--:-- Dwarkesh Patel: Okay, today I have the pleasure of speaking with Caleb Watney, who is the Director of Innovation Policy at the Progressive Policy Institute. He's recently written a piece in The Atlantic arguing that America's innovation engine is slowing. So before we get into the weeds, Caleb, do you want to describe the main reasons why you think America's innovation engine is slowing?
--:-- Caleb Watney: Sure. America's innovation engine is a very broad and multivariate thing, but I was trying to focus on three particular trends that I thought were troubling and make major impacts on America's ability to innovate. Briefly, those are: high-skill immigration and international talent flows; number two, our cutting-edge university system that's really the best in the world; and three, the industrial clusters, the agglomerations that are the geographical regions where a lot of this innovation comes from. Each of those were seeing negative trends before the pandemic, but the pandemic might serve as a breaking point in each of them.
--:-- Dwarkesh Patel: Okay, so let's start with remote work first, or at least the agglomeration effects first. What is so important about the physical agglomeration effects of being in these high productivity cities? Why can't remote work make up for that?
--:-- Caleb Watney: Sure. Economists have for hundreds of years been studying the effect of these close geographically connected networks of entrepreneurs, innovators, scientists, academics, supply chain managers. When you have all of those people in a geographical region where they're running into each other spontaneously, sharing ideas, having long conversations in the hallway or in their offices that they wouldn't otherwise be able to replicate simply online, that ends up leading to more creativity, more ideas, more patents. We have a lot of evidence for that in the real world.
--:-- As you mentioned, some people are starting to counter and say that in the last 20 years we've seen this huge array of online communication platforms pop up, and can't those replicate a lot of the agglomeration effects we're getting in a physical environment? I think that's certainly true to some extent, but the right comparison here isn't necessarily physical agglomeration versus digital agglomeration. Right now, if you're in one of these geographic clusters, you're getting the best of both worlds. You're getting the best that physical agglomeration effects have to offer, plus you have tons of great conversations on Twitter. So it's really digital and physical agglomeration effects versus just digital. I think the difference there is still pretty large, and it would be a shame to lose a lot of the high productivity clusters that we currently have.
--:-- Dwarkesh Patel: Let me make the case that remote work is actually going to increase some sort of agglomeration benefit, and you can convince me that it's wrong. First of all, it's going to connect the entire world into this hyper-productive city that we can't have in the real world, not just because of the limitations of how big cities can be and who can move in there, but because of the inevitable zoning laws that cities like San Francisco passed that artificially limit how many people can join into the innovation circus there. Now we're going to have developing countries and the talent there joining into these agglomeration benefits, as well as people in rural areas in America. Why won't this massively outweigh the loss of physical agglomeration?
--:-- Caleb Watney: I think digital agglomeration, these digital networks, can have huge benefits. You're exactly right that there are a lot of ways you can arbitrage both bad regulations like zoning laws, but also physical distances. But again, the relevant alternative here is not where you're just getting physical agglomeration or just digital. Right now, living in Washington D.C., Silicon Valley, or New York, you can have those conversations with people from all over the world. You can even hire remotely. But you also get a certain kind of cultural context that you only get by being in person and sharing conversations.
--:-- I certainly experienced this in my own life. I get a lot of ideas from having conversations on Twitter, but it's still very different to have an extended two-hour long conversation with a friend in their office, making eye contact, having spontaneous ideas that are hard to replicate even over Zoom. Partially because Zoom meetings tend to be a little bit more scheduled. It's hard to have a flexible time. Frequently the best conversations that happen in the real world are ones that weren't intended to start off as conversations. They just spontaneously arise. You start off sharing one thing, and then it extends into a whole conversation.
--:-- There's also a kind of Zoom fatigue that you see. People, especially right now in this era of remote work, are having so many Zoom calls. They don't want to stay on longer just to have a conversation with somebody extra. So again, these aren't completely contradictory. We can have both. That's really what we're getting in the big physical agglomeration effects of today. I think we need a better taxonomy of what we're talking about in terms of the kinds of interactions that lead to innovation. There's a spontaneity and really in-depth conversation that you can only get in physical spaces, but there is a huge effect from having way more people that you can have access to in digital spaces. Rather than asking which is better or more important, let's break down what are the specific attributes of these kinds of interactions that are happening in the physical or the digital space.
--:-- Dwarkesh Patel: That's very interesting. In your Atlantic piece, you talk about how there might be aspects of remote work that advantage the businesses that are using it, but for the net rate of innovation, it's actually quite bad. Can we understand this in the context of you can't explicitly say that you just want to hang out at this coffee place or in your office just in case somebody wants to help you leave this company and start another company, or maybe you'll find a mate there. But getting on Zoom requires some sort of explicit communication of your motives that you can get away with in the real world.
--:-- Caleb Watney: I think that's exactly right. Some workplaces have tried to experiment with this, like what if you just left your camera on as if you were working next to somebody in an office? But people seem to be really uncomfortable with that. They don't seem to like it. It doesn't seem to have the same sort of easy, spontaneous conversation that you can mimic in the real world. I'm not sure exactly why, and I don't want to rule out the case that 30 years from now, or maybe sooner, we have some sort of virtual reality technology that's really blossomed, and everybody, regardless of where you're living, is working in these virtual workspaces. If that's the case, then maybe physical agglomeration effects really start to diminish in importance. But I certainly don't think we're there yet.
