Nobel Prize Shocker & Economics as a Science w/ Noah Smith

Hey, Upstream listeners. This week, we're releasing a conversation from Turpentine's show, Econ 102. Noah Smith and I talk about the 2024 Economics Nobel Prize, what it means for economics, how to think about the economics of AI, and more. If you like this episode and want more from Noah, please check out Econ 102 linked in the description. Please enjoy the episode.

Hey, how's it going?

Hey, it's going great.

I'm here in New York. How are you doing?

Oh, awesome. What are you up to in New York?

I am at 2 conferences. There was a media conference yesterday and a venture capital conference tomorrow.

Oh, cool.

Yeah, I try to go every couple months, every 3 months, at least every quarter, see some friends. You having a good week so far?

Oh yeah, yeah. All right, cool.

How do we pronounce the the guy who just won the Nobel Prize and who wrote Why Nations Failed? Was that how to pronounce the last name?

Asimoglu.

Asimoglu.

I mean, I think the Turkish pronunciation is more like Atçamoglu, but like, I think he just goes by Asimoglu. He's not born in America.

Okay, I want to get into his book, and I want to get into his thoughts on AI because they touch upon another piece you wrote recently. But First, maybe we just zoom out. You had this interesting reflection on the Nobel and how sort of the prize in economics, it signified a little bit of a change this year where you wrote about how it was, we were moving to this idea that the field is something of a natural science that has sort of, you know, can be measured empirically and has more practical applications, but has sort of moved to back what it used to be, which is more around who has more cultural influence in the field. Separate from perhaps its real-world impact. Is that how you see it? Or when you talk about broad reflections on the—

Right. So, so traditionally, you know, if you give a physics prize, it's because you discovered a real thing in physics that works. Either you made an invention that works, like, you know, I don't know, fiber optic cable or something, or you make a— or this year they gave it for AI, which is kind of, I don't know, a little weird, but like, but you know, AI is a tool. It's a, it's a digital tool. You know, but a fiber optic cable is a physical tool. So, so then, but the thing is that if you made a theory that didn't, you know, that didn't actually match observations, that didn't, you know, explain data, like, that you couldn't actually use, you wouldn't win a Nobel Prize for that. And you, you still won't. So like, you'll notice that no one has ever won the Nobel Prize for string theory. There's a reason for that, because no one's ever managed to find evidence for string theory. And, and so Like Ed Witten is a, is a fricking genius and has done, you know, arguably more work than anyone else in the string theory field. And he won a Fields Medal for it. Like he invented new, important new math just to do string theory and actually won the most prestigious prize in math, which is actually harder to get than the Nobel Prize, statistically speaking. Plus you have to do it before 40. And so, and yet he hasn't won a Nobel Prize and he never will because, you know, String theory doesn't work, or at least yet, maybe someday it will, but, but no one's ever found a way to test it, to match it to experiment. So that's how physics prize works in chemistry, biology, similar stuff. But I'm, I mean, medicine, medicine prize. But in economics, if you look at who gets the prize, it was people with influential ideas within the economics profession. So that could be anyone from people like Hayek, you know, who, who just, you know, his stuff was all written down in sort of like expansive discourses of words. Today he'd be a blogger. Right. But then, you know, he didn't do math, but then so, so people like that. And then, then, you know, later on it was often given to people who made theories that could never really be proven or, or which theories which were just like dead on arrival because they didn't just didn't match anything from the data. So like in 2004, they give it to this, this theory called real business cycle theory and, or RBC. And what RBC says is that recessions and booms are due to changes in the rate of innovation. Technological innovation. And, or, you know, you could have policies like taxes, like self-inflicted stuff, but primarily it's just like this year our engineers discovered more stuff and that's why economies get better and worse. And then when people, when suddenly you have this giant flood of mass unemployment, it's because, you know, we discovered things at a slower rate. And so people's, you know, wages won't grow as fast. And so people are like, you know what, I'll take some vacation by going, by just quitting my job. I'll just take some years off of work. That is not a very good match to anything like that. You know, you can just show all kinds of data showing that this is full of crap from data on, you know, the rate of productivity growth to the, you know, data on people quitting their jobs or just, you know, all kinds of things like this is, this is not a good model for how the economy works, but it was very influential. And the math that they used to do it was not very hard math, but it was math that people liked because it was hard enough to, you know, to sort of, well, I could speculate on why people liked it. But anyway, the point is that that math became very popular in macroeconomic modeling. That math is something we call dynamic stochastic general equilibrium or DSGE, if you've ever seen that name.

Now we're getting to the good stuff.

