Callum Williams: You know, you look at things like, I mean, there's, there's crazy things like how quickly it takes to put up a building, you know, that takes ages now, you know. San Francisco put up the Golden Gate Bridge in three years. The Empire State Building went up in a year. Like that just would never happen now.
So I think there's loads of examples, and I don't think that, like, AI, which obviously isn't, I'm not trying to say AI isn't amazing, because obviously it's completely amazing, I use it all the time, but, like, the notion that AI by itself undoes all of that stagnation strikes me as just completely wrong.
Steve Hsu: Welcome to Manifold. Today, my guest is Callum Williams, a senior writer for The Economist. Callum was educated at Cambridge, Harvard, and Oxford. He is the author of a book called The Classical School: The Birth of Economics in 20 Enlightened Lives. Callum, welcome to the podcast.
Callum Williams: Thanks for having me, Steve. It's really, really amazing to be here.
Steve Hsu: Yeah, I'm really looking forward to this conversation. I find you to be a reasonable and intelligent economist, and most economists are intelligent, but not all of them are reasonable in terms
Callum Williams: the nicest thing anyone's ever said to me. Thank you very much.
Steve Hsu: So we have a list of topics to discuss. And for the listeners who follow me on Twitter, this conversation actually started because I posted some Fed data on U.S. wealth inequality and that started a conversation. And then Callum, and I said to Callum, Hey, let's talk about it on the podcast. So that is, but that is just one of many topics that we're going to discuss in our conversation because of what's happening with our God, emperor, Donald Trump and Vladimir Putin.
Um, I thought we'd start with something which is related to current events, which is the current rapprochement between the United States. And Russia, and now Callum's employer has a certain editorial line on this, and I'm sure he doesn't disagree with them, but we're not going to put him on the spot.
I'll, I'll say all the things that are pro Donald. And Callum doesn't have to do that. Why don't you start out, Callum, with just maybe talking about how surprised or whether you are surprised with the turn of events in the last week or so.
Callum Williams: In the past week or so no, I mean, no, not really. I mean, it was all, it was all, it was all telegraphed really. And, you know, if you look back at the coverage of the Munich security conference last year it was precisely the fear of people who were at the conference back then. Right. I mean, it was that Trump was, you know, doing well in the polls.
He had often sounded kind of skeptical of NATO. He had often sounded skeptical of supporting Ukraine against Russia. And so really, you know, Trump has done what he said he would do. So in that sense, it's not a surprise, even though, you know, it's a, people still find it shocking in some regard, I suppose.
Steve Hsu: I think in the MAGA world, the surprise is not how Trump felt about all this or directionally where he would go with it. There's always a chance, as in the first term, that the crowd around him would accidentally appoint a bunch of neocons or deep state people around him and they would restrain him from actually doing what he promised.
That doesn't seem to be the case, the, the turn that Marco Rubio made from, you know, prior to his appointment, I, I was thinking, or even prior to a speech that he gave recently, I always thought of him as like a kind of neocon guy, but then he gave this speech as his first speech, I think, as Secretary of State about multipolarity, which really shocked me.
And so it seems like this time Trump may have put in place a bunch of people that are aligned with him and that are, you know, willing, willing to like to sprint in the same direction as he is, in this administration.
Callum Williams: I mean, yes, you know, there is reporting that suggests that what Rubio was saying to people behind closed doors is different to what he's saying out loud. in public. I mean, I don't know. I mean, this is not something that I'm an expert on, but I mean, certainly the kind of, you know, all of the horses are running in the same direction at the moment.
But, you know, then again, we are only in like month two or month one really instead of the administration. He had, you know, won convincingly in the recent election. So he's got a huge amount of political power. I mean, I suppose for me, if I were to look for the positives here, The glimmer of hope, I guess, is that this is kind of how Trump negotiates, typically, where he, he kind of butters up people that he wants to do a deal with and then some kind of compromises reach.
Now, who knows whether that's what is actually going to happen this time around. Maybe it's the case that he really genuinely would kind of like to give a lot of Ukraine to Putin with those strings attached. but we don't know, obviously. And, you know, this, this is basically how he, how he's always negotiated.
It happened with. obviously happened with Canada and Mexico a couple of weeks ago but it happened with the Abraham Accords. It happened with North Korea. You know, it happened with the EU over, over a bunch of washing machines and all that it, it, this is the modus operandi of, of, of, of Trump.
So I, I, that's the only thing that's making me think that it might turn out differently to how people expect at the
Steve Hsu: Now, my, my people who listen to the podcast or follow me on Twitter know that my views on the Ukraine conflict are probably most similar to those of two former guests on the show, John Mearsheimer and Jeffrey Sachs. And, you know, that would be the view which in Europe, people would say, you are repeating Russian disinformation.
Which is that this thing was a long time coming. neocons really wanted to create a proxy war against Russia. We could have stepped off of that trajectory that led to outright conflict. Many, there were many opportunities to step off and we didn't take them because we wanted it. At least a certain faction that was in power in the United States wanted it.
And now a different faction is in the, in power that doesn't want it, wants really to be friends with Russia because they don't want the Russian-Chinese alliance to be something that they have to deal with. and so to me, it was a, an on off thing that took place over about a decade, and we're now in the kind of off phase sort of like risk on, risk off, and I don't know if you've, if, You're allowed to comment on this because I think your, your employer has a very different view from mine on this.
Callum Williams: Yeah, yeah, yeah, no, I mean, definitely The Economist does, and we do, I mean, the, The Economist, Does not subscribe to the kind of Mearsheimer sacks also Cummings to a degree on the Cummings argument on, on, on Ukraine. I mean, I'm, you know, I'm not, I'm not kind of involved with it in the reporting in a very detailed way.
I mean, I, I, I think that I'm, you know, closer to the economist side than, than, than the Mearsheimer side. I mean, I think the kind of cock up versus conspiracy point, I think is valid. I mean, the notion that. There was some kind of engineered proxy war to me sort of falls down in a kind of prima facie sense that it's just too, just too difficult to do.
Um, I mean, I mean, what is clear, I think is that, and this is where the economist has departed a long way from the kind of, as it were, Western elite consensus is that the expectation that EU and American sanctions, we're going to be able to quickly destroy the Russian economy and it's going to bring it to heel has been proven completely wrong.
And it was pretty obvious that was going to be completely wrong right from the start. That is something that kind of weirdly, we haven't kind of oddly had much engagement with our, I think, correct reporting that the Russian economy has proven very resilient. Like, you know, we haven't had people reaching out to us who've got it completely wrong, who said.
Like I'm, I'm, you, you guys were right and we were wrong, you know, there's been none of that kind of introspection I think from, from, from, from other, other economists. So in that sense, we, we kind of have departed quite a long way from the sort of mainstream consensus on, on the conflict.
Steve Hsu: Good. Yeah. So that hence my term reasonable when I said you were
Callum Williams: Oh yeah. Okay.
Steve Hsu: So, so, yeah, I would think someone who makes a strong prediction. The way it's supposed to work in science is when you make a strong prediction and it's wrong, you come back to the other side and you say, yes, well, I was wrong about this and let's all update our, you know, our priors together.
Um, in the case of how well the sanctions are going to work. The thing that popped into my head being, as you know, very familiar with, like, what's going on in the Chinese economy is I just sort of thought, like, let me just flip through Aliexpress and Taobao, which are like two places where you can literally buy like anything in the world.
And now, Americans know it because Timu is very similar to that. And I just thought, like, is there anything on here that they're going to be able to order in Moscow that they're not going to be able to order, you know, from Aliexpress in Moscow? And then I thought, like, nah, I think they'll be fine.
Right? They may have to switch over to BYDs and Huawei phones or something, but I don't think it's going to cripple the economy. And I think that largely turned out to be correct.
Callum Williams: Yeah, absolutely.