--:-- Regardless, it seems like there's a big experiment being taken. I'll be very curious to see what are the effects of all of these big tech companies—Facebook, Twitter, others—that are moving to all remote work for the foreseeable future. You pinpointed something in your earlier comment. There's both productivity within the firm that I'm concerned about, but in some ways, you could say that's internalized. If the companies themselves are seeing lower productivity measures and lower idea generation, they should be able to internalize that themselves. But the thing I'm particularly concerned about from a public policy perspective is the public spillovers—the fact that you have fewer new firms being formed. Facebook doesn't internalize that cost, so they have no reason to consider it. But as a matter of public policy, we do care about new firm generation and the kind of idea spillovers that are better generated in these public externalities.
--:-- Dwarkesh Patel: Is there a reason why agglomeration effects are so salient even today when we have online courses or open source code? I would expect these sorts of more public and transparent displays would reduce the effect of having to be in the right cluster at the right time.
--:-- Caleb Watney: I think they have reduced it to some extent. But at base, humans are extremely pro-social animals. We do all of our thinking and idea generation in groups. Of course, you have some lone geniuses that are introverts, but I think that's a pretty small minority of the population. Another way to think about this is how long Silicon Valley has already lasted and how high the rent costs have gotten. The fact that everybody continues to live in Silicon Valley and pay these exorbitant rent fees despite all of this online technology getting so good is perhaps evidence that the productivity benefits of being in those clusters are actually really quite high. Although we're seeing the rent prices crater recently, so we'll see what happens in the future.
--:-- Dwarkesh Patel: Yeah, we'll see. Let's talk about the big picture before we get into these different points. Why does it even matter if America's innovation engine slows and somebody else takes over? Why can't I make the case that whatever country in the future is the innovation hub will be the one that is free, prosperous, and has adopted the policies you're advocating? Otherwise, it wouldn't be the innovation hub. So why does it matter if somebody else takes the mantle?
--:-- Caleb Watney: Great question. One thing you've seen in this last generation of communications technologies is that there are a lot of cultural values that are conveyed in the building of technology platforms that end up becoming the underlying infrastructure used in the rest of the world. Technology has a great degree of path dependence. Sometimes it's easy to be a technological determinist and say whatever the best technology will always win out. That's true within some bounds. But it's also the case that sometimes we can get caught in a particular infrastructure, whether physical, digital, or even cultural.
--:-- If you look at a lot of the social media platforms, the big ones internationally are primarily American firms. They've built into those platforms certain kinds of cultural mores around free speech, the separation between the private and the government sector, intellectual property law—a huge number of different choices that you end up baking into your platforms. If we were in an alternative universe where all of the main social media companies being used globally were Chinese or run by the Chinese Communist Party, there would be very different values embedded in those platforms. That would really shape the direction the technology plays.
--:-- Especially as we're starting to look into the future of technologies that might be even more monumental, like artificial intelligence, I think it is worth thinking very carefully about where the leading firms are geographically and what kind of cultural attitudes or social infrastructures you end up building in. I would much prefer that the United States and its broadly liberal democratic values are the ones ending up building into the emerging technologies of tomorrow than the Chinese Communist Party.
--:-- Dwarkesh Patel: But what about the claim that's turned into a truism at this point—that liberal democratic countries are just necessarily going to have higher rates of innovation? So as long as the Chinese Communist Party stays authoritarian, they're not going to be the next innovation hub for the things we care about.
--:-- Caleb Watney: I hope that's true. But we certainly can't have complete confidence. If you compare a lot of predictions from liberal democratic economists 20 years ago about what China's expected development path is, they've certainly outperformed the vast majority of predictions. Some of that might be catch-up growth, but they've had much higher rates of catch-up growth than almost any other country. They've been much bigger, and they've maybe had other issues that they've been able to overcome surprisingly.
--:-- The other thing is, I wouldn't underestimate our ability to shoot ourselves in the foot. Absent actively bad policy, this free liberal democratic model should work and should outcompete, especially because the nature of innovation relies on free thinking and iterative thought that's different from the consensus. That tends to work better with these sorts of cultural values. But that only works so long as policy is equally conducive in both cultural environments. Right now, the United States is certainly not playing into its role as the global R&D hub of the world. We're shooting ourselves in the foot in many ways that undermine our own ability to play that role.
--:-- Dwarkesh Patel: There are some questions raised about the effectiveness of Chinese R&D. Patrick Collison has a question up on his website about why, given how much China spends on R&D, there aren't any influential papers in the field he cares about that come from China. Is there something about the way they're doing science there that's not that productive?
--:-- Caleb Watney: I share some of those same concerns. I would look at this in an expected value model. There's some probability that China stagnates or can't really push out the technological frontier—they might be really good at catching up, but that's an entirely different question than actively pushing out the frontier. That's actually where I'd say the majority likelihood is. But we can't completely rule out the chance that they're going to continue making progress in the way they have.