Right. So 2004, we got the, we got the prize for that. And that's how the Nobel economics Nobel was for a very long time. And so But in recent years, I would say the last, the last 10 years and especially the last like 6 or 7 years, we've seen a shift and we saw a shift toward empirical stuff. And now this shift follows at a lag, the, a big shift in the economics profession. So around the 1990s, but really starting in the 2000s, economics started to get much, much more empirical. Data became much more and statistics became much more of a thing. Whereas, you know, theory alone sort of declined rapidly. And, you know, this is all because of computerization, right? We got better tools. And so when you, when you get computerization, when you can, when running a regression doesn't require like, you know, getting at your paper spreadsheets and then doing a whole bunch of calculations and blah, blah, blah, blah, blah for some, for one simple little regression where you can just run a million regressions with like a little script that you make, then, you know, the, the cost of doing empirical work fell. And so people started doing data stuff and lo and behold, a lot of the data stuff they found really contradicted a lot of the leading theory. And so I, and at this, however, at the same time, you had a shift in theory, whereas people who were working on sort of fundamental theory, the actually difficult math stuff, they largely shifted from doing macro models where you'd model like the whole economy, you know, to, to doing micro things about how like agents interact in an auction. And I'm not, I'm not talking about shifts of specific people, although there were a few of those. I'm talking about shifts in just like, if you're a smart person coming up in the econ world who does math, what kind of theory do you work on? And there, there was this shift, I think, from general equilibrium, which you can see declining a lot in the, in the number of papers that, that use it, and an increase in game theory, you know, Nash equilibrium and things like that, strategic interactions instead of This sort of like, let's throw everything together into a big imaginary economy and see what pops out. So there was this shift toward game theory. And one thing is that that shift made economic theory much more practical. So you saw, for example, Google is entirely powered with auctions designed by a leading auction theorist, Hal Varian. And so that's how they use these auctions to predict how much people will pay for ads. And that's incredibly useful to Google's business model. And so, you know, lots of online sites, Amazon, eBay, all those people always used like pricing strategies that economists would come up with, like basically theories that you could then test really, you know, well in real time. And of course now with, with, you know, AIs trading against each other, you have, you have new theories. So there was a shift toward that. Plus like, you know, the algorithms that we use to assign kidney donors to recipients, those were the, like economists made that up. Or to match medical students with like their residency places, those matching algorithms that like economists won a Nobel Prize for that. They made that up. And so you saw this, this, and then in all kinds of things like, you know, auctions and competition policy and blah, blah, blah, you see, you've seen a closer correspondence between theory and practice. And then sometimes they still give the prize to macro people, but it's, it's often something that's extremely like currently policy relevant. Right? Like, so for example, Bernanke got it for doing, you know, macro models that involved a financial sector. Well, he was heavily involved in, you know, applying those models in real time to the Great Recession. And so, you know, there was a— but you didn't— you have— what you haven't seen as much in recent years is big think, is like taking on— like attacking these gigantic questions. And you've seen some old guys like, like Lant Pritchett, like Angus Deaton, You've seen old guys criticize the Nobel Prize for not focusing on the big issues, for focusing on people who get, you know, more credible answers to small bore questions like, you know, does this education program help these kids in Ethiopia? Right? You can get a very good, pretty good answer to that. It's actually a lot harder than you think, but you can eventually get a pretty good answer to that. But, or, you know, does minimum wage cause a lot of unemployment or does Does like a big surge of low-skilled immigration actually hurt the job market for native-born people? And so those are all questions that people sort of won Nobel Prizes for asking in recent years. And so then people said, those are too small bore. You need to look about big, you need to look at big, big things. And so I think this prize shows more, it's more about big think. It's the first big think prize in a while. So who got it? Daron Acemoglu, Simon Johnson, and James Robinson. Got it for their theory of institutions, inclusive institutions. The idea that when you have a society whose whose sort of rules and systems are set up to to include the you know the opinions and the talents and the preferences of the vast majority of people, then that country will grow faster over the long run and do better than a country that simply treats people as a resource to be exploited. And that's the Big Think Prize.

And why Asimoglu? I know he's, as you write, he's a beast of a researcher, but what are his great contributions? I know that you've viewed some of his works somewhat skeptically. Obviously, no denying his prolificness, but—

I view a lot of his work very skeptically. I think that it may be related to his prolificness. When you take that many shots on goal, you're going to miss some. But I think he could probably stand to slow down and do some of these things a little more carefully.

But what do you think is his great contribution, basically?

You mean if I had to pick one great contribution by Acemoglu and say this is his big thing? I don't know what I'd pick, but you have to pick if you're going to give a— you know, he dominates top journal publishing. Like, he is the beast. He's the closest thing that econ has to Terence Tao in terms of how much he publishes.