I mean, the way that I'm thinking about it is, is that it's, and the data I think supports this and the, and the anecdotes support this quite clearly is that it's, It's basically still possible to get anything in Russia that you want. The question is how much you're willing to pay. And if you want to have a Prada handbag or you want to have a bottle of nice Burgundy wine, you need to pay 20 percent more than you used to, which is, you know, sizable.
And there was, you know, sizable inflation in the Russian economy, particularly in early 2022 when the invasion first happened. You know, the ruble massively depreciated, and it just became much, much more difficult to to get hold of, to get hold of Western goods. But what you've basically seen since then is, yes, you've had that sort of substitution away from iPhones towards Huawei phones or whatever, totally.
The substitution away from other kinds of Western made goods towards Chinese made goods or Indian made goods or whatever. But then crucially, and this is where the, you know, this is probably where I do think the EU has, has really kind of fluffed its lines. What's also happened is that this enormous parallel trading system has Has begun whereby exporters in, in Italy and Germany and, and to a lesser extent, the UK, but you know, big European countries will export German made Italian makers to say places like Armenia, Kyrgyzstan, Azerbaijan, and then it will be rerouted to Russia.
So you see this absolutely enormous surgeon import, say, for example, from Germany to Kyrgyzstan. In Georgia in the opening days of the war, and it hasn't really come back down. Now, obviously this rerouting adds costs because, you know, you need to, you need to pay the middleman, you need to pay for the extra transport, blah, blah, blah.
But like, it's not that, it's not, it's not that much. And, you know, with time, these supply chains become more and more efficient. So the effect has a sort of half life that sort of decays after a while. So in that sense, like purely if you're thinking about imports. Yeah, you're totally right.
Your supposition that you could go on to, you could go to AliExpress and do it was, was essentially correct, I think.
Steve Hsu: Yeah, you know, on this evading sanctions kind of thing, I guess one variable which I didn't know how it was going to turn out was the oil thing, but I didn't really think we were going to be able to stop, you know, India and China from buying oil from Russia. You know, as far as sanctions go, you know, we have this interesting phenomenon that exports from, say, Vietnam or Mexico to the United States went way up and there's a lot of evidence that that stuff is just stuff that's actually made in China.
But then it's assembled in Vietnam and then exported here to evade the sanctions and and even more drastic sanctions like. Stopping NVIDIA from selling its top GPUs into China, those seem to have been largely evaded too. If you look at the NVIDIA earnings, you know, over half of their revenues come from sales to the greater Cynosphere, which includes Singapore and Taiwan, and I think a large number of those GPUs ended up in China.
They didn't end up, some of them ended up in data centers in those countries, but probably at least half or more ended up just being trans shipped. into China, and my sources tell me there's literally no shortage of H100s or even the most advanced GPUs. In China, there's actually a glut right now.
Callum Williams: Right. And you hear all these crazy anecdotes about people getting on commercial planes, right, with suitcases that have them inside and all this kind of
Steve Hsu: Yeah, I mean, if you can, for the average guy, if I, if I can open a fake data center using my apartment in Taipei as the address, and then I order a bunch of, and NVIDIA's very happy to sell me a 30, 000 H100 or something, and then I literally get on the plane with it in my suitcase, and Taiwanese authorities don't care, and then I make 5, 000 bucks, you know, just on the flight, and I have a nice vacation, I come back like, I think a lot of people did stuff like that.
Callum Williams: Totally. And I, you know, I think, I think the sort of intellectual mistake I think a lot of both the Western policymakers who imposed sanctions on Russia and to a degree tariffs on China and sanctions on China, but if we're talking about Russia, we can leave it at that. The intellectual mistake they made, I think, was to believe that we were still in a 90s or even 80s world where the U.S. economy was kind of coterminous with the global economy.
And therefore that is what the U.S. decided if the U.S. wanted to throw its weight around. It could get what it wanted, and I just think, you know, just purely by dint of, of, of GDP becoming, you know, sort of more equally distributed between the kind of global South and the global North, that has changed. and it was kind of, I guess, with retrospect, it was, it was odd that, I mean, maybe, maybe, maybe you'll say it wasn't odd because these people don't know what they're talking about, but in most ways it was odd that people didn't consider that at the time, you know.
So, the idea of Russia exporting more oil to India. So what, like 25 percent of Russian oil goes to India now, something like that. is, is kind of so simplistic that a child would, would have come up with it, but it's almost like that you're so blinded to the narratives that are in your head that you're unable to imagine that world existing.
Steve Hsu: Yeah, I think if, if someone like, let's just imagine some, you know neocon strategist and, you know, in the White House or something and on the National Security Council, you know, like plotting, like, oh, yes, if we can get, if we can just get them to start a war, then we'll sanction them and the Putin regime will fall.
I mean, I think people really thought this at the time, if you and I were confronting that person saying, like, let's sit down together and just do a couple Google searches. And we'll just test whether your view of the world is right and you could have just Googled like, okay, what, where does India get its oil?
And could they, is it realistic that we're going to stop them from getting it from Russia? And then, you know, so, so I think their worldview could have easily been collapsed, but, but they were so insular, like the bubble that they were living in was so insular that no one, I think, effectively challenged them and got them to really, you know, try to sanity check their worldview.
And I think the worldview could almost just be compressed into saying, Okay. As you said, they just thought it was 1990. So in 1990, the post Soviet collapse, the Russian military was shit. It was not going to be able to do anything. The whole society was dysfunctional and the entire global economy was the U.
S. economy, right? And if, if that, if you're just stuck in that era, then you're going to think, Yeah, great, as soon as we put sanctions on them, Putin will just collapse.
Callum Williams: I think that was, I think, so I do agree with that a lot. The other thing that I think was going on was something subtly different, but kind of added to the problem, which I think rightly there was. enormous outrage globally when Russia launched its invasion of Ukraine. And there was obviously enormous sympathy for Ukraine and enormous sympathy for the Ukrainian people.
And I think as a result, it's very hard as an analyst at that point too, you basically really want Russia to get punished for this, right?
Steve Hsu: Yes.
Callum Williams: You really, really want them to be punished. And so your tendency, I think, as a result is to look, you have a lot of confirmation bias as a result. Where any kind of data point that comes out of Russia that looks bad, you're like, right, that's it.
This shows the economy. That's about now. One thing I should say is that I say this with the benefit of hindsight about what happened with Brexit, because in 2016, when the UK voted for Brexit, I, like most people at The Economist, was super anti Brexit. I think it was a really bad idea. And so when the UK voted for Brexit, I know when I, when I sort of look deep into my psyche.
There was a, there was a sort of phenomenon where I was like, I, I kind of want the UK economy to, to, to, to go down the, down the, down the toilet as a kind of punishment for voting for Brexit. And I think a lot of people did that. And I think you're, as a result, you were looking for evidence. that the UK economy was about to collapse.
And in practice, again, that didn't happen at all. And I kind of internalized that lesson being like, okay, when something happens that I don't like, I cannot allow my dislike of that event to influence how I judge the economic data or whatever. It's very hard to do in practice, of course, but I, as I say, yeah, I think that was a big thing that happened in 2022.
And what was the, what was the crazy thing was, and again, I won't reveal names, but so there was this period, the interesting period from about sort of April 2020. to onwards where it became increasingly clear that the economy was, the Russian economy was not, was not collapsing. In fact, it's going to see a sort of mild recession and then by the end of 2024, it was very clear, but even then a lot of the sort of more conventional forecasters on the Russian economy was still predicting a big recession in 2020, 2023.
They were saying, Oh, no, no, no. It's just been delayed. It's going to happen in 2023. And then it didn't happen in 2023. And then again, they said, no, no, no, this time, I promise it really will happen in 2024. And now we're seeing it again in 2025, where if you look at what's been published over the past few days or past few weeks, there's so many articles being like, is this the year when the Russian economy finally collapses?