--:-- To that point, I do think there are a number of examples of Goodhart's law—the idea that when a metric becomes a target, it ceases to be a useful metric. You see lots of examples of that in China. One of my favorite examples is they made a big push recently for AI and they said they were going to have 50 world-class leading AI textbooks and 50 world-class leading AI online classes. That seems to miss the point of what a good textbook or a good online class is supposed to do. After you have the first five really good ones, I don't know what the other 45 are really getting you.
--:-- Regardless of whether or not China stagnates, it's good on the merits by itself for the United States to try to play this role. It's good if China does end up playing the threat that we fear, but it's also good even if they end up stagnating and don't. It's kind of a win-win either way.
--:-- Dwarkesh Patel: Let me give you my optimistic scenario. Maybe the U.S. shoots itself in the foot, and China starts to take over as the innovation hub. Innovation is the sort of thing that you only recognize its importance when you can saliently see its lack. So you see China making innovations, we're not. Then we reverse the mistakes that we made that led us to less innovation. We already have good universities and a history of innovation. So if we start to lose our edge, we'll correct course and become the innovation hub in the future. Is this a persuasive model?
--:-- Caleb Watney: It's certainly within the range of possibilities. It does ring true that the United States seems to coast for long periods of time, and then when they feel particularly threatened by another great power, they kick into a whole other gear that a lot of the world forgets we have. Look at how galvanizing it was for the U.S. when the USSR was able to launch Sputnik and get into orbit before the U.S. did. That was a great political awakening. You could use the term "techno-hegemony"—the idea that the United States has traditionally, since the 1940s, been the clear, unrivaled leader in science and technological development. When we have the chance, we'll coast in that role. But as soon as somebody explicitly takes up the mantle as a challenger, that kicks us into another gear. We'll see if that can happen.
--:-- The case for concern would be that American governance institutions are much more sclerotic than they used to be. If you look at the Vannevar Bushes of today, would they be willing to work in the NSF? I think you can have a pretty compelling case that they wouldn't. There are so many more roadblocks to wanting to be an inventive, interesting person pushing out the frontier in useful ways while also working in government. That means even if we get our cultural attitudes right, our institutions might be too sclerotic to really kick into high gear in the way they once were.
--:-- Dwarkesh Patel: As far as these institutions go, there's always been a debate about whether you reform these institutions or try to create new institutions that are better. When it comes to institutions that are funding science, political institutions, whatever else, do you think the best course to increase innovation is just to make them less bureaucratic and have less red tape? Or is it to start alternative institutions?
--:-- Caleb Watney: I don't know that I have strong priors on that. For some of these questions, it's really a matter of political opportunism. You have to play with the cards you're dealt. I would guess that reform seems much more likely than outright abolishing these institutions or creating new ones. But if the chance presents itself to create new ones, that might certainly be worthwhile.
--:-- Part of it is we're trying to create a governance infrastructure that can easily scale to the kinds of problems we see ourselves facing or the new roadblocks that existing institutions are running into. For example, with science funding, one of the easiest, lowest-cost things we could start doing is to introduce more randomized control trials within science funding. New Zealand did an interesting experiment recently where instead of going through this lengthy science funding application process, they just randomly distributed funds to projects that met some qualified metric. So far they found it to be pretty successful. They were trying to randomly distribute it so it would have some external validity.
--:-- You can imagine running 10 different versions of that experiment in the United States, tweaking little variables within the current science funding apparatus, and seeing how each of those randomizations ends up making the results better or worse. Then you can take those results and, depending on how willing the current bureaucracy is to incorporate them, decide whether you want to reform or tear it down.
--:-- Dwarkesh Patel: The concern there is that the people who would be running the experiments are the ones that have a vested interest in preserving the current structure. It reminds me of a story that Thomas Sowell always tells. He used to be a Marxist and then he worked at a labor department for a summer. He had this idea of studying whether the minimum wage actually increased employment, and he was completely rebuffed by this department because they were saying this guy's going to make us all lose our jobs. It's interesting how you go about instituting those kinds of studies.
--:-- Caleb Watney: For sure. There's the whole meta problem here of whether you can even break through the current sclerosis enough to be able to do things that would test the level of sclerosis. It becomes almost circular in some sense.
--:-- Dwarkesh Patel: Okay, let's talk about immigration now. I want to bounce some anti-high-skill immigration arguments off of you and see how you respond. Senator Tom Cotton has proposed limiting Chinese students from coming to the U.S. and studying in STEM fields. This is a direct quote: "If Chinese students want to come here and study Shakespeare and the Federalist Papers, that's what they need to learn from America. They don't need to learn quantum computing and artificial intelligence from America." The case here is that it's just some sort of arbitrage where they'll learn the information that taxpayers have subsidized in the U.S. and take that innovation to China. Why is that a bad argument?
--:-- Caleb Watney: I think it's a pretty foolish argument, and it would be even worse if we ended up running some version of that. There are a number of factors here. One, it's good to recognize what the CCP themselves recognize as their biggest barriers to achieving technological leadership. They've consistently pointed to their lack of talent acquisition. They've said multiple times that their innovation strategy is a talent-first strategy because one of their single biggest problems is that all of their best and brightest don't want to stay in China. Instead, they go to the U.S., Australia, or other foreign universities, and then want to stay and work there.