But if he wasn't around, would we have a different understanding? Is there something that would be missing in our understanding of modern economics or something? I mean, I do want to shift to the—

There would certainly be differences in the way economists think about things. So the institution stuff has been incredibly influential within the economics profession and in the public through books like Why Nations Fail that he wrote with Robinson. So that stuff has really, really affected a lot of public discourse. And a ton of economists have made their whole careers, you know, doing research and following this up. Similarly, his work on AI and automation is incredibly influential in, well, it's influential within the public, in the public discussion, but also in, you know, tons of people. If Asimov writes about something, 100 people will follow that up and write papers about that. Maybe not 100, maybe 20. But, but yeah, so that, so in terms of influence, there's been a ton. But in terms of how much we actually know about the world versus if Asimov had decided to go be an electrical engineer, I, I don't know what I'd say. I don't know what I'd pick because I think that the institution stuff, although I like the idea, it's an idea I really like personally, I don't think it's been proven or validated to anything resembling the degree where I can say that we know this or we know that. There's just not much we know about it.

Let's talk about his book, Why Nations Fail, because that's been widely read and impactful, certainly when it came out. I'm curious how it holds up and just to zoom out, there's been a lot of people who've written about why nations succeed and fail. And Peter Zeihan has this— he focuses on demographics, geography, and energy. Someone like Balaji focuses a lot on human capital and technological capacity. And Acemoglu seems to focus on the quality of institutions. And that's kind of a broad term and unclear to what extent culture is wrapped up in or tied into that. How do you kind of evaluate Asimov's claim that he makes in his book.

We'll continue our interview in a moment after a word from our sponsors. Hey listeners, it's Erik. If you're looking for more expert-level conversations, but specifically about top negotiation strategies, then I have the perfect podcast recommendation for you. DealCraft: Insights from Great Negotiators, hosted by Jim Sebanias, Harvard Business School professor and renowned negotiation expert. DealCraft features interviews with the world's greatest dealmakers and diplomats. From former Secretary of State Henry Kissinger to Blackstone's Steve Schwarzman on their most challenging negotiations, from which Jim draws lessons for listeners. Not to mention that The Guardian recently called DealCraft a best podcast pick. New episodes release weekly, and I'd recommend starting with their first episode, which reveals how legendary entertainment attorney John Branca was able to successfully negotiate for Michael Jackson's ownership of his master recordings and the Beatles catalog. Almost unprecedented feats at the time, so don't miss out. Follow DealCraft: Insights from Great Negotiators on Apple Podcasts, Spotify, or wherever you get your podcasts. Hey everyone, Erik here. You know, at Turpentine, we're always looking for ways to streamline our work. That's why we use Notion for all our internal documents. It's been a game-changer for us, and I want to tell you why. Notion is this amazing all-in-one workspace that combines your notes, docs, and projects. But what's really cool is their new Notion AI. It's like having multiple AI tools built right in. You can search, generate, analyze, and chat. All without leaving Notion. What sets Notion AI apart is how it personalizes responses based on your content. It's not just another generic chatbot; it understands the context of your work. Plus, it can search across thousands of docs in seconds, even connecting to tools like Slack and Google Docs. And get this— Notion is used by over half of the Fortune 500 companies. Teams using it send less emails, cancel more meetings, and save time searching for their work. It's designed to keep everyone on the same page while protecting your privacy. Wanna give it a try? Head over to notion.com/upstreampod. That's all lowercase. notion.com/upstreampod. You can try Notion for free, and when you use our link, you're supporting our show. Give the powerful, easy-to-use Notion AI a shot today at notion.com/upstreampod. Right.

