually tickles me and amuses me, but, I think the kind of rapport you're talking about, what I think of as...

• 1:12:14Speaker 00

  I think listeners love to listen in. And that's why your example of explaining it for the audience like I'm somehow, I don't want to hear that. I want to be the fly on the wall when two interesting people are sharing their thoughts. And sharing is better than interviewing, right? Sharing is much better than, I've got a list of questions, I just go down them one by one, and when we're done, I say, "Thank you, Dorash, for being part of Econ Talk." When in fact, what's really fun is that two people who respect each other, as I think we do, have a more peer-to-peer conversation than this famous person I'm trying to interview who's way above me professionally and intellectually and so on.

• 1:12:59Speaker 01

  And can I just say one thing, which is that when I look at my most popular guests, you would think it's like, oh, it's the famous people, or something like that. So there is this risk, as you mentioned, where somebody might not be that good at talking, and you read their book and you're super interested in it, but then the podcast didn't turn out well. That's actually very rarely happened to me. I can't even think of a single case that might have happened. It's so much more common, if I look at my most popular guests, I've interviewed people like Mark Zuckerberg, and Satya Nadella, and Demis Hassabis, and Dario, and Ilya Sutskever. They are not my most popular guests. My most popular guest is Sarah Payne.

• 1:13:41Speaker 01

  She was a, before I interviewed her, she was not a publicly well-known person at all. She's a military historian, has written about every single conflict of the 20th century and others. I loved, her books were amazing military history. And as a result, you know, her views, her videos have like 3 million, 2 million, I did four separate ones with her, 1 million, 2 million, 3 million views each. And it was just the, the other example is David Reich, who is then my second most popular guest, has like 1.5 million views. And again, this is like, I write, I liked his book about ancient genetics, I had no idea how it would be as a presenter. And it might not even be clear if he's like a quote-unquote "good presenter" in the sort of traditional sense, but he understands the ideas, he can like really riff on the

• 1:14:25Speaker 01

  details of these interesting models. And once you get through those, then you get to the Satya Nadellas and the Mark Zuckerbergs. And this is also a reason why I'm...

• 1:14:36Speaker 01

  You can be a little more, not confrontational, but a little more cavalier when you have these big names on and be like, this is, this will still be a real interview because at the end of the day, you know, it's, it's, what moves a ball forward for me, both from the perspective of the audience and what makes me enjoy my job, is having interesting conversations that I enjoy. And so we're going to do that regardless of who the guest is because who the guest is like, just doesn't reliably end up mattering.

• 1:15:04Speaker 00

  Great, great minds think alike. My next question was about Sarah Payne. So I'm looking into you because I get your book and I see this interview with Sarah Payne. I've never heard of her. And then I see it's got a lot of views. And then I watched 30 seconds of it's like, this is fantastic. Half of the fantastic is her delivery is just so authoritative. It's like, obviously it's the truth and, which it's not, but it could be. And the second part is she's, she knows a lot and she's talking about a lot of interesting things. And I also looked, found an article about you where it mentions that because you interviewed

• 1:15:44Speaker 00

  Sarah Payne, Noah Smith interviewed her. And I'm thinking, I'm thinking about that going, yeah, I was thinking about that and I might still. How did you find her? And how do you find people like that more generally? Because I think the other thing that, part of the reason she's popular is the same reason that the hole in the wall restaurant's popular rather than the famous one that everyone knows. Everyone likes the idea that there's this really, there's a brilliant person that you only know about. Most people don't. You know, I used to feel that way about Scott Alexander. Everyone's caught on. But when I first read him, a zillion years ago, it was like, oh my gosh, this guy is so interesting. And so you find that and it's exhilarating. It's so fun. How'd you find Sarah Payne and how do you look for her? And people like her.

• 1:16:27Speaker 01

  I had this phase where I got really interested in World War II. You know, my interests are like 30 years older than I am or something. So I asked on Twitter, who should I interview about World War II? And then Tanner Greer, who is this great blogger, suggested Sarah Payne. And he said, she's not famous, but she has been to all the archives on all the continents. She's been to the archives in China, in Russia, Japan, just name it. She's done the work. She's done the, she's got the deepest models. And so I started reading her books as a result. And, you know, I don't know, maybe had like five Amazon reviews or something, just like was not...