--:-- If you look at AI PhD students from China that study in the United States, when given the opportunity, 90-plus percent of them end up staying in the United States. That's a really big problem for China if they're trying to develop the cutting-edge AI companies of the future. They've been pretty explicit that talent retention is their biggest issue. They're undertaking a bunch of programs to try to fix that, like the 1,000 Talents Program and the 1,000 Foreign Talents Program, both aimed at recruiting Chinese nationals back home and talented scientists from abroad.
--:-- In some sense, if we took extremely broad measures to shut down all Chinese students from coming here to learn STEM, we'd be doing them a huge favor. That's exactly what they're hoping for: "Great, now we can keep all of these smart Chinese STEM students here in China rather than in the United States." It would play into the CCP's hands.
--:-- Two, I don't want to downplay that Chinese industrial espionage isn't a problem, because there are certainly many examples of it. One problem is we don't have a good sense of how big or small the problem actually is. You hear tons of anecdotes or tangential stories, but it's hard to know systematically, by the nature of espionage, how widespread this is. I would be supportive of more narrowly targeted ways of doing more background checks before Chinese nationals are working on sensitive government projects or receiving government funds for research. We can run counter-espionage networks.
--:-- Additionally, sometimes the uncertainty about Chinese students' ability to stay in the United States ends up giving the CCP leverage. If you give them a clear path to permanent worker status in the United States upon completion of their degree, that reduces a lot of the leverage the CCP has. Broadly, we could certainly take many more targeted actions to reduce the threat of espionage. But to take overly broad actions that end up playing into the CCP's hands would actually be a mistake.
--:-- Dwarkesh Patel: The fact that their top talent wants to come here and stay here is such a good card to have. It's amazing that people want to burn it down.
--:-- Caleb Watney: Absolutely.
--:-- Dwarkesh Patel: Okay, here's another argument. Eric Weinstein argues that in the long term, in a free market, wages and demand equilibrate, so there's no such thing as a labor market shortage. So this entire business about having shortages in STEM fields is a conspiracy from the NSF to reduce wages for STEM field graduates to help employers.
--:-- Caleb Watney: I think that's pretty silly. It really misses the positive-sum nature of fields in technology. The fact is, we don't have some set number of top-end machine learning jobs that are just waiting to be filled by American nationals. If you keep foreign nationals out, that means more jobs and higher wages for Americans. That may be true in individual cases, but in a dynamic equilibrium, it's totally not true.
--:-- Foreign immigrants that come here have much higher rates of patenting. They make their American counterparts more productive, and they patent at higher rates when they've had exposure to these international entrepreneurs. They also start businesses at higher rates, which is by definition net positive in terms of the number of jobs. He's completely missing the positive-sum element of this. There's sort of a word game about whether there's a STEM shortage or not.
--:-- Caleb Watney: But I think it's again not particularly relevant, because it is true that to make progress on the cutting edge of emerging technologies, you just need more really smart people working in a conducive environment. And the more really smart people you have working in a conducive environment, the faster your progress will go. So again, I would prefer the United States to have that progress inside. I want all of the best and brightest world talent to be coalesced in the United States.
--:-- As one additional point, there's also a long line of academic research backing up a very strong case for high-skill immigration. But there's a new paper recently that came out in NBER that specifically showed when you cut down programs like the H-1B program, it actually increases offshoring. When companies don't have the option of bringing their talented people here to the United States, it's sometimes easier to just move your headquarters somewhere else where you can hire them. So that's a whole other dynamic that he's not considering.
--:-- Dwarkesh Patel: Oh, that's a very interesting piece of evidence. And it's probably going to be exacerbated by remote work. It's going to be much easier to just offer labor. And yeah, it seems our intuition works here. When we're just talking about native born talent, there's more CS graduates now than there were 20 years ago, but there's more IT jobs because every new CS grad is a probability that a new business is going to be created and more jobs will be created. Sure. No, go ahead. Sorry.
--:-- Caleb Watney: I would say though that I think there's a lot of complaints against the H-1B program that are completely valid. I have a long list of reforms that I would like to see the H-1B undertake. But just taking it as it is and cutting it in half or slashing it, I think would be very counterproductive.
--:-- Dwarkesh Patel: Yeah, I know those things personally. My dad is a doctor, so when I was eight we moved to the US, and I'm 20 now. We're still waiting in the queue because there's per country caps to get green cards. We'll certainly get green cards, and I'll probably get kicked off unless Congress passes a bill to let kids stay on. I'll probably get kicked off and I'll have to start the process again to get my own H-1B off my CS degree. At that point it might be easier to leave the country and go to another with more lax immigration laws.
--:-- Caleb Watney: Yeah, that's such a mess.
--:-- Dwarkesh Patel: Okay, so how would you go about reforming the H-1B visa?
--:-- Caleb Watney: I have a number of specific reforms for the H-1B, but then as a meta thing, I think in some sense the H-1B is an overrated visa program. I would like us to move our focus more broadly. But I'll start with the specific reforms.