So first of all, I think it, it looks right to me in, at least I think it's part of what's going on. Based on casual observation. So you look at North Korea versus South Korea, right? South Korea, you can start a business, you can vote, you can do all these things. And South Korea is very rich. North Korea, you can't do any of that stuff. And so people just spend most of their time trying to like dodge the security services, you know, or whatever. And so like, and you can, so South Korea or North Korea is poor, right? You look at East Germany versus West Germany. Obviously institutions matter in some sense, right? Like they, and they can matter huge. Like you can fuck up your country bad. You can, you can screw it up. You can screw it all up. You can take a good thing and you can just waste it. Right. But whether historically that was the main thing that made some countries get rich before other countries is an open question. I, you know, you know, it can work. And I do believe that inclusive institutions are important based on casual observation of just how things work. I think that, yeah, I think that, yeah. And in terms of technologies too, technologies that, unleash the talents and skills of a ton of people tend to be more successful than technologies that unleash the skills of only a few people. And so I think that's basically right about what's going on in the world, but my thinking that it's right is not science, right? It's just like, obviously there are so many issues that are too big for science right now, social issues like war and peace, like Yeah, there's people who do game theory about that, and you can do some statistical analyses. But I think ultimately, like so much of of that is is beyond current science, and you just have to go with your your you know you can use you can do empirics. You can look at data, but that doesn't make it science, right? That doesn't make it as validated as like Newton's laws of motion or like the germ theory of disease or any of those things, right? You you. you can gather data, you can get like, you know, informed, basically it's what we use is informed guesswork because these things are too big to be put in a lab and studied. And we do the best we can, you know, and that it doesn't mean we should just completely reject empirics and go with our gut. It means that we'll never, our conclusions will never be as reliable as for science. And so I think that that's true in the case of institutions too, because when you look at, so for example, Acemoglu's most famous paper, Acemoglu, Johnson, Robinson, Their most famous paper looked at European colonies. They said, which of these European colonies got rich over the centuries and which didn't? They look at the ones that got rich now, you know, America, Canada, Australia, Singapore, places like that. And they say, okay, those places have what we would call good institutions now. They have property rights, they have rule of law, they have all these things we think are good. Now let's look at the history of those colonies. And what they found is that places where Colonists themselves died usually of tropical diseases, did not get, you know, now are not the places that are rich and not the places with good institutions either. You know, so that would be like Nigeria, right? You can't, if you're European, you can't really go settle much in Nigeria in the 1700s, 1800s, because malaria and a million other things like that will just kill you, right? Tropical diseases will just, you'll, you'll die. You're, you know, you're not set up to resist those diseases. Whereas, you know, Nigerians sort of have Endemic resistance to those diseases, but they you know they they die a lot too. But basically because of that, so Asimoglu and Johnson Robinson said, okay, what happened was when Europeans could move in, they set up property rights and other institutions, you know rule of law for themselves, for their own per for their own benefit. And when they couldn't move in, they they didn't set that up; they just sort of relied on their armies or local cronies to kind of extract mineral resources, and then and then just didn't really you know change the system of government much. That's their explanation. And then some other economists like, like Ed Glaeser came along and said, look, I hate to tell you this, but there's another obvious explanation for why places like America, Canada, and Australia did well and places like Nigeria and Pakistan did not do well. And that's that people moving in bring human capital with them. So if you are, if you're like an engineer who knows how to build some machine or whatever, right, in the 1800s, you don't go to Nigeria and you don't have Nigerian friends that you tell about it, right? Nigeria is part of the British Empire, but because it doesn't have many British people in it, there's not many people that you can share this with. There's not many people that can start a company in Nigeria that does engineering and then link up with Britain, sell stuff to the British market and whatever, buy parts from the British suppliers. Those connections aren't there and that know-how isn't there. And all that stuff just isn't there because there's very little, because most of the stuff is just personal connections, you know, and personal education, culture, whatever. And so because of that, Europeans moving in, so, so, so Asimoglu et al. thought that Europeans moving in forced them to develop good institutions, which then benefited them over the centuries. And then these other, Glazer, et cetera, said, no, no, it's just that they moved in and their descendants stayed there in those, you know.

Whatever.

That's it. And so who's right? We don't know. But oh, and this is by the way, you know, Esimoglu, Johnson Robinson wrote another paper called Reversal of Fortune, where they show that places that were poor, which by which they mean places that were not very urbanized in 1500, became a lot richer among European colonies only. So they looked at like, but where are those? Where are the places that were like poor and not urbanized before the Europeans got there? They were America, Australia, New Zealand. Singapore, Canada. Those are those places, right? And then, or, you know, maybe some, some places in Latin America too, like Chile, Argentina. And then, and then a bunch of Europeans moved in to those places. And those are the places that ended up getting rich and the places that didn't. But once you, once you take that out, once you account for that, you, you definitely see the places that were richer when Europe colonized them are richer now. So, so I think these empirical results are pretty shaky. They rely on some assumptions that, you know, a sparsely inhabited North America was sort of the same raw material as a densely populated Nigeria or Pakistan that probably just don't hold true. That doesn't mean I think that inclusive institutions are not important or that the institutional theory of development is wrong. I just think that we haven't confirmed it. And the Nobel Prize Committee saw fit to include a disclaimer. in its explanation of the prize, saying that you can't really trust these estimates. Well, isn't that bizarre? Like, you're saying, here's a thing, we're giving this thing a Nobel Prize, but you actually can't trust the work. You know, we, you know, we think it's, it's suggestive is the word they use, or strongly suggests, they say. They say it suggests this, and we're giving a Nobel Prize for suggesting this, but, but you can actually have to take this with a grain of salt. They actually use that expression, grain of, you have to take these estimates with a grain of salt. If you have to take them with a grain of salt, why are you giving it a Nobel Prize? Like, it's astonishing the degree of non-empiricalness that is accepted in a prize given for empirical work here. Which doesn't mean that I think that Acemoglu et al. suck at their empirical work, although some people allege they do, and there's backs and forths. So some people allege they do. And then, so what was I saying?

We were talking about the Why Nations Fail book, or, you know, suggestive You know, they're mostly right, but are directionally right. Yeah.