• 1:17:09Speaker 01

  And I couldn't even find that interviews of her because she was just didn't have a public profile. Now, what I learned afterwards, and she will never say this publicly, but it's true, is if you go on Amazon and you look up her book, the author is S C M Payne, not Sarah Payne. She's in her 60s, I'm not sure, but maybe 50s or something. When she was coming up in military history, I don't think saying Sarah Payne instead of S C M Payne, was it helpful? And so there's a question of, well, why is this such a compelling presenter? Somebody who has deeply studied all these different conflicts. Why are they not already famous? And you know, after I interviewed her, she went on Time, she got a profiled in the Time Magazine, Noah, obviously, many others.

• 1:17:52Speaker 01

  And maybe, I think the answer might just be that, you know, like military history, at least until recently, was not a field where female scholars were recognized to the extent they should have been.

• 1:18:06Speaker 00

  Yeah, it's an amazing thing for me. And I talk about it sometimes with respect to economics and the rarity of public economists in the past and how many there are now and because of technology, because of podcasting and other things, blogging. But just, the world's full of interesting people. And, and one of the advantages of being an academic, and you're not, right? But I, one of the things that I liked about academics was being a professor is that you get to spend your day with smart, interesting people who tell you things you haven't thought of. So if you're a curious person, academic life has an appeal. Now, it's overrated, big time. And it's, and it's less true now probably than it was before because it's gotten even narrower

• 1:18:50Speaker 00

  and less, less, uh, interesting. But one of the reasons Sarah Payne is so attractive, and I'm using her as an archetype, is you can tell she's not trying to be famous. She's not selling herself. She's not trying to build a brand. She's just a thoughtful person and who's read a lot and knows a lot. Like talking to Claude, but, uh, in the flesh. So it, it is a, there's something special about authenticity and about a disdain or clearly a lack of interest in self-promotion that, that also makes her, makes her attractive. But I do think what makes it fun for you and me to do what we do is the chance to

• 1:19:34Speaker 00

  talk to really smart people and ask them questions. It's an extraordinary blessing. What a privilege.

• 1:19:39Speaker 01

  Yes, yes. Can I ask you a question?

• 1:19:41Speaker 00

  Yeah, sure.

• 1:19:42Speaker 01

  So you say, look, if people want this kind of lifestyle where people can explore different kinds of ideas, um, meet interesting people, academic life can provide that. Um, I'm curious what you make of this. But part of me wonders whether more people could, um, do the kind of, do this kind of thing independently. And so they, you know, nothing is preventing you from starting your own blog, your own podcast. I mean, podcasts especially is helpful and kickstarting here because instead of having to come up with your own worldview from scratch, you can first interview a bunch of people who have just thought about these topics much longer and in prepping for them and just working with them, you can build up this repertoire of conceptual tools.

• 1:20:25Speaker 01

  But you can just independently do this. And I often give this advice and one of my friends, um, teases me. He's like, you're like, you're like a country bumpkin who won the lottery. And you're like, you're telling every, every person you know, like, you gotta get a scratch pad. You know, buy as many as you can.

• 1:20:44Speaker 01

  But, but, um, maybe, yeah, my sense is that this is a field where it's, it's not some, like, I'm, I'm gonna become, uh, America's Got Talent or something. I think this is like a viable career path. And it's obviously incredibly fun. Do, would you disagree with that advice that you should like, consider being an independent? High, content creator is a terrible term, but you see what I mean?

• 1:21:07Speaker 00

  Yeah.

• 1:21:08Speaker 01

  A podcast blogger.

• 1:21:09Speaker 00

  Well, yeah, I don't know if you've heard my interviews with Adam Astriani, but he's somebody who is a, you know, very thoughtful person. He's a trained psychologist, had a normal academic career and decided he'd just not do that. He'd still do research, he'd still write about the things he cared about. He's gonna do it on Substack and see if he can make a living that way. I don't think, I think that talent is scarce. That's my first thought. We won't name names, but there are podcasters who aren't very, don't seem so smart or curious. And some are thriving, but on average, it's a hard thing to do.

• 1:21:47Speaker 00

  And it's a hard thing to do regularly. Most people don't want to sit and read for a week or two before interviewing somebody.

• 1:21:54Speaker 01

  Right.