--:-- There are a couple problems to address. One is the current lottery system adds a lot of unnecessary uncertainty and doesn't prioritize the most highly skilled candidates. Because it's just a lottery, if you've met certain qualifications, it gives equal weight to the top-end machine learning engineer with a salary offer of $400,000 from some top firm, and the IT specialist that is still important but would be making $60 or $70,000. If this is a high skill immigration program, we should be giving priority to those with higher salary offers.
--:-- Some suggestions have said we should do a salary ranking where the higher your salary offer, the higher priority you should have in the queue. I would be open to something like that, but I would also want to modify it to include equity, because a lot of times small startups can't compete on salary wage, but they can offer lucrative equity packages. So long as you have some sort of equity-adjusted salary ranking, I think that would be one easy way of making the H-1B much better targeted.
--:-- Another issue is that the H-1B is tied categorically to a very specific employer and it's hard to transfer. This leads to complaints that maybe you can underpay workers because they have no other option. But that's not a problem with the workers or them bringing their talents here. It's a problem with the structure of the visa that doesn't let them switch. You could either try to make it easier for H-1B workers to switch between employers or to even start their own business. Oftentimes people come here, want to live and work in the United States, and maybe they have an interesting startup idea, but they end up having to work at a big company for six years while they're in line for a green card. Once they get permanent residence, then they can start.
--:-- Dwarkesh Patel: Yeah, that'd be ideal.
--:-- Caleb Watney: Which gets to the larger meta problem that the H-1B is pretty poorly structured to actually attract and integrate the kind of talent we want. I've been looking at alternative programs. One I'm particularly excited about is the O-1 visa for immigrants of extraordinary ability. It's a pretty underutilized visa right now, but in principle it's very flexible because congressionally it has no cap. It's supposed to be for immigrants of extraordinary ability, and USCIS has an eight-factor test evidentiary criteria for choosing who fits. You could conceivably have a new USCIS director come in and either issue new regulations to change those eight criteria or rewrite the internal guidance documents, and end up letting in dramatically more talented individuals through that program. The nice thing is, unlike the H-1B, it's not categorically tied with one employer. You could use an O-1 to start a business or work for a startup.
--:-- Dwarkesh Patel: Isn't that called the Einstein visa because of how hard it is to get it and how rare it is?
--:-- Caleb Watney: It is, yeah. It's very hard right now. You really have to be top, top, top end. Even then it's a 500-page petition process, which is awful. But because it's so subjective and the congressional mandate is so broad, the scope for possible reform without needing to go through Congress is also very broad.
--:-- Dwarkesh Patel: Has Joe Biden spelled out how he would liberalize high skill immigration if he became president?
--:-- Caleb Watney: He has not given a lot of details. He's certainly spoken positively about trying to undo some of the damage of the last four years and make it clear that the United States wants to be the destination for high skill immigrants. But understandably, voters have not shown a strong preference for detailed policy plans, so I don't think that's his highest priority right now. In the months after the election, if he wins, there will be a lot of attention to getting the details on board.
--:-- Dwarkesh Patel: I hope so. Speaking of what's going on at this time, another argument I forgot to bounce off of you is: high-skill immigration is usually good, but right now given the high unemployment rate and the economic catastrophe, we need to be emphasizing our own laborers. So for a little while, let's pause the H-1B while our economy adjusts.
--:-- Caleb Watney: We have lots of economic evidence that regions with higher rates of H-1B historically have higher rates of job growth for native workers and higher wages. Even a pause doesn't necessarily make sense. The kinds of job categories that have been hardest hit by the pandemic are not the ones where H-1B employers are typically hiring from. It's mostly service sector jobs or jobs that require a lot of in-person interaction with customers. So it's pretty unconnected from the specific job market issues of this pandemic.
--:-- Dwarkesh Patel: To the extent it is connected, it's through medicine, which is increased demand now, if anything.
--:-- Caleb Watney: Exactly.
--:-- Dwarkesh Patel: Okay. I think you said somewhere that immigration is the one lever you change first to increase levels of innovation. Why is it so much more important than other factors?
--:-- Caleb Watney: I would consider it my number one lever right now simply because it's so far away from optimal. If we look at trade policy as an analog, the free flow of goods is pretty close to optimal. We still have tariffs and dumb regulations, but moving another five percentage points in favor of free trade on the margin just doesn't matter that much. But we're so far away from the global optimum on immigration policy that a 5% change in how many really smart, skilled, talented people could come to the United States could make a very large difference. If you follow the idea that people are the ultimate resource, they are sources of creativity and innovation. If you're trying to affect innovation, you want the thing that can most directly come up with new ideas, which are people.
--:-- Dwarkesh Patel: That's definitely all very persuasive. Let's talk about big tech. We were talking a month after Congress held hearings considering antitrust action against big tech companies. The concerns are: they're too big, small startups can't compete with them, they limit innovation, they buy up new startups. There's a lot to unpack. You've also written a paper about decreasing the barriers to entering the AI market. The concern is they have too much data and how are you supposed to compete with the petabytes of user data that Google has? They need to be broken up so there can be more competition and innovation. Will breaking up these big tech companies increase innovation?