So, so, yeah, I mean, they, they basically, they've given a prize for big think with empirics that can't really, can't really, you know, support it. And, and this is true of, you know, they mentioned other work by the, by, you know, these guys on, on democracy and growth. Well, it has similar problems. I won't go into it all day, but I guess the, the, that's the, that's the point. The point is that This represents the Nobel Prize moving away from really, you know, verifiable, concrete, definite stuff, whether theory or empirics, and back toward kind of, you know, vague big think that we think is kind of cool, like what Hayek wrote.

And by the way, quick side note on Hayek. One of Hayek's, you know, big ideas, of course, was gone around like, you know, sort of the, you know, decentralized system of markets being critical to discovering the true price of things. And I'm curious if you think in an AI-driven world, you know, there will be more effective centralization of price or centralized organizations that have data and can use AI to synthesize data and make sense of it will be able to have better price discovery. And does that cut against Hayek a little bit?

Maybe so. Certainly a number of people have postulated that, but I don't know yet. I don't know I sort of think Hayek's right, but it's hard to, you know, that's big think that it's hard to verify with science too. And my guess is no, because AI will generate just as much information as it analyzes. And that information will need to be aggregated by AI markets. That's my intuition, but I don't know. And yeah, and I don't even know how important the stuff Hayek talked about really is quantitatively. I think it's obvious the mechanism that he described by which markets can aggregate information, you know, if like oil gets expensive, that tells you to drill for more oil and then you drill for more oil and then oil gets cheaper again. Like that's real. We have all seen that happen. How efficient those price signals are, how many price changes reflect those signals, what they tend to be a signal of, blah, blah, blah. Like those are all empirical questions that I think that Hayek didn't even try to answer. We didn't even think about trying to answer, really. And I think that, you know, because econ was just philosophy at that point. And so, but, you know, now we, for example, there's a, there's a real question as to whether or not inflation expectations motivate inflation and macroeconomists argue about this again and again. Well, that's sort of what Hayek was talking about, you know, inflation expectations of prices should influence the price that you set, which should influence actual prices. There's a feedback effect. But, but we don't, you know, people argue back and forth about how much that happens. how much that matters.

Yeah. That's a segue to this piece that you wrote about AI. Well, and also Asimov, because one of the places it seems like you disagree with him a lot is his thoughts on AI. But also, you wrote about this idea in your piece that people often equate the strengthening of or improvement of our models with sort of the improvement of accuracy and truthfulness, truthiness. And you write about how this idea of, hey, maybe power and truthiness are actually not correlated. And to the extent that they're not correlated, that will present opportunities for humans to be in the value chain or add value along the way. Maybe we could start with why Asimov is wrong on AI, and then we could get into your nuanced view.

Well, okay, so I don't know that Asimov is necessarily wrong about AI. And he's done a lot of valuable work. Remember, I think Asimoglu is primarily talented as a theorist and kind of honestly mid as an empiricist. And that his virtuosity as a theorist causes people to overbelieve in, you know, to have too much confidence in his empirical results. So for example, a paper he did with Pasquale Restrepo found that companies that do more automation hire fewer people and then, or that have more industrial robots. Hire fewer people, and then industries where there's more industrial robot adoption hire fewer people. So they're like, okay, robots, industrial robots destroy jobs. But then a whole bunch of other people did this with better data, you know, and then, and I don't know, whatever, more companies, longer time horizons, better definitions of things, basically better data. A whole bunch of other people did this thing and they found the exact opposite. They found that companies that hire the, that buy more industrial robots actually hire more humans. Overall, and that's that has become a pretty robust finding, and it directly contradicts their empirical result, even though their empirical result got a lot more play and a lot was a lot more influential than the subsequent many many papers that rebutted it. And at the same time, so then they were like, okay, well, it's because those companies just become more dominant. You know, the that's why they hire more humans. But then at the at the whole industry level, you'd see. You'd see this shift where the industries that automate more, you know, destroy, destroy jobs. And then, and then people looked at the industry level, and then that, you know, the industries that automate more also gain jobs. And people are like, but the jobs are getting destroyed somewhere, the job destruction must be somewhere. And so eventually, you end up doing this at the level of countries to try to look for the job destruction at sort of the most general highest level. And at that point, you get to the point where regressions can't really find anything because like the data is too sparse. There's just not enough data points and it's contaminated by other stuff. And like, and at that point you're like, okay, we declare victory because we took it to a level where we weren't able to actually get good empirical work. Therefore, our theory wins. Well, that's not very scientific. And I think that the, the job destroying effects of industrial robots and automation in general are just incredibly hard to find in the data. And ultimately everyone's got a job now. So like, however much automation there's been, it didn't stop people from having jobs. You know, middle-class wages are the highest they've ever been. Obviously, this is counterfactual. Maybe things would be even better. Maybe middle-class wages would be even better without any automation. But it's, it's, it's really, really hard to find the job-killing effects of automation in the data. And then, you know, Anasimoglu has often played fast and loose with this theory too. For example, he wrote a paper called The Simple Macroeconomics of AI, where basically he broke, he used theory to break AI's potential macroeconomic effects, effects on jobs into like multiple things it could do. AI could make you more productive at the things you already do. It could just replace you flat out. It could, it could create new things for you to do, or it could make your tools just more productive, whatever, blah, blah, blah. But the point is then the whole thing where it could create new things for you to do, he just assumes that away. He says, I assume this is zero. And, and then of course, all the results say that AI destroys jobs because you've assumed away the main mechanism by which AI creates jobs. Congratulations. That, you know, he's a talented theorist for breaking down the potential effects of AI into this very clear taxonomy that you could then potentially measure, maybe if you had good enough data. But then as a polemicist, as a blogger, as a public intellectual, he just turns off a piece of his model in order to assume his desired conclusion that AI kills jobs. That's malpractice. And, you know, as a blogger, that's blogger malpractice. All right. Probably, I mean, research-wise, it's not malpractice. It's just, you can make whatever assumptions you want and people can decide how smart they think that is or not.