So I think people still have not been able to separate their dislike for or hatred for the Russian regime, which I think is entirely justified from a sort of sober analysis of what's really going on in the economy.
Steve Hsu: Yeah, I totally agree with you. And I would say exactly what you described is mirrored in the military analysis, like on NATO, especially the UK military side when they estimate Russian casualties or losses. I think they're heavily overestimating. You know, what happened, the amount of damage that the Russians took in all this and a similar thing because they were looking for it.
And so there's a, there's just a little bias to, to take the higher end of the estimate. whereas when it does, the dust will never really fully settle, but I think 10 years from now, when some very dispassionate historians are going over the evidence, I think everything will be adjusted downward on the Russian side relative to what a lot of people in the West believe. this idea of like not being able to do dispassionate analysis just freaks me out because as a physicist, like, you know, usually people don't get very emotional if I talk about, like, what the quark is doing inside the proton, like, you can, you don't get emotional. It's like, oh, what does the data say?
And okay, yeah, maybe the wave functions peak here or something. but then as soon as you go in to get into history or social science or these other things, it's like, yeah. Wow, people just cannot decouple. And if you're caught saying something like, Hey, the Russian economy seems to be doing pretty well, immediately people start accusing you of being a Putin sympathizer or agent or something.
It's just crazy.
Callum Williams: Yeah. I mean, we never got accused of being a Putin sympathizer, I think, because we had had such a strong anti Putin line for many, not just when the invasion took place, but for many, many years before that. I think we were accused of, okay, I think two things. One was being duped by statistics that the Russian authorities were putting out.
So that was, that was a sort of, that was like the last readout of the bears, right? Right. Right. They would say, well, it may look as though unemployment is 2%. It may look as though inflation has come down to 2%. It may look as though consumer confidence is at an all time high, which it is, by the way. it may look as though there's not been much of a recession. It may look as though the economy grew by 4. 5 percent last year, which it did. But you can't possibly trust the Russian data. Now, you know, we've done a lot of work and spoken to a lot of people and analyzed the data ourselves and got input from others. We have absolutely no doubt whatsoever that the Russian data is basically okay, and doesn't materially mislead.
It's also very consistent with all of the private sector data being produced. So yeah, no, I mean, it's, it's, it, so that was our main accusation that we, that we had been duped by the data. But then also, as I say, it was like, there's a kind of temporal critique as well, which is like, you may be correct now.
But you're missing the longer term, bigger picture, which is that the Russian economy is screwed. And there's something unsatisfying to me intellectually about resorting to a full sort of speculative forward looking analysis as your only way of conveying a point. Because, like, in the here and now, what matters is, like, is the Russian economy screwed up or is it not?
And obviously, it's got massive imbalances, running huge deficits, and inflation is high. And so I'm not trying to say the Russian economy is, like, the best performing economy in the world, because it definitely isn't. But the notion that it was going to force Putin's hand was just laughable, basically. So yeah, that's, that's where we found ourselves.
Yeah,
Steve Hsu: going to be something like, Well, you know, if Trump hadn't won this election and Kamala Harris had won it you know, and we had kept a strong, you know, a set of sanctions in place, then, you know, In the far asymptotic future, the Russian economy will definitely, you know, collapse, not, maybe not collapse, but they're screwed, right?
And so we were actually right. It's just unfortunate this crazy Donald Trump guy got into office and
Callum Williams: It's possible. I mean, you know, there's a, there's a, you know, I equally know I've, now I've kind of staked out this position of being like, quote unquote, relatively bullish on the Russian economy. There's a risk, of course, that you overcorrect emotionally and you actually don't see the. you don't see the bad data coming out.
I mean, one thing that did make me question when there's two things that made me question how the, how the economy is going to do in the next kind of year or so, one is that there were quite materially tighter sanctions that were implemented in the autumn of last year on like Gazprom Bank, which basically is responsible for processing a lot of payments of exports of oil and gas, whatever.
and that caused the ruble to massively depreciate. So people were, you know, investors were scared by this. And then the second one was that there were new sanctions implemented on Russian oil tankers, which were much more serious. Now there is a world in which that does start to bite. And the data would, the latest data would suggest that like Russian oil flows are being materially, I'll say in a, that's a vague word, but materially disrupted by the tanker sanctions.
So like. It could turn, it could turn, but like the latest data that we're seeing, I'm seeing points again, GDP growth in the region of three to 4 percent real times, which is not bad. Like it's like Germany is at zero. The UK is at zero. France is like one. It's like, it's not, it's not actually that bad,
Steve Hsu: Right. Right. So before we leave the Russia, Ukraine business, any, any, do you want to make any predictions or things that you suggest the viewers look out for?
Callum Williams: about the conflict or about the economy,
Steve Hsu: Either one, anything.
Callum Williams: either one. well, I think on the, I mean, I think on the conflict, I suspect that some kind of arrangement will be, will be, will be reached. Yeah. I mean, my concern, I suppose, is that, like, Putin's time horizon is just much longer than Trump's and so I think, notwithstanding the critique of always looking too distantly, distantly, distantly into the future, like, I think it would be worth sort of wargaming in your head how sustainable any kind of deal that's reached between Putin and Trump will be, either to the benefit or to the detriment of Ukraine, I don't know.
But I think, you know, I, I suspect not this, this won't apply to The Economist, I, I'm, I'm sure, but the tone of a lot of the news analysis, if and when it is reached will be, this is the final act of the play. And, you know, that's just not going to be the case.
Steve Hsu: If it turns out that we sort of normalize relations with Russia, we kind of cool off the Ukraine conflict and the Trump administration really pushes for a Sino Russian split, okay, so let's suppose that, I mean, that's one possible future that could happen here can you model in your own head, because you're much more familiar with their psychology than I am, how the UK and European elites will react to that new world? So, so, so suddenly, suddenly it's like massive charm offensive with Russia because any small thing that is a difference of opinion between Trump and them, they're going to paper over because they want Putin to basically start doing things which are not threatening to China, but we, you know, are weaken that alliance significantly, but in that world, we're just, Trump just doesn't want to hear any bad stuff about Russia, right?
He's, he's like, I'll forget that, right? yeah,
Callum Williams: I mean, I suppose to slightly change the terms of the question. I mean, my sense was that the economics of the Sino Russian alliance were always massively overstated. I mean, not least because China is just so much bigger in economy than Russia is. But like, even if you look at networks of investment and networks of trade, they were still really tiny after, even after this supposed alliance.
So like, you know, it was one of those things that like the optics massively dominated the data. So in that sense, I would say probably nothing will really change. I mean, again, and even like there was this period in the UK when I was living there in the 2000s, where like everyone went mad for Russia, everyone loved Russia, like Roman Abramovich took over Chelsea, all these Russian people moved to like Hensington and Chelsea and Knightsbridge and stuff.
And it became a kind of cultural meme that like there were Russians everywhere in London. But again, if you, if you, if you look back at the data and like, well, actually, how much did Russia truly invest in the UK? It was tiny. It was like. You know, the stock of investment in the UK economy was something along the lines, it was like 0.
5 percent of my total foreign investment in the UK was from Russia, like it was really, really small. So I guess if I, if I were to, my kind of, what I would urge people to do if that starts to emerge is to, is to actually look at the economic data. Look at the bilateral data and see how that's changing because it basically didn't really change after, after this supposed alliance between Xi and, and, and Putin.
Steve Hsu: Great point. My view of this was not that the economic tie up was the primary thing. I think for Chinese strategists, it was always that Russia was, if Russia maintained an antagonistic relationship with Europe and the United States, that gave the Chinese a much better option in the event that things got hot.
So in the event the Americans tried to block the Malacca Strait or a hot conflict. Pacific, having Russia on your side is just useful, right? And if, if Russia is still at loggerheads with the U.S., they're not going to let China go down because if China goes down, then Russia's next, right? So then like their nuclear arsenal becomes Not like you control their nuclear arsenal, but if the Americans launch on you, the Russians are gonna launch on the Americans, right?