--:-- Caleb Watney: There's a lot there. In broad strokes, if we're looking at this in expected value terms, there is some probability that if you broke up these big tech companies, maybe you get lots of innovation. Personally, I would assign a pretty small value there. I just think it's unlikely given many arguments we can talk about. In Brussels, I think it's pretty unlikely. The other factor is that it's a very high risk strategy. If you break them up and you're wrong about the effect on innovation, then you've just destroyed the few golden geese actually producing productivity gains in our economy. That's extremely bad, and you've given a jump to China.
--:-- A much lower risk strategy that maybe gets at some of these same concerns would be a big national push to tear down entry barriers, make it much easier to compete against these companies in a variety of ways, reduce the moats they have around them, increase the gale of creative destruction. If they can survive that level of increased competition, then that further increases the evidence that they are efficient and competitive. If they end up falling because some new startup has out-competed them, then you've gotten the outcome you wanted. As a meta strategy, I would prefer to first try the strategy of increasing the level of competition they're faced with, not by artificially tearing them down, but by making it much easier to compete with them.
--:-- One thing connecting this to the immigration discussion is that a lot of these big tech companies have a huge advantage in the immigration market because they have large HR departments capable of braving the immigration bureaucracy. If you look at the career preferences of top-end talented immigrants, a lot of them say they would prefer to work at startups or start their own company at a much higher rate than American native-born workers. But in actuality, immigrants are way more likely to work at the big firms because that's the only place they can get an immigration visa. If you could make it just as easy for startups to have access to this high skill talent, you would unlock a whole pool of entrepreneurial talent that is currently stuck at big tech firms.
--:-- On the data side, we sometimes get confused and think of data as this homogeneous thing where the more you have, the better it is. In reality, data is very context dependent. It almost has no value outside of the original circumstances that created it. Utilizing data requires a massive internal infrastructure. If someone gave you all of the data from Amazon Web Services or from Google search results, it's pretty unclear how you could use that, even if you're trying to build a competitor. Even if you have this one-time static flow of data, that totally misses the dynamic flow of data that actually matters, because it's what happens when you make a change in your platform and then you can see the results.
--:-- There might be isolated cases where I'd be in favor of interoperability, but we want to be pretty careful. Making data useful in the first place is a very expensive operation. It requires building a massive internal infrastructure with data engineers that can collect the data, think carefully about what kind of data to collect, process it, turn it into useful insights, and then apply it. You need lots of talent at each step to turn raw data into something useful, which is actually where the barrier is.
--:-- There might also be interesting things you could do with government sector data. The government has lots of data about the business sector and geographical regions. There have been plenty of examples where the government has made information that only they have public, and that has led to pretty large gains. One of the best examples is geolocation data from GPS satellites and how that created an entire industry. I would be interested in looking at various open source data sets or data sets you could make open source.
--:-- Dwarkesh Patel: You've been writing about how the attention the news media gives to the EU's regulation of tech companies is so misplaced. If the claim is that they're really going after the big guys, you have to explain why only one of the top 30 big tech startups came from Europe. These new startups are the ones that would be competing against Google and Facebook and Amazon. They're the ones these companies are afraid of. The EU puts up such barriers against starting such things that it benefits the very big companies they're presumably trying to target.
--:-- Caleb Watney: That's exactly right. If you look at something like the GDPR, which Europe was touting and you still see tons of conversations in the United States about adopting the EU approach, after the GDPR went into effect, it only increased the market share of Facebook and Google. Similar to the immigration discussion, the more you have these large bureaucratic burdens that you need to meet, the easier it is for large companies to handle that because they have the staffing resources to have huge teams of lawyers to make sure they're compliant. The United States is the one actually producing top-end companies that are competing internationally. They're by far the most innovative in terms of patents and where interesting startups are coming from. The EU talks a big game, but their regulations in practice just make it even easier for the big companies to succeed there.
--:-- Dwarkesh Patel: I like a thing Jonah Goldberg says, which is that complexity is a subsidy. If you have more cognitive or legal or whatever other labor resources, the more resources you have, the easier you can navigate the complexity.
--:-- Caleb Watney: That's a great framing.
--:-- Dwarkesh Patel: Let's get a little bit into the politics. This might be a meaningless question, but given that there's an election coming up, which party cares more about innovation or would have a better pro-innovation strategy? You can look at Trump and say he might be good on regulation and taxes, but he's bad on immigration. The Democrats on balance, which one is better for innovation?
--:-- Caleb Watney: Talking about parties is in some sense not the best framing, partially because the parties are so broad. One party incorporates everyone from AOC to Joe Biden, the other incorporates everyone from Donald Trump to Jeb Bush. I think a better way of thinking about it is that there are sub-parties within the existing national parties, and they're competing. Whoever wins the primary ends up getting the national thing.
--:-- I would say a Joe Biden agenda would be better for innovation in the United States than a Donald Trump agenda, especially when you consider what I see as the biggest drivers of innovation that we're missing: high skill immigration reform, federal R&D reform, and a lot more housing in these big agglomerations and industrial clusters. Those would probably be my top three. On each of those, Joe Biden has expressed more interest than Donald Trump has.