But it doesn't pass the sniff test.

No, you shouldn't do that. And so, and then he wrote this book with, with Johnson, Power and Progress, that they just had a whole lot of historical howlers.

Did we ever, did we ever do that review on, did we ever do a I think we alluded to it, but we haven't done a dedicated segment on it, so we should.

Right. Okay. So a lot of it is just economic history. And basically the economic history is meant to tell the story that, that tech bros want to destroy jobs. So you can't let the tech bros invent what they want. You have to force them to invent stuff that creates jobs. That's the basic message of Power and Progress.

We'll continue our interview in a moment after a word from our sponsors. Hey everyone, Erik here. In this environment, founders need to become profitable faster and do more with smaller teams. Especially when it comes to engineering. That's why Sean Lanahan started Squad, a specialized global talent firm for top engineers that will seamlessly integrate with your org. Squad offers rigorously vetted top 1% talent that will actually work hard for you every day. Their engineers work in your time zone, follow your processes, and use your tools. Squad has frontend engineers excelling in TypeScript and React and Next.js ready to onboard to your team today. For backend, Squad engineers are experts at Node.js, Python, Java, and a range of other languages and frameworks. While it may cost more than the freelancer on Upwork billing you for 40 hours but working only 2, Squad offers premium quality at a fraction of the typical cost, without the headache of assessing for skills and culture fit. Squad takes care of sourcing, legal compliance, and local HR for global talent. Increase your velocity without amping up burn. Visit choosquad.com.

But we won't know what creates jobs, right? Isn't it, uh, kind of difficult to anticipate?

Yes, that is one of about 1,000 reasons why this book is just one of the worst nonfiction books I've ever read. So for example, they talk— they say they're making a point technology can be used for good purposes or bad purposes, and then they list new inventions that brought nothing like shared prosperity. Here's an example. I'm quoting from Asimoglu and Johnson's book. At the end of the 19th century, German chemist Fritz Haber developed artificial fertilizers that boosted agricultural yields. Subsequently, Haber and other scientists used the same ideas to design chemical weapons that killed and maimed hundreds of thousands on World War I battlefields. In case you don't know about the Haber-Bosch process, this is what allows artificial fertilizer, which is responsible for feeding the entire world. If artificial fertilizer were not usable, if this one chemical discovery were not usable, literally billions of humans would starve to death. We could not, we could only support a population of like, I don't know, 5 billion or 4 billion or something like that. We could not support the number of people, like billions would starve without this thing. And it's just the most important agricultural innovation in all of history, even more important than the Green Revolution. And so they said, oh, but because also, Similar chemical reactions were responsible for chemical weapons in World War I. We don't know if this is good or bad. That's in, that's not, that's intentionally dumb when you say something like that. That, that was the worst howler in a book full of howlers. Actually, it's just, it's just chock full of howlers. Like, but I just can't imagine how they allowed themselves to put that on paper.

It doesn't even pass the sniff test. Yeah.