You can be pretty confident that's gonna happen. And or if the rush, if a hot naval campaign breaks out, the Russian ships are gonna be supporting them. Chinese ships, right? And submarines. And so it was, it was all like, kind of tail risk kind of stuff. But all these military planners live in the tail risk world where things get really hot.
And, and so it wasn't really like, oh, we're, we're doing X billion more trade now with Moscow. That's important. That's actually not the important part. But if we really needed the oil, we needed something. You know, fuel from the Russians, because the U.S. is succeeding in cutting off what we buy from Iran.
Being able to rely on that a little bit, in, in the crunch, allowed them to plan differently. I think that, I think that's really more this, where this stuff lies.
Callum Williams: Totally. I suppose one thing I, the only thing I would say to that, and here I'm venturing far from my expertise, but my kind of model of Russian Chinese relations going back over history was that they were always, they were always in a very unstable equilibrium, right? I mean, the, the, the split in the, in what, 1968 or whatever, over nuclear weapons was like a massive event.
And I think even now they've still got big disputes over some part of northern China. Like, is it Russian or is it Chinese? Some islands or whatever. So, I mean, I, I, but then again, this is all, I think as an economist, I'm sort of slightly, I sort of end up believing that a lot of this stuff is quite word-sell.
Like, it feels a bit kind of geopolitics, hand wavy stuff. but no, totally, totally take a point about, about the tail risk stuff.
Steve Hsu: Yep.
Okay, let's move on to the topic that actually got us together, which is U.S. inequality.
Callum Williams: Yeah.
Steve Hsu: for the listeners, I, I tweeted, retweeted a graph from the Fed. Which looked at the amount of wealth held by different segments of the American population. So, you rank the individual families, say, by their percentile of wealth.
And then you aggregate within that band. So you could say, like, oh, all the people that are in the bottom 50, 50, 50 percent of families for wealth. So they're the 50 percent least wealthy families in the United States. You aggregate all their wealth. And that comes out to be very close to zero, right?
They have very, you know, because a lot of them have debt and et cetera, et cetera. But anyway, their aggregate wealth is very close to zero. And then as you look at more and more wealthy chunks, like 90th percentile families or 99th percentile families. They start taking up more and more disproportionate chunks of the total wealth of the United States.
And so that kind of graph makes it seem like the American, America is kind of a really dystopian place because like, oh, half the country has like little to no wealth. And of course, it's much more nuanced than that because of course, a lot of the people in the bottom 50 percent are young people who haven't had a chance to accumulate any wealth.
So it's, it's, it's sort of misleading in that way. But I got a lot of pushback on Twitter just because people didn't like, you know, some people just don't like. Any favoritism toward equality or, or, or, you know, redistribution. So people just got angry at me for even pointing this out. Callum wasn't one of those people, I think.
But, Callum probably had a much better idea of the actual numbers you want, which is, you know, control for age, control for a bunch of other things, and then look at wealth inequality. And it's not as bad as that Fed graph showed. But Callum, maybe just tell us what, like, how would you explain if some Martian landed and they wanted to know, like, well.
How unequal is America? Like what would you tell this Martian?
Callum Williams: Well, I think you summarized it really well, and I, and your tweet was, was, you know, factually completely correct and you, and you, and you added bits to it as, as the conversation went on. So you didn't make any mistakes or anything. And it's, it's very complicated. I mean, I think what, you know, one, I think one extreme argument that one can pursue is to say that wealth is kind of a meaningless measure to begin with.
So, you know, you know, there is an argument to say you shouldn't be interested in wealth to begin with. One reason for that is, as you alluded to, there will be many, many people. Well, let's start again. What do we really care about? We, we, we care about the capacity to consume different people.
That's really what we care about. Like, if they want to, what kind of living standards can they attain? Person A, what kind of person, what kind of living standards can a person achieve? Person B, right? That's what we care about. Now, one issue with wealth. is that there will be a lot of people who will, you know, have taken out a large debt on an absolutely enormous house.
3 million, 5 million will be paying that back, or they're paying into their private equity funds and they've got a huge liability there, you know, and so they will have deeply negative net wealth, but will be by any reasonable measure, extremely well off, right? So I suspect that most general partners, maybe not general partners, most partners in Silicon Valley, VC and private equity funds probably have.
close to zero or highly negative net wealth. So, that illustrates the conceptual difficulty of focusing on wealth in the first place. Then there's a question of, of, of, of, of measurements or really a question, I guess the second question is, what do you include? And you allude to this in your tweets.
So when people think of wealth, they think of the Donald duck jumping into the pool of gold coins in his dungeon, which is a form of wealth, but of course it encompasses lots of different things. Encompasses housing. but it also, as you said, encompasses things like social security, and obviously what you've had over the past, I don't know, a few decades, 50, 60 years, is a huge expansion of the welfare state.
Um, not just in Europe, but also in America. Now, I think by any reasonable standard, that should count as a form of wealth. It's a, it's a sort of pot of money that's sitting in some account or it's some kind of promise being made to you by the government. For flow of income to come in your direction, that is basically what wealth is.
It's a stock of money or a stock of value. So I think you therefore need to have a question about what counts as wealth. Like, do you count, do you count social security? Do you count company pensions? Or do you count, you know, the promise that the government has made to give you food stamps if you, if your income falls below a certain level?
I mean, that is a form of wealth also. And then there's the question of how you measure it. Now, to begin with, there are only. Three or four countries where surveys on wealth even approach the level where it would be reasonable to estimate how much wealth a given person has. I think right from memory, those are the U.S., UK, Denmark and France, I think.
And basically elsewhere, the surveys just don't exist. You know, you can, in theory, sometimes look at people's tax returns, but that's difficult because people will often misestimate wealth and often don't report on tax returns either. So you've got all these kinds of big problems. And so as a result, my view, I suppose, is that it's sort of just a bit too complicated and you should really be interested in income or even ideally like expenditure inequality.
Like how much can people actually consume? So, so, so, so what you get, what you've, what you've had in the U.S. is a lot of work by Piketty and Saez and Zucman. Those are like them, those are like the head honchos of, of, of wealth inequality studies. And, you know, that their work has been enormously influential, obviously, and it's shown that, you know, over the past few decades, really since the 1970s, mid, mid 1970s, early 80s wealth inequality, as measured by, say, the top 1 percent share of all wealth, has, like, exploded, it's gone up loads.
But I think there now, more recently, really over the past 4 or 5 years, has been a, I think, a quite convincing pushback on both the wealth stuff, and on the in, income inequality stuff, to say that, You know, hang on, once you actually think about the difference, the difference in the sort of balance sheets between rich households and poor households, things change.
So I think the easiest disagreement to, to understand is, is sort of as follows. So the way that studies estimate wealth when they cannot observe wealth directly is to look at the income from assets and then multiply it up by an assumed interest rate, right? So, you know, if you've, if you've got 10, 000 but you assume a 5 percent interest rate, then, you know, multiply it by 20, 200, 000.
That 200, 000 is the wealth of which you get a 5 percent interest rate, so you get 10, 000 a year. So it's a sort of, it's a calculation, an estimation of wealth, rather than directly observing the wealth that someone has, like, to hand. Now, the assumption that, yeah, so the, the assumption, well, it could be stocks, could be it could be bonds, could be just bank accounts.
The assumption that has typically been made in the Piketty work, is to assume that the rate of return across households and the income distribution is the same. So it's 5%, 5%, 5%, 5%. In reality, it doesn't quite work that way because often people at the bottom have poor financial literacy will put all of their income into a bank account that might pay almost no interest, right?
Well, maybe it pays 1 percent interest, or maybe they buy a certificate of deposit and it pays 2 percent interest. Whereas the very rich are disproportionately likely to put all their money into things like equities or corporate bonds or something. where the rate of return is typically higher. So if the rate of return is typically higher for rich people, then once you back it out, the assumed amount of wealth that they're deriving that income from is much lower.