--:-- Dwarkesh Patel: Oh, I didn't know that he was proposing zoning reforms in the big cities.
--:-- Caleb Watney: He's not really talking about many of these things, but in terms of where the energy is within the people that would likely be put into the various cabinet positions.
--:-- Dwarkesh Patel: Good. I'm happy to hear that. But what about the fact that he wants to raise the corporate income tax rate or probably increase the regulations that Trump cut down?
--:-- Caleb Watney: That would be bad in some ways. Everything is a series of trade-offs. I would take a trade-off where we get more high skill immigration and more federal R&D experimentation in exchange for slightly higher taxes or slightly higher regulations in some ways. The better way to look at these things is on a case-by-case basis. At some level, practically, you have to choose which package you prefer. That's kind of what I would prefer at this point. But that doesn't mean I'm not going to also try to push the Biden administration in a better direction on some of these things.
--:-- Dwarkesh Patel: I'm happy to hear that. With these three areas, what is the most undervalued phenomenon relative to its impact?
--:-- Caleb Watney: I think we still have a pretty poor understanding of why some of these industrial clusters form in the first place. Sometimes it seems like accidents of history, that it just so happened that such and such a company was here, and then a number of competitors formed around them. But how you get from a fledgling industrial cluster to something full scale like Silicon Valley is pretty mysterious. We've had lots of artificial experiments to try to seed new industrial clusters in other places that have failed pretty miserably. Sometimes they work. A much better understanding of how and why these agglomerations form is something I would love to see.
--:-- I think also the structure of federal R&D support is pretty understudied. As a percentage of GDP, basic R&D support from the federal government keeps declining. We're also not seeing a lot of experimentation in the structure of these grants. I'd love to see a science of science, to know that the structures we're using are the best versions of them, because I think we're probably pretty far away from the optimal. Given the huge public spillovers of well-structured, well-targeted public subsidies for R&D, it matters a lot.
--:-- Dwarkesh Patel: I have a sort of naive argument against more government spending in R&D. Matt Ridley's new book on innovation talked about this. The problem with government spending more on innovation is that they don't know where the next big areas of innovation are. They might be misspending their money, and worse yet, they might be crowding out the priorities of the actual businesses that have a hands-on practical understanding of what the interesting problem areas are. The researchers are working on the problems the government wants them to work on instead of the problems that businesses would have a better understanding of.
--:-- Caleb Watney: A lot of these arguments could be true in the abstract, but not necessarily true on current margins. If you look at something like the R&D tax credit, which is probably the biggest in terms of spending that the federal government does, it is not directed towards a particular thing. It's basically saying, companies, invest in the R&D that you think is most valuable, and it will reduce the amount of taxes you need to pay. The biggest tech companies today are by far the biggest R&D spenders. The top five are Amazon, Google, Apple, Qualcomm, in some order. It seems to be successful in prompting our big tech companies to spend on R&D, which is good.
--:-- The other thing is we maybe need a better theory of what kinds of science we want the federal government to be investing in. Private companies will invest in R&D so long as it is sufficiently predictable in terms of outcomes, sufficiently timely in terms of being on a profitable timeline in the next five to ten years, and sufficiently cost effective. It has to be profitable in some sense. But there are a lot of ways in which valuable questions don't fall into one of those three categories. Previous scientific advancements have not met one of those criteria. Especially in terms of predictiveness, we're constantly getting scientific advancements in fields completely unconnected from what we think we're going to be studying, partially because scientific breakthroughs kind of have to be unpredictable, otherwise they would have already been discovered. There's not really an incentive for Google's Revolution Cardivision to be pouring money into something that might suddenly have an advancement for molecular biology. The more unpredictable in terms of how this is going to pay off, the less likely companies are to do it. When you start focusing on what kinds of scientific questions fall into those three buckets, the less relevant the crowd out question is. If private companies are not going to be spending on these areas anyway because it's sufficiently untimely, sufficiently unconnected, and not sufficiently profitable, then you have way less concern about crowding out.
--:-- Dwarkesh Patel: There's also the question of science that is inexcludable and inexhaustible. Maybe it's not worth it for companies individually to invest in that R&D, but for the economy as a whole, the tech is more than worth it.
--:-- Caleb Watney: For sure.
--:-- Dwarkesh Patel: Okay, so the next question is, what is, relative to its impact, the most undervalued future technology?
--:-- Dwarkesh Patel: This is like stuff like crypto or AGI, biotech, VR. Yeah, that's a hard question. I mean, I would say, sort of it depends on, if AGI does end up being feasible, then probably it's AGI, just given how wide-ranging its effects would be across the economy. But of course, that's in some sense begging the question. We don't know if AGI is going to be completely possible to create. There might be difficulties that we're not running into yet that we don't know about.
--:-- Caleb Watney: So I think one thing I would like to see a lot more emphasis and research on is big climate megaprojects that I think are being very underhyped. We're getting to a point where to really avoid some of the worst outcomes of climate change, we're going to start needing pretty dramatic action now. But I think any sort of political approach that basically just asks consumers to take massive welfare hits themselves is going to be completely untenable. It's not going to get anywhere.
--:-- But there are lots of big environmental megaprojects that we could do. Eli Dorado has been paying a lot of attention to this. He's talked about olivine as this kind of mineral that I think you can put on a lot of beaches. It ends up absorbing a lot more carbon than traditional sand does, but it has a very similar texture. The main difference from sand is that it's green, so it kind of even looks cooler.
--:-- Or, you could do something like Yellowstone produces just an absolutely massive amount of geothermal energy, partially because there's a massive supervolcano underneath it. You could tap Yellowstone for geothermal energy supply, a very high percentage of U.S. national energy needs just via that, and you would also be reducing the risk of that supervolcano erupting. So it seems like a win-win. Those are just a few examples. But I think as an area, it's really underleveraged.
--:-- Dwarkesh Patel: Oh, yeah. It's super fascinating. I actually had no idea about the specific things you talked about. And also, if you think climate change is a serious threat to human civilization, it makes perfect sense to take a few risks, just see what happens if you throw some stuff in the oceans, if it's going to save people as you know it.
--:-- Okay, so relative to its impact, what is the most undervalued social or geopolitical threat? This is stuff like falling fertility rates, the rise of Asia, stuff like that.
--:-- Caleb Watney: Yeah, I mean, I think earlier I would have said maybe falling fertility rates just because it's only really in the last couple years especially in the U.S. that it's become a major political phenomenon. But you are starting to see a couple of books now that are starting to discuss that more explicitly. Ross Douthat's *Decadent Society* makes that a big central focus. Matt Iglesias' upcoming *One Billion Americans* talks about that. So I'm glad to see that start becoming a major focus.
--:-- But it does seem like demographics are just so connected with a wide range of outcomes. Culturally, I think it makes a society more conservative and less risk taking. When you have an older demographic, Megan McArdle had an interesting example a couple of years ago when she was writing about this. She talks about an average town that has a median age of 27 versus a town that has an average age of 57. When you're that much older, you are more experienced and maybe you have more human capital or technical skills, but partially because you have less time and you're more embedded in a family, you're less willing to work the 18 hour days and to leave your job to go create the new startup. That ends up having cultural effects, both in terms of company formation and idea generation and innovation, that I think end up resounding. Even young people that are traditionally surrounded by an older society, I think will themselves be less risk-taking too.
--:-- So yeah, maybe I would still say demographic aging, just given the fact that it's starting to get a bit more attention now, but it has such wide-ranging consequences that it's not really talked about in the national discourse in a major way yet.
--:-- Dwarkesh Patel: Oh, yeah, that was the scariest chapter in *The Decadent Society* for me. There was a chapter on authoritarianism coming to liberal democracies, and the one on falling fertility was so scary.
--:-- Okay, so by the way, does that mean that you expect sub-Saharan Africa and parts of Asia to be a much bigger deal, given that they have rising populations?
--:-- Caleb Watney: Oh, totally, yeah. I totally buy that. At the end of *The Decadent Society*, Ross says one of the more optimistic versions is some sort of Afro-Eurasian futurism. If you imagine classically European cathedrals being filled by sub-Saharan Africa, that's a great vision and something that could end up providing both a multicultural society that we want to see, while also trying to help solve some of these birth rate problems.
--:-- Dwarkesh Patel: Wakanda forever. Yeah, that's the future.
--:-- Okay. All right. So to close it out, what kind of advice would you give to a 20 year old, maybe when you were 20, or just to a generic 20 year old?
--:-- Caleb Watney: Yeah, I think I would say think hard about the kinds of questions that you think matter for society, and where you might be most effective. One of the things that I was trying to think about in my own career is, I think problems that the profit motive can be brought to bear on, I just have a higher certainty in my own hand that those will eventually get solved. If there's money out there, someone's going to solve it. But where there's not, that's maybe where there's more attention worth throwing at it because in some ways it runs counter to self-interest.
--:-- I think something like policy analysis, the reason I'm in there is because it does seem like the missing element in the American growth engine. If you think about an O-ring theory of growth, where you need all of the components to really be working well for the rocket ship to take off, it seems like bad policy is really the thing that's strangling the American growth engine. Trying to fix that bad policy is not something that right now is going to be economically profitable, but could end up having a really big impact. It just so happened that my comparative advantage was in writing and communicating and whatnot. So that's why I went into that.
--:-- But yeah, if you think about it using a traditional effective altruism framework, what's the highest impact combined by what your efforts on the margin are going to have the highest impact on? I think that'll get you pretty close to what you should do.
--:-- Dwarkesh Patel: Yeah. I'm glad you're doing the work you're doing because innovation is the most salient fact about our civilization. If we just look at what makes our life better over time, it's clearly innovation, but our incapacity to see future innovation and how much better our lives could be just makes us undervalue it so much. I'm glad you're doing the work you're doing.
--:-- Caleb Watney: Thanks. Yeah, I'm hoping to get other people excited about this too.
--:-- Dwarkesh Patel: Okay. Well, thank you so much for talking to me. This was a lot of fun. I learned a lot.
--:-- Caleb Watney: Definitely. Yeah. Great conversation. Thank you.