It doesn't pass the sniff test at all. And then, yeah. So anyway, I list like a whole bunch of these and I'm just like, the history in this book is just terrible. It also doesn't cite sources. It just has a bibliographic essay where they just mention all the places that things they read to produce this book, but it doesn't connect individual assertions to individual sources. So you don't know where it came from, which is, that is, that's terrible. Like, how do I, how can I disprove a lot of stuff? So I had to go reading all those papers. This took me like, this, this review took me like a week to write because I was reading all these papers and hunting down these things because they didn't cite their sources. Anyway. I really don't like this book. Also, as power, when they define power, it's not some Marxist thing. It's actually, they just say power is the power to persuade. If you persuade people to allow you to invent job-killing technologies, you have power. Why do people, why are people able to persuade people to allow them to invent these technologies? I don't know. Maybe it's just luck. So power is like rhetorical luck to these guys. That makes no sense to me. Like nobody thinks that's a good definition of power. And they, they, they basically think that accidental success in a nonviolent marketplace of ideas is in the same conceptual category as slavery and feudalism, which they compare it to. They say those were cases where power determined economic outcomes. Also, tech bros writing, you know, VCs writing op-eds that, that, that, you know, reduce appetite for AI regulation. That's also power if people are persuaded by those op-eds. Well, that's dumb. That, like, how can you write that in a book and have people think you're like the top intellectual on the subject? I don't understand it. It was just terrible. And there's a whole lot of other problems. So my review is about 7,000 words. You can go read it. And then there's just like lots of things. There's empirical mistakes too. There's like the facts they claim that are just like the opposite of right. And then, and there's lots of stuff they just don't grapple with and they just avoid. Anyway, it's bad. It's a really bad book. I didn't like it. It's fractally bad in that you'll see like a whole big claim will be bad, but you'll zoom in and see tons of tiny little incidental off-the-cuff bad claims they didn't necessarily need to make in order to support the big bad claim, but that they make anyway just because they can and which are bad. Yeah, I'm very frustrated. In case you can't tell, I'm really frustrated. I spent a lot of time reading this book. It's possible to frustrate me to the point where, you know, I lose my cool and this book just really, oh man.

7,000-word takedown. Amazing.

I want that. I want those weeks of my life back, man. Bro.

Eddie wins a Nobel. Crazy. Well, let's, I wanna, let's flesh out your thoughts on AI because they've been evolving. You know, we've done a conversation with Dario obviously a few, you know, a month ago or so, one with Nathan Benz before. You've been writing about it and it seems like you have some synthesis here. Or when you share kind of the latest thinking.

I don't know. I think, you know, I didn't know a couple of weeks ago and I don't know now, but I think that absolutely everybody is sort of doing a more casual version of what Asimoglu did in his paper, The Simple Macroeconomics of AI, which is to say, absolutely everybody is simply refusing to imagine the jobs AI might create and focusing all their attention on the jobs that AI might replace. Is this appropriate? Should it, should we assume that AI is simply a people remover? And that that's the assumption we should make. And that other things, like so many other technologies in the past have created jobs for humans to do. Should we just assume that AI is fundamentally different? And that AI, AI is the people remover? Remember that a lot of those old technologies people thought would function primarily as a people remover as well, ex ante, right? People thought that like steam shovels would displace all the construction workers. And people thought that like factories would displace everybody. People thought that like, there's a lot of these, a lot of these examples of technologies that people just thought like, there'll be nothing for humans to do. And then there were. Now that doesn't mean the future will be like the past. We have no assurance that the future will look like that. Right. And in fact, I think there's a very good argument to be made that there's a difference in the job effects between the first and second sort of computer revolutions in the first in the '80s and '90s and the second in the like late '90s and 2000s. I think internet technology looks like it behaved differently than database technology. Those are the 2 big things, by the way, that in the 2 big ways computerization affected business, you know, databases meant you could just sort of, you know, you moved all your office work to digital stuff and then like Excel spreadsheets to Excel. And that was huge. Like, you know, Oracle exists because it helped, it held people's hands. SAP exists, right? Because they, they hold people's hands and they teach them how to use databases. That was like the first B2B. you know, IT software stuff was database. Teach a company how to put their shit on a computer. That was your first wave. And your second wave was teach companies how to use the internet. And so when you get to internet stuff, I think there's actually some decent evidence that this on net killed jobs and reduced wages for middle-class people. That the internet, that the technology that Acemoglu et al. believe that AI is, the internet is the actual that. And I have some half-baked hypotheses as to why that's true. I think that the internet, primarily what it does is it allows you to, that a lot of human labor was local redundancy. You'd have like, you know, salesperson in every county, right? Because you had to be on the ground. And now the internet allows you to just blast stuff out, just one person behind a computer blasts it out to a million people. So I think that my suspicion is that delocalization, that a lot of human labor was local redundancy and that the internet got rid of a lot of that. And therefore, that's why it had ultimately, so far, I think the internet has largely had a job, not necessarily job destroying effect, but certainly promoting inequality. And so I think that a lot of times economists and also pundits tend to fight the last war. They tend to look at the last big technology that came through and think that the next big one will be just like that. I think that that's why everyone was complacent about the internet. Everyone thought, oh, well, computers have done, you know, Everybody's got an email job now with computers. Everybody's got a job. People used to just write these things on paper and put them from the inbox into the outbox, right? Now, you just have a digital inbox and digital outbox. Now, you just do Excel. You type things in Excel. It's great. Still work for people to do, and that's going to be fine. It's just going to make people more productive at their existing tasks. And I think that people were, in 2004, people were incredibly complacent about what the internet would do to human labor because the computer revolution had gone well. And now I think the internet revolution, I think, did not go great for human labor, although I think that's a very general statement and I think not always true. Obviously, many new jobs are created like digital marketing consultant, you know, like that wasn't a job before. Like, um, yeah, so those new jobs are created and I won't say the impact of the internet was uniformly bad, but I think that overall it probably promoted more inequality. I think internet made the world a bit more unequal, also by outsourcing, offshoring, all that stuff it facilitated. And so I think that everybody expects AI to do this partly because we've inundated ourselves with science fiction stories about AI that does this, but also partly because that's how the last thing went. Like, we're fighting the last war here opinion-wise. And I think we do that all the time.