So you get a lot of these kinds of adjustments that are made. And once you do all that what you, what you typically find is that there has been a kind of U shape of wealth inequality in the U S and lots of other countries where it was super, super high in the kind of great Gatsby days. And then it fell off during world war two.
And after governments, you know, they put in capital controls and wealth taxes and all this kind of stuff. And then, since the 1980s, it has been coming up a bit, but to a much, much, much lower degree than what the Piketty research would suggest.
Steve Hsu: So, just to recap that for the audience. So, so, you know, I remember, it probably has been like more than 10 or 15 years. When did Piketty's book come out? It was a while ago, right?
Callum Williams: Yeah, I mean, I mean, the really famous one was like 2012,
Steve Hsu: Yeah, so almost 15 years, 10, 15 years, so yeah, that had a huge impact and I remember looking at it and being very, you know, struck by the figures in the book, but I just wasn't sure how much I should trust his inferences and so now you're saying that there's been a fair amount of empirically grounded pushback against his,
Callum Williams: I would say there's more discussion over the income inequality figures than the wealth inequality figures. My sense is that academics are generally. generally believe that the income inequality stuff is more fruitful, basically because there are fewer sorts of imputations you need to make, because in theory, at least, at least in the U.S., somebody will report all of their income on their tax return.
And so you can look at what someone's income actually is, whereas for their wealth, they don't report all of their wealth on their tax return. So if you're interested in looking at tax returns, which you kind of need to do to understand the finances of the very wealthy, then you're kind of pushed in the direction of analyzing income inequality rather than
Steve Hsu: Yeah, you know, I, I have interest in both of those things. So I'm interested in the ultra high net worth world. But when I put out what I was actually interested in, I was just looking for a figure that captured the level of precarity. Of the really poor. So you, you read these, you hear these factoids like, you know, X percent of Americans are one paycheck away from being out in the street and homeless.
Right? And I was actually just trying to get at that. Like, like, what if it's 1 percent of the population that's in that situation and it's actually because they have schizophrenia or something. That's a very different situation from like, single mom who is working really hard, but if she loses her job and things just, a couple of bad breaks, and she's living in a car with her kids.
I don't want to live in that country. I want to have a country where that second scenario isn't happening so much. And I was, so my, my posting of the Fed numbers was a very poor way of getting at that question for various reasons. But after some investigation where we, somebody finally found data where it was normalized by age and all this other stuff still does seem like 10, 15, 20 percent of Americans, I don't know where the number is, maybe it's closer to 10 than 20, but Some kind of uncomfortably high fraction of the population is living a kind of precarious existence.
And so that, that's, that's kind of what I was getting at. And then I, for the people that I interacted with on Twitter, I was like, look, let's just figure out where it is. And you tell me if you're comfortable with this, right? So you could, you could say I'm totally comfortable with this because, you know, after all, without that, we wouldn't have, you know, breakthroughs in AI or whatever shit you'd like, right?
Like, oh, we couldn't have the dynamic good stuff if we weren't like. If we didn't have all this precarity at the bottom, which I don't necessarily believe, but people could make that argument. Right?
Callum Williams: Yeah, totally.
Steve Hsu: Yeah. So do you have
Callum Williams: I guess what I would say,
Steve Hsu: is the
Callum Williams: Yeah, it's a very interesting question. I mean, I think the, I think the question is, so I guess what I would say is that using wealth inequality or wealth data to get at that answer is tricky. And the reason why that is, is because wealth accumulation strategies are heavily influenced by the welfare state.
So, to give you an example, Sweden has extremely high wealth inequality. Now Sweden is, in the eyes of many people, a kind of socialist welfare state paradise. So the question is, why is that? And the answer is really because If you're at the bottom, there is not that much of an economic incentive to accumulate wealth, because you know that if you face financial difficulty, the state will step in.
So there's no need to insure yourself personally, because that's the purpose of social insurance. Now, what you've seen in the U.S., as the welfare state has massively expanded over the 80s, 90s, and 2000s, is that the household savings rate has come quite markedly down. Now, people often point at that data set or data series of a falling household savings rate and say, Americans are becoming like spendthrift or frivolous or whatever.
In fact, I think a large chunk of that reflects the fact that there's just much less need to insure yourself privately now. You can insure yourself publicly. So that would be the first thing. so I think we need to look at another data set. So the, the other one that, sorry, you go, you go.
Steve Hsu: Can I just drill down on that?
So are you saying that these people who are in danger of being kicked out of their rental because they can't make the rent are somehow knowing that they will be saved at the last minute by the government? Or, because, I mean, maybe that scenario that I have in mind is not, maybe it's only because, like, Hollywood filmmakers like to make movies about this happening to people, and maybe it almost never happens.
Is that, is that the answer? Or, or like, does the government actually save these people from becoming homeless? Because I do see a lot of homeless people in the
Callum Williams: Well, yeah, absolutely. And we can talk about why that, why that may be. I think it's wrong to ascribe conscious motivation to economic decisions always, um because people make decisions about pricing and about saving strategies based on a lot of unconscious or subconscious things. But no, I think that's right.
I mean, I think it is important to remember that the, in, in the U S the, the yeah. The welfare state has massively expanded, right? So it's like 1 in 8, 1 in 8 either people or households in the U.S. is now on food stamps. Medicaid has massively, massively expanded in recent years. And, you know, it was only during the pandemic that the federal government was, was, had had, you know, increased the rates of unemployment benefits to 600 a week, right?
So it was, it, it, no, there's been a pretty substantial, substantial change. And I, I think it's not unreasonable to expect that that would affect. economic, economic behavior you know, I mean, look, there may also be evidence of more kind of spendthriftness than there was before as well. I mean, it's, these things are obviously overdetermined, but I know I, I, I, I, I do think that that could explain it.
But I think the broader point is that there are probably better ways of capturing what you're trying to, what you're trying to look at. I mean, obviously the best way of doing it is actually to look at flows. In and out of homelessness, which is difficult, but possible. I mean, one that people often look at, which has its own problems is like, and you alluded to this, like when the fed, there was a fed survey that, that asks people, I forget the exact amount, but it's something along the lines of if you had to produce 600 or 800 at a moment's notice, could you do it? and that, you know, the answers to that are, are, are quite high. but you know, there's questions about how that survey is designed, that kind of thing. And then the other question, which I really do have a problem with is, do you live paycheck to paycheck? Because some insane proportion of U.S.
households, including U.S. households that earn well over 100 grand a year, will say they live paycheck to paycheck. So there's a question of like, is it really a meaningful measure of economic precarity? I would say, I would say probably not.
I mean, for me, I think the thing that is one of the great successes of the U.S. economy and indeed many other economies in the rich world over the past really 10 years is that it has become very easy to find a job by historical standards, right?
You know, the unemployment rate in the U.S. is like 4 percent or whatever. It's pretty low. The unemployment rate in most rich countries is, like near or at an all time low. And the unemployment rate, like the share of people actually in work, is at an all time high. And so if you're thinking about, like, what are the channels through which people are completely unable to meet their, meet their bills, one of, obviously, probably the biggest way in which that happens is because they can't find work and that has become less and less likely, not more and more likely over the past few decades.
Steve Hsu: it. Well, on that 4 percent number though this is, this is probably a really basic question that, you know, shows that I'm not a real, I'm not an economist at all.
Callum Williams: Yeah.
Steve Hsu: There's some other figure, which is like the fraction of working age people who are actually engaged in work. And isn't that alarmingly low, like, are there a bunch of quote, discouraged workers that just aren't counted in that 4 percent or something?
Callum Williams: yeah, so this was much more of a problem in the sort of aftermath of the financial crisis. It is true that labor force participation among both men and women in the U.S. It's quite low by international standards. I mean, there's like a crazy stat, which, which was true the last time I checked it a few years ago, which was that, like, The labor force participation of Japanese women is higher than American women, which, you know, in the Western stereotype of like Japanese women being like very homebound and kind of
Steve Hsu: Yeah, that's surprising.
Callum Williams: completely blows that up.
But that has again, that has changed in recent years because the labor market's been so strong. I suppose the only other wrinkle to this is that I would emphasize that the U.S. is very unique in this regard for having a kind of fairly low decline or decline. labor force participation rate.
If you look at almost any other rich country, you don't have these problems, right? So it's a particular U.S. phenomenon. And there's this question of why it is like some economists have argued it's video games, which obviously can't be correct because there's video games outside the U.S. It could be the opioid stuff and the painkiller stuff.
I mean, that seems plausible. The China shock, some people have said that it's that, you know, competition from imports from China has destroyed various towns and cities in the U.S. like maybe, but like, obviously. Lots of countries have faced this kind of shock and had this problem. So there is a bit of a puzzle there, but no, the U.
S. is not like the best economy in this regard, but things have been improving recently.
Steve Hsu: It's amazing how, you know, like, obviously, you're a professional, like, following these kinds of issues. I'm just an amateur. But, you know, even between us, we don't have a real feeling for, like, if I go to some Rust Belt town, a little town in Michigan, I meet some 27-year-old who should be working, and he's vigorous and, you know, healthy.
Turns out he's living in his mother's basement and playing video games all day. Is that real or is that just some story that the journalists find and they blow it up? Like we have? No, I, we have no idea. Like, is it 1%, is it 5% of 27 year olds? I just don't have any clue.
Callum Williams: Yeah, it's a really interesting question. It's also that that's become much more of a thing in the U. K. as well. In recent years, this idea that I mean, you've had this like in the U. K. I know this for sure. They've had this amazing surge in people taking time off work for a long time because of mental health stuff, a really massive kind of like booming.
In doctors giving people notes to say you, you are excused from work for three to six months or whatever. Now, I mean, I'm open to the idea that this is the product of some kind of like, you know, generalized like social, societal degradation linked to, I don't know, what it may be, social media or technology or whatever, but No, I agree, we don't, we don't have a, we don't have a great insight into this, but You know, I think it's, it's weird because for the, for the longest time, In economic policy, giving, getting people into work was like the number one goal, you know, it was the thing that kept politicians up at night quite recently, right?
You know, there was a period in the early 2010s when the U.S. unemployment rates wouldn't come down, just wouldn't come down, and that has now kind of gone away as a topic. And, you know, if you look at what happened, you know, look at what Trump was tweeting about when he was president last time, for like, say, 2017, 2018, 2019.
He would always say jobs, jobs, jobs, right? Block capitals, jobs, jobs, jobs. He doesn't tweet that anymore because it's now seen as assumed that it will be quite easy to find a job. So it's a really interesting change that's taking place.
Steve Hsu: I still have this residual suspicion that it's just there are a lot of people that are not counted in that 4 percent thing and we've just like, we just like push them off to the side. Like, well, we don't worry about them, but for everybody else it is 4%. Great, you
Callum Williams: Yes, yes, yes, that is, that is, that is more true for the U.S. than it is for, for, for other people. Then again, on the flip side, the U.S. has created loads of jobs that probably are measured badly by the stat, by the stats people, because they're given to people who are not here legally, And so there is this very interesting question about, like, how many jobs, how many workers are there actually in the U.S., we don't really know.
Steve Hsu: Yeah, let me, let me close this topic off by pointing out I'm revealing a little bit of your personal information here, but I believe you live near the marina in San Francisco.
Callum Williams: That's true.
Steve Hsu: Okay, now, that's like one of the most expensive, upscale, amazing places in the world, but also because you live in San Francisco, you don't have to go very far before you could, you could step in some human feces, see some needles on the ground, maybe witness a store being you know, emptied out by a flash mob or something like that.
So how do you feel about, like, this, getting back to this precarity question, and is the U.S. a polarized society of big winners who live in the marina? People living in a dystopia or is it, or is that just overblown?
Callum Williams: Well, obviously, in terms of one's lived experience in San Francisco, it definitely is not overblown. because there are, you know, there are large chunks of the city that are, you know, the kind of wealth generators of the city. I'm not one of the wealth generators. I'm a journalist, so I don't generate anything, but the wealth generators, they don't go there.
You know, they don't, they don't really go downtown. They don't go to the Tenderloin even though it's very close. The question of why San Francisco has so many people living on the streets and taking drugs is obviously a very complex and difficult question. I have gone back and forth over what is the true cause of this.
I was, for a long time, very convinced by the argument that it was all to do with housing. San Francisco has extremely expensive housing, which is undoubtedly the case. And as a result even if you are in a pretty well paying job, it is very difficult to afford a house and it's the pathway from even a minor disruption to your income and not being able to make the rent strikes me as quite logical.
Um, however, I have more recently after reading and living here for a few years, come to the view that housing plays a role, but perhaps more important, is basically the role of drugs. Um you know, you can see people who are mentally, physically in absolute, complete crisis. These are people who are San Franciscan, like natives or whatever, whatever that word you want to use from San Francisco.
But I think a lot of them are not. I think a lot of them are drawn despite what a lot of the nonprofits will say. A lot of them are drawn to San Francisco because both the supply of drugs is plentiful and also there are lots of well meaning, but I think misguided non profits who will, will support your decision to take drugs and to live on the street.
Now I don't think for those people, giving them a home is the solution or it's not the only solution. I mean, they did this during the pandemic where there were lots of hotels in the Tenderloin and downtown which were empty because no one was coming to San Francisco. They did actually give people a home. You know, the city government did actually turn over a lot of these hotels. It has not made any difference to the problem whatsoever. If anything, it's made it worse. A lot of these hotels were, to be honest, completely destroyed as well. So it's really not, not working. So, I mean, the question of what you do about it now we have this problem is even more thorny, but I think, I don't see what's happening downtown San Francisco as a product of American precarity, to be honest.
If, if anything, it's, there being a lot of resources, not, not too few.
Steve Hsu: Good. Let's move on. Because although we have like two hours, there are so many interesting things for us to discuss.
If we could, we could fill the whole two hours with just any one of these. So another topic we want to discuss is European divergence. You know, any number of metrics, GDP per capita, wealth, et cetera.
One of the striking things for me, because I, by coincidence, my wife was a visiting professor in the Netherlands in the fall, so I was traveling a bit in Europe and I saw a lot of things there you know, you hear people talking about their net income, which means the income they take home after taxes.
And even in sort of rich European countries, let alone like Greece or Italy or something. I was shocked at how low the net, what was considered a healthy, you know, upper middle class net income per month was, you know, it could be a couple thousand three thousand euros or something like that would be considered pretty good, even for a professional class person.
Um, and for Americans, it seems like, wow, these guys are really like, I hate to say it this way, but you guys are poorer than the people in Mississippi. Right? And so that's how the debate often emerges. Enlighten us about this. What is really going on? What's going on? Divergence between the U.S. and Europe.
Callum Williams: Yeah, it's such an interesting question. So I guess I would say two things by way of context. The one is that there are, there are ways in which the U.S. has clearly recently diverged from the, from the EU. As you say, GDP per capita is one, GDP growth since the pandemic asset prices have really gone bananas.
I mean, I'm thinking of the stock market really has, you know, they've really gone bananas in the U.S. They haven't done that in Europe. You know, investors are basically treating European equities like bombs, like they, you know, they, they sort of pay out some dividends, but they don't really grow. and so that's happened.
The second thing I would say is that the U.S. has really since, I don't know, the 20s or the 30s, been richer than Europe, right? And that's, so it's not just because of World War II, but it has been richer than Europe, richer than the UK for a long time. I mean, if you talk to people, I mean, I had a good conversation with someone at Stanford who had done, who was a Rhodes Scholar at Oxford in the seventies and asking him like what it was like to travel from the U.S. to, to Oxford in, in 1973 or whatever, like it was, he was like, you know, it was like going, it was like a different planet. Like it was so much poorer than America. So that's, that's existed for a long time. The question has, you know, has the U.S. become substantially richer than Europe in the past few years?
I Which is the kind of Twitter kind of zeitgeisty narrative at the moment is less clear. I think what has really happened is that, I suppose it ultimately depends on what you're interested in. But the thing that I'm interested in is the best measure of living standards.
Like what, again, it's getting back to this question of capacity to consume, like what really matters is what is the bundle of goods and services you're able to afford. And the best measure of that, I believe, and I think I'm right to believe this, is something called actual individual consumption, which is something that the OECD and other stats people produce.
And it basically says how much, what is the market value of what someone buys with their own cash, with their post tax cash, plus an assumed market value for all of the public services they consume. So healthcare, if it's, if it's socialized. Public education, public transport, the quality of the roads, all that kind of stuff.
And it's actually measured pretty accurately and, you know, at, at pretty high frequency. And what you see there is that over the past kind of like 10 to 15 years, Europe has actually kept, you know, has actually, you know, kept up with America pretty well. And in some, in many cases, it's actually improved relative to America.
So to use, in fact, to use your example of the Netherlands, which is where your wife is, the actual individual consumption per person in the Netherlands is higher than in the U.S. Now, I, I know that that might be sort of odd to, to square with the, the conversations you were having over there, but you know, the value of, of things like public education and healthcare is very, very high, in the Netherlands, you know, people basically, you know, in the U S you have to take a big chunk out of your salary to pay for your health insurance, like the same pressures just don't exist in the Netherlands.
So that helps to close the gap in a big way. I mean, you know, I think the outlook for the European economy, that said, is pretty bad. You know, despite all of this, these debates that just go round and round and round and round in circles, you've got a number of big structural political impediments to the EU becoming a genuine single market.
The U.S. is a genuine single market. And I think that fact alone will explain more and more the divergence that probably will start to take place with the U.S. really, really keeping going. Whereas the EU is going to be kind of trapped in a low growth system. So I suppose I'm, I guess to sort of summarize what I said, I'm more optimistic than you are when we're looking at history and looking at the current day, but I'm more, I'm quite pessimistic when we're thinking about the future.
Steve Hsu: Got it. You know, the, this, the divergence has reached a point where, like, even when we were, when I was with other professors, professor at the University of Amsterdam, having dinner with me after I give a seminar, and there's a bunch of physicists there, and they're aware that, like, U.S. salaries are, like, way higher, like, the Nobel Prize winner You know, the University of Amsterdam is paid like less than maybe paid less than me.
Okay. And, they were aware of that and they started to cope. They and, you know, like, of course, yes, they do get free health care, but I also get free health care through my job. Right? So, so it's like. It's like, but they would even resort, they even resorted to things at dinner, like saying things like, yes, but you know, have you been to a grocery store?
You know, its food is much more expensive in the U.S. and the Netherlands. And I'm like, well, like, yeah, it's like 20 percent more expensive, but frankly, it's not a big deal in my monthly budget, you know, and then they would be enraged by that or something. So it does seem like something's going on.
it could just be that the euro, you know, it could a lot of this just could be affecting stuff where the dollars are very strong right now in the euro is very weak. So, um,
Callum Williams: Yeah. I mean, that definitely explains the different GDP per capita or GDP measured in, in, in, in in dollar terms for sure. but you know, again, bear in mind, like, there was, you know, in the, in the 90s, right, particularly in the 90s, there was this massive exodus of British academics to the U.S..
You know, I'm thinking of Neil Ferguson, Adam Tooze, people like that. They all left because the salary was so much higher. So, and that was a point in the 90s was a point where the UK economy was actually doing really well. So, you know, these differences have persisted for a long time. They have, I don't think they've got materially worse in recent years would be my guess.
And of course, I mean, I don't know, I mean, you'll, you'll know more about the Netherlands university system than I do, but like, my guess is that a lot of it's publicly run. And obviously the kind of salary pressures that are going to be much greater than in a lot of the private U.S. universities. And then even the public ones have to, like, match salaries to a much greater degree than they would, then would be the case in, in the E.
Steve Hsu: Yeah, the system is much more competitive here. There's much more poaching people, which, which
Callum Williams: I mean, I, I, I go back and forth with this topic because, you know, I'm British, and, and go back and forth between here, between San Francisco and, and London all the time. And You know, I, I, I suppose it, it, it's absolutely undeniable that you can earn a much, much, much larger salary, particularly in San Francisco, but really anywhere in the, in the, in the U.S. and actually, you know, like, I think a lot of the public services in the U.S., the kind of classic quote from the, from people in the E. U. is that, or in the, in Europe, is that like, U.S. public services are all terrible, which I don't think is actually true, to be honest. Like, it isn't actually that, that isn't the case.
Like, the public transport in San Francisco is actually pretty good. So that, but that said, like there are a lot of, you know, I think the, I think most EU cities are like way, way, way nicer to live in than most American cities. I think you are, you are much, much less likely to be shocked and like killed and killed in a car accident and all that kinda stuff.
So ultimately it's very hard to make this comparison. Also, by the way, I always lose weight whenever I go to the uk. There's something in the food here, which is very, very weird. Even if your lifestyle doesn't change at all, like you always lose weight.
Steve Hsu: they give you huge portions in the U.S. and they don't in Europe necessarily. And also, like, there's so much walking because, you know, you're walking to the tube and I think you've just burned a lot of calories. Getting around in Europe, which is healthy, which is good. I would say that, you know, I really love being in the Netherlands, you know, because my recollection of it is like riding around on a shitty bike, slightly buzzed most of the time.
And of course, no risk, no violence, no nothing. And then you get on this really nice bus or Metro system. Everything works. And I had real just sort of dislocation because, because I spent time in Europe. I spent time in the U.S. And I also spent some time in China in the fall. And there the infrastructure is even better than in Europe.
And so it's just crazy reconciling. Oh, how much do people know what material well being here versus here versus here? Because there are these important systematic differences and also you know, purchasing power adjustments that need to be made. It's just crazy.
Callum Williams: Totally. I mean, the, the, I think one way I've thought about answering this question is I think the only or the cleanest way of measuring this is looking at, it's not perfect, but it's looking at the number of, if you're, if you're considering two countries, right? So like the UK or the U.S., how many Britons live in the U.S. versus how many Americans live in the UK?
Because those flows, I think, will indicate which country is genuinely better to live in. Now, the problem, I mean, the slightly confounding factor is that the immigration regimes differ between the two. So it might be easier to move, you know what I mean? But it is the case that way, way more British people live in the U.S.
Steve Hsu: Yeah.
Callum Williams: then vice versa. So, to me, that would suggest that life in the U.S. is better than in the U.K. So maybe you could do that for other countries that you're interested
Steve Hsu: It's a revealed
Callum Williams: I mean, reveal preference. I mean, you know, you've got loads of Chinese people, like, desperately trying to get into the U.S., right? So that, to me, says that life in the U.S. is still better than life in China, despite the better infrastructure, but, you know.
Steve Hsu: What's going to happen in the future? So I think you already touched on this. So, and I think your main driver was because they're really not a unified economy. They're actually balkanized into small states, with their own regulations and et cetera, et cetera. There will continue to be a big advantage in growth terms for the U.S. economy. Is that the main driver? Are there other drivers that you want to point to?
Callum Williams: Well, I think that's the thing that ultimately matters. I think, you know, if you're a company that is in the EU that wants to grow quickly and needs to get a lot of capital, right, you just can't really get it in the EU, you have to get it in the U.S.. The capital markets are just so much deeper in the U.S..
And so your incentive to list there and to set up shop there, there's just much higher. I mean, you know, you taught, I mean, I talked to people in San Francisco founders and, you know, founders in Europe, and they will say, you know, we're growing really well in the UK. We'd love to expand into other European countries, but it means we now have to have, like, we've got to translate everything into French, translate everything into Italian, translate everything into German.
We have to make sure we're complying with all the regulations in each of these countries, whereas in the U.S., you just set up shop and you just go, boom, because it's all basically the same, right? I mean, the great, great problem with the single market, which has been the case for a long time, maybe to 15 years or so, is that they have, they have completely failed at reducing the sort of tariff barriers equivalent on services trade, right?
You can now send goods completely freely between all the EU countries, which is great. So, you know, Germans don't pay tariffs on French wine or anything. But, you know, a German trying to set up a legal service in France, for example, like just, just still cannot, cannot do that. So. As a result, like, you know, for the kind of companies that like to drive innovation, they're just not interested in expanding in the EU.
That's like, that's a big problem. And, you know, I think even with the best will in the world, and even if they do manage to get these tariffs, these non-tariff barriers down, the language barrier is still a problem. You know, that is still a
Steve Hsu: Oh yeah, when,
Callum Williams: they're kind of consigned
Steve Hsu: Yeah, when I talked about Balkanization, language is one of the big factors too. It's not just the legal regime or the regulation, it's also language.
How much of the delta between the U.S. and continental Europe is just that we culturally have somehow embraced, rightly or wrongly, a very optimistic view of risk assets and risk taking.
Whereas like, you know, apparently Germans don't own any stock, right? So they don't even own houses if I'm mistaken.
Callum Williams: They don't own
Steve Hsu: Like if I have a culture that doesn't believe in Taking on, you know, or doesn't believe in the equity risk premium. If you know what I'm saying, you know, they just don't believe that they're going to, it's going to pay off for them to tolerate a little short term vol, then they're screwed, right?
Because how can they grow their economy, right? They're not going to have deep capital markets, right? So some of this is just like cultural storytelling to yourself, or maybe a realization that you know, markets deliver returns, excess returns or something. I think there's a big sort of sociological part of the story.
Callum Williams: There totally is. And it's, it's kind of worse than that because. You know, the money that's not invested in equities and stuff is then invested in government bonds. So you then reduce the cost of your government borrowing, which means you fund the welfare state expansion and that kind of thing. So it becomes this kind of circular problem.
Um, which is, which is definitely a big thing. I mean, I guess, I'm always struck by being in the U.S. And particularly in Silicon Valley with how motivated people are. And it's kind of cool in a way. It is also kind of lame in a sense, because I mean, everyone in Silicon Valley has enough, right?
If these people were going to apply themselves to learning about, like, Thomas Aquinas or or or Boston or something, I would totally support that. But like the kind of acquisitiveness, you know, John Stuart Mill talked about this a lot when he visited the U.S. Right? That kind of acquisitiveness. That needs to be constantly richer and more powerful is both amazing, but also quite corrupting.
Whereas, you know, you go to someone in Italy or France and, you know, you go to a restaurant, you go to a seaside town or something. I mean, it's just amazing. Like you, you, when you're there, you're like, I don't need any more than this. This is like, this is, this is great. Like this is enough. And, you know, there's a sense of enoughness.
Steve Hsu: I totally agree with you. I mean, in the Netherlands, you know, as long as the sun was shining, okay, the, the, the pitch darkness during winter, that's a whole different thing, but just kidding. But, but, but when the weather was good, I definitely felt like, yeah, I could live on a thousand or 2000 euros a month here.
No problem. And it's great. I love it. Right. And why can't I relax? Why, why do I feel the need to like, oh, I gotta, I gotta message this investor about, you know, something, right? So, yes, that divergence is, is, is for sure there. And it doesn't seem like it's going away.
Callum Williams: I mean, it's interesting because something that's again, comes up a lot on kind of online, I guess is this idea that Japan has kind of found the answer because, you know, the streets are really clean the food's amazing, everyone's really skinny, everyone lives for ages. and so on. And of course, GDP per capita basically hasn't grown or GDP hasn't really grown in, in, in many years.
And, you know, so when you go to Japan, you're like, well, actually maybe a bit like in Italy or something, you think, well, maybe GDP growth is not what it's cranked up to be. Of course, the difficulty is that because both in Japan and, and the EU, either the sort of, I mean, the politics, I suppose, have, have, have meant that governments are unable to run budget, budget uh surpluses or, or even balance the budget. And so you have these massive accumulated debts and massive contingent liabilities to pay pensioners for, for many, many years to come. And so you're in a problem where you really do have to increase GDP. Like if you had, if Italy genuinely decided to have a zero growth economy, where it also balanced the budget and also didn't have massive contingent liabilities to its pensioners, then it would, it genuinely really wouldn't have a problem. But the reality is that it has to cut its GDP, its debt to GDP ratio, because at any moment it could all explode.
Um, I mean, what's interesting is that there are some countries where they have actually bit the bullet and are now sorting themselves out. So Greece is a really interesting example. It's kind of gone a bit under the radar, but like, you know, Greece's debt to GDP ratio has fallen by like 50 to 60 percentage points of GDP in the past four years. Like that is insane, right? They got, they got dealt a really good hand. They had high inflation, high nominal GDP growth, low interest rates. When you combine these three together, your debt to GDP ratio is going to collapse very fast.
In addition though, they have actually decided after the whole like Syriza, Yanis Varoufakis thing to, you know, to seriously implement reforms and their economy's growing pretty well. You're also seeing this in Spain to a degree. You're seeing this in Portugal. Like, I don't think Europe is definitely screwed, because there are some glimmers of hope, but it's mostly screwed.
Steve Hsu: Okay, good. Excellent.
Okay, our next big topic, AI and jobs.
Callum Williams: Yes.
Steve Hsu: I believe you actually covered it. This is like one of your focuses, right? Uh,
Callum Williams: I've written about it. I mean, I've written about most stuff we've talked about, apart from, apart
Steve Hsu: Okay. Sorry.
Callum Williams: stuff. But, yeah, no, it is. It is. Yeah. Yeah.
Steve Hsu: So let me tell you what something that struck me recently as, you know, an entrepreneur that's actually working in this space. And so, in our case, we build these AIs that can handle things like customer service, and the voice's voice abilities or capabilities are really good.
Now the eyes we build can really. answer phones, deal with account issues, delivery issues, ordering food, whatever it is, pretty much at the level that a human can do it. And, probably cost much less, like, actually less than one tenth the hourly rate of humans to do it. But the rate of adoption is still very modest.
And when we get a look into these call centers, they have typically the back end systems are very antiquated, the humans that manage the call center are very risk averse because their job is mainly about nothing that can blow up. If something blows up, I will get fired. Some whiz bangy Silicon Valley guy comes and says, Hey, this black box is going to replace 80 percent of your head count.
And the managers, the guy who owns that operation is thinking to myself, wait, I'm going to give away 80 percent of my headcount to this guy, and I have no idea how that black box works. What's going to happen to my job? Right? Because after a while, then all the, all the future improvements or management of that function are whoever runs the box, not him.
So there's, there's a huge amount of just, I would call this like human decision maker friction, which is preventing you know, obviously things should equilibrate if I can do the same job at 1 10th, the cost for you. You would think there's going to be some equilibration, but the time scale is just shockingly long.
There's just so much resistance to it. And it's not like kind of, it's not kind of this dystopian singularity resistance where they think, oh, is going to kill us or is bad. It's, it's, it's actually just very mundane friction that prevents these organizations from shifting their workflow. And so one of the interesting things for me is that even though in this narrow space, we have delivered an AI that can do the job, the adoption is still extremely low. And it's going to take years before it really is, is, like a big chunk of the total work hours in space. So any reaction to that?
Callum Williams: Yeah. Super interesting. I think the way you outlined it was theoretically really interesting because there's a kind of principle agent problem, right? Isn't there? Because