• 1:21:54Speaker 00

  So I think it selects people who are really good at, at talking, obviously, but also good at absorbing information and also good at this weird skill called interviewing. That is, uh, looks really easy. And I think is, I'd be curious if you think it's as hard as I, I think it's incredibly hard to do. Either because I'm not very good at it or it's just hard. So I don't think it's gonna be, uh, I'm not convinced it's gonna be a, a way for many, many people to, to pursue an intellectual life. What I think is available is listening to podcasts. And one of the things that, you know, is not very well, surprisingly not. Happening is aggregating podcasts in any kind of way.

• 1:22:38Speaker 00

  And people have thought about it, worked on it, talked about it. But, you know, maybe it's not even very necessary. Maybe that's the reason it hasn't happened. You can still, just by Googling or using AI, you can certainly find out so much extraordinary material to teach yourself stuff and to amuse yourself and to grow intellectually as you walk your dog, exercise, do the dishes and so on. Do you think it's hard?

• 1:23:05Speaker 01

  Yeah.

• 1:23:05Speaker 00

  Do you think it's hard? Is it hard for you?

• 1:23:08Speaker 01

  Um, it is hard for me. I do think, if I'm being honest, I have friends who have better things to do with their life, maybe. But, uh, I mean, you are like, you're running a college, right? And so if you, I mean, many other people in your position would not make the time to also spread this content and this knowledge to the world in the form of a podcast. And they would just be doing their, the thing that, you know, their deja vu is exciting them. And I have other friends like this who are like, who I think would also have good podcasts if they didn't have better things to do with their time. I fortunately or unfortunately don't. And so, um, but I, my sense is that when I go to dinner with them

• 1:23:52Speaker 01

  or something, I'm like, and especially when we invite somebody who's like in a different field and they start asking them questions, I'm like, yeah, if you, if you got into this field, I'd be worried for my job. And I get that feeling more than, um, more than a little. I mean, on the point about, well, it is the case that many people, you should, this shouldn't be your sort of, I'll keep doing this for five years, even if it doesn't work. Cause it is the case that there is a skill to it and you, um, maybe you lack it. I will say, look, we tell people to start startups and we think that's a sort of totally viable and praise worthy thing to try with your career at some point. And if it doesn't work, I mean, like 1% of them succeed. If it doesn't work, it didn't work. You can do something else afterwards. But it's worth trying in case it succeeds.

• 1:24:38Speaker 01

  This is such a fun job that I feel like having it actually become a career is the equivalent of having a successful startup or something like that.

• 1:24:46Speaker 00

  Sure.

• 1:24:47Speaker 01

  In terms of how much I value it. And if we think it's viable for like you to take a, you know, 1% shot at starting a startup, and if it doesn't succeed, you can do something else. I just think it's like sort of maybe should be a thing where, yes, go in knowing that it's a risk, and it often doesn't work out, and there's power laws. But give it a real chance. Like, spend three months making it your main thing. You don't have to quit your job, you don't have to do anything irresponsible, but like, actually try. At least I feel lucky that people gave me that encouragement early on. Otherwise, I would have just been the kind of software engineer that might be automated to, and I don't want to be cavalier about it, but, and did this instead.

• 1:25:28Speaker 00

  But that's because you're really smart. I'm guessing, I don't know you very well, but I'm guessing it's because you're really smart and you can observe a lot of information fairly quickly in that one to two week period to do a good job. So I'm not sure it's good advice in general. For a certain kind of person who has different alternatives, it might be good advice. What's hard about it for you?

• 1:25:56Speaker 01

  Finding the main thing. Finding, with any guess, you could talk about a million different things, but there's often a key crux where if, not only should you press on it, but you should keep pressing on it until you have totally exhausted any potential confusion or any chance in which you think they're not addressing your question or something. Because a lot of ideas do have this, like, I don't know, if a Marxist came on your podcast and you didn't just like hassle them a little bit about like this like key foible in their worldview, then it would not be a successful interview.

• 1:26:37Speaker 01

  And so maybe just like, even as much as I try to emphasize the main thing, I probably underemphasize just really hammering on it.

• 1:26:45Speaker 00

  Yeah, there's nothing worse than you get to the end of the interview, you say, you think, that was really good, then you look at your notes and you realize, I forgot to ask him the one key thing. It's still a good interview, it's just there was one thing that would have made a great interview.

• 1:26:58Speaker 01

  What is it for you, by the way?

• 1:27:04Speaker 00

  For me...

• 1:27:08Speaker 00

  It's a few things. I try to keep them to about an hour. So, and I'm now, we're about an hour and 28 minutes of this, and we're gonna go a little bit longer and, but when it's, when I used to try to keep it to really right around an hour, I'd be very wary and conscious of the time. And so I'm balancing, listening to the argument, trying to figure out whether I should ask a follow-up question, a challenging question, trying to think, should I skip to the next topic, looking at the time, and keep my own thoughts straight. And, you know, I, I think listeners, some listeners think I talk too much. I've talked too much in this interview, which is interesting, because I think I've done so many AI interviews. I think I have something to say about it.

• 1:27:52Speaker 00

  My mind's just ran for a few different times. But the hard part is knowing how much of that to do, and how much of it is me. And if it's none of it's me, like, some people, some guests, most guests, I want them to shine. I want, I'm trying to bring out their insights. And then, but then, oh, I have so much to say too. And the answer is, of course, sometimes you don't. And you should just shut up and let the guests talk, as some of my listeners complain. So that's one, that's what's, those, those trade-offs, and balancing those things are what's, are what's hard for me. And as you get older, like, if you've had a long day, like I have, it's 7:30 here in Jerusalem, I forget, I lose my train of thought, much

• 1:28:36Speaker 00

  more than I did when I was younger. And I'll sometimes stop and say, well, we'll edit this out. And listeners don't hear that, because we edited it out. But that's hard too. And it's, it's kind of sad. But anyway.

• 1:28:49Speaker 01

  I mean, there are a lot of scenarios where I, in retrospect, I'm just like, gosh, you know, I miss asking about that, or there's an obvious follow-up here, or something. And maybe it's something that's sort of harder to appreciate when you're just listening to it, is, as an interviewer, how many things you're sort of, you were like, thinking about how much time is left, and what is the other topics you want to cover, and what is the follow-up to their specific thing they just said, and how does, what are the transitions, and so forth. And so, often in all that froth, the, certain things can get missed. I mean, I would say, with respect to that particular thing that people point out, As compared to your audience of hundreds of thousands of people,

• 1:29:32Speaker 01

  there is a failure mode of updating too much on what a couple of people who send emails say. And I personally, as a fan of Econ Talk, actually really appreciate that you have back and forths with your guests. Because these people have done other podcasts, right? So if you just want their raw spiel, you can get it somewhere else.

• 1:29:52Speaker 01

  Yeah, keep the Russ style, I would say.

• 1:29:56Speaker 00

  That's sweet. What's your favorite thing about it?

• 1:30:00Speaker 00

  How many have you done, by the way?

• 1:30:01Speaker 01

  Have you done a talk?

• 1:30:01Speaker 00

  No, no, no, about yours, about being a podcaster. No, yeah, tell me more. Enough about me. What do you think of me? No, I meant first, how many episodes have you done? How many interviews do you think you've done? Do you know?

• 1:30:15Speaker 01

  Probably close to 100.

• 1:30:16Speaker 00

  Okay, so what's your favorite thing so far? Not favorite guest, I hate that question. What's your favorite thing about doing 100 interviews?

• 1:30:28Speaker 01

  The fact that it gives me the opportunity offline to meet incredibly interesting people who teach me a lot. There's different ways in which you can model why the podcast improves over time or grows over time. And I actually kind of believe that slow compounding growth is fake. For maybe a reason, basically, I think good stuff actually does boil up pretty fast.

• 1:30:54Speaker 01

  And it's not even that, like, just doing the process itself has made me that much better. Because you can keep doing the same thing again and again. I think one of the lessons in life is that you keep doing the same thing again and again in life. And if you're not getting any sort of signal from it, you know, you won't improve. The main way in which I've improved, and the main thing I got out of the podcast is, hopefully, I do a decent enough job on a topic. And that gives me the chance to reach out or be reached out to by some practitioners in the field who can teach me more. We go out to dinner, coffee, and so forth. That helps me do a better podcast on the topic, helps me meet smarter people, dot, dot, dot. And that, especially in AI, has been the main source of. I mean, I'm genuinely just, you're very kind to suggest.

• 1:31:37Speaker 01

  But I know in college, I was like in the middle of my class or something. I'm I'm I'm not trying to be humble or something, but I do know that like a lot of like the way in which I'm able to ask maybe good questions about AI or any other topic is that I am surrounded in San Francisco by this hive mind. I mean, going back to the value of the hive mind. And the key is to try to replicate that in other fields as well.

• 1:32:02Speaker 00

  You see, you're in the middle of your class, but that was in a class called computer science. This is a different class. This is the class called curiosity. This is the class called synthesizing a bunch of different things. It's a very different skill. So I appreciate your humility, but I'm not sure it's deserved. But it's nice to say. Let's close with curiosity. In prepping for our conversation, I realized it must be a cliche. Everyone must know it already. But what I find is that the more interviews I do, and the more interesting people I get to spend time with, my curiosity does not get sated. It gets stimulated. And

• 1:32:47Speaker 00

  what a strange thing that is, right? Like, I want to know some stuff. Well, okay, here's some stuff. Now you know a lot more than you did before. You don't say, Well, yeah, I think I'm close to knowing enough. It doesn't. You want to go deeper and deeper. And that, to me, is I suspect it's genetic for me, but I think I've fed it over my lifetime in a way, because I haven't been much of a specialist. And I did a lot of reading outside of economics all my life. And, you know, it's bad advice. You're talking about bad advice. Here's some bad advice. Yeah, read whatever you want. Don't pay attention to school. Books aren't important, and it doesn't really matter. Just do what you care about. And whatever your passion is, pursue it. Yeah, well, if it doesn't pay, it's usually a problem. So

• 1:33:32Speaker 00

  it's not the only thing, but it's not irrelevant. But for me, the great thing is, and I don't know if this will last much longer, I'm 70 years old, but I might be as curious now as I've ever been. I'm not as good at digging into that central limit theorem. I'm not as curious about that. But I care, I find lots of things exciting to discover. And I think it's a gift, but it's a gift you can, I think it's a gift you can build by exploring different things. Or maybe it's just genetic and you're stuck with it, with whatever level you're at. I don't know.

• 1:34:09Speaker 01

  Well, the, hopefully I'm getting to that point, but you've definitely had the opportunity and do this is that if you've done a thousand interviews and obviously are, have done your own research in the past and so forth, you are aware of the contradictions, the open questions, the different hypotheses, the mysteries in so many different fields. And that is such a rewarding experience because when you're, when you're a kid or something, you can maybe ask a question and you're curious, but it's a curiosity that's very often responded to in very cavalier, you know, why is the sky blue? Oh, there's this explanation or whatever.

• 1:34:50Speaker 01

  And then over time, I've gone to the, with some of these conversations, you start talking to these scholars and you're like, nobody knows this. And they have a hypothesis. I mean, it's similar to the point you made about Dario saying it's an empirical trend. We don't know why it works. Especially in AI, there's so many questions like this, but even outside of AI, like David Reich, I obsessed about the interview for months after I did it because what I learned is we did, a lot of the things we thought we knew about human evolution are just like, we don't actually, we don't know them. We don't know where it happened, like how much of it was the human groups from the antarThals outside of Europe mixing back into the human gene pool in Africa. We don't know when it happened exactly.

• 1:35:29Speaker 01

  We don't know how it happened, like what were the sort of mutational or cultural changes? And just this total shock and the open questions that I had, which some of which even David's lab has not yet discovered the answers to. And getting to do, I mean, the world is so endlessly fascinating and our job is just to find the smartest person in any given field and then just pepper them with our open, why this, why that? How come not this? Why do the, how do these ideas connect? And that's our job. And it's, I'm incredibly delighted about that.

• 1:36:04Speaker 00

  My guest today has been Dorash Patel. Dorash, thanks for being part of Econ Talk.

• 1:36:10Speaker 01

  Thanks for us.

• 1:36:17Speaker 00

  This is Econ Talk, part of the Library of Economics and Liberty. For more Econ Talk, go to EconTalk.org, where you can also comment on today's podcast and find links and readings related to today's conversation. The sound engineer for Econ Talk is Rich Goyette. I'm your host, Russ Roberts. Thanks for listening. Talk to you on Monday.