Yeah, that makes sense. Just because I do have to leave early, we have a couple minutes left. Is there anything we haven't yet covered on this topic or a previous topic we spoke about that's worth making sure we get into?

Well, sure. So I guess in economics, one big problem is that the glamour divisions are not the most reliable in terms of science. And this is why people think econ isn't a science, because the economics that they interface with, that they see in their daily lives, is the least scientific part of it. They see people prognosticating about inflation, which is very, very hard to forecast for anyone. They see people prognosticating about growth and employment, which are very, very hard to forecast. They see people making big pronouncements about why nations fail and why nations succeed, and growth and all these big things, these economists making grandiose pronouncements and promoting these big theories. And then those theories are big and vague enough where you can certainly find casual evidence to say, oh, look, that didn't happen. Or you can look at individual inflation forecasts and say, oh, look, that didn't happen. These are the glamour divisions of economics because they grapple with the big, big questions. I'm not saying you shouldn't grapple with the big questions, okay? I'm not saying you should focus only on the things you can get solid answers to and then ignore all the other questions. Don't ever think about why nations get rich because you'll never know for certain. Don't ever think about why, you know, whether institutions are good because you can never know for certain. No, I'm not saying any of that. You should grapple with these big questions. But the fact that the people who are getting much more solid, reliable answers and making much more solid, reliable theories in economics are not getting the attention. That the big think people do and that the big macro people do. Macro gets all the attention. And then micro, which is like most of economics and which gets incredibly solid results and often generates many, many billions of dollars in corporate value and improves lives in any number of ways and makes all kinds of policies better, does all these things like this humble, effective plumbing or dentistry, as Keynes called it, this dentist economics. is actually by far more common and often involves harder math, honestly. And yet it gets results. It really gets results. And I'm frustrated because people think economics is this branch of woo-woo philosophy, which in the past led people to think economics is just a whole bunch of free market libertarian bullshit. And in the future, we'll get people to think, oh, economics is just a bunch of like commie anti-technology bullshit. Right? Already all the VCs and tech people are just looking at the Asimov stuff. They're being like, what is this guy on? What he's on is big think, right? And people thinking econ sucks because it's all about this big think. I don't want to say big think's bad, but I do think that we economics— I was happy with the way the Nobel Prize was foregrounding the humble, effective stuff over the vague, big think, fun, indeterminate stuff. And this looks to be a reversal. We'll see if it's a permanent reversal. We'll see where the Nobel Prize goes in in future years. It may be that Acemoglu is kind of a one-off. He's the only macroeconomic theorist, you know, the big, vague, big think guy who's like influential and well-published enough where he pretty much had to get the Nobel Prize. And now, like, after that exception, they'll go back to people doing more humble work. That's possible, especially because there's just a lot fewer guys doing the kind of stuff that Acemoglu does in econ now. Now that it's shifted to empirics, there's fewer would-be Acemoglus than there used to be by quite a lot. And so it may be that the Nobel Prize will just snap back and this is just a one-time anomaly and blah, blah, blah. Or it may be that whoever's in the Nobel Prize committee has just decided, you know what, these critics are right. And guys like Deaton and Pritchett and guys who said that we weren't rewarding big think enough are right. We need to start awarding more big think. And we might see a shift back toward grandiose, unverifiable big think as the thing that gets rewarded in the economics profession. I hope we don't because I like science and I think big think is necessary, but science is cooler.

Yeah. Well, that's a great note to wrap this podcast on. It's a good discussion on economics as a science. Noah, as always, until next time.

Until next time.

Upstream with Erik Torenberg is a show from Turpentine. The podcast network behind Moment of Zen and Cognitive Revolution. If you like the episode, please leave a review in the Apple Store. Hey everyone, Eric here. At Turpentine, we're building the first media outlet for tech people by tech people. We're the network behind the show you're listening to right now. We have a slate of hit shows across a range of topics and industries, from our AI and investing cluster of podcasts. To shows that drive the conversation in tech with the most interesting thinkers, founders, investors, and influencers like Econ 102 with Noah Smith. We're launching new shows every week, and we're looking for industry-leading sponsors. If you think that might be you and your company, email me at erik@turpentine.co. That's erik@turpentine.co, and let's partner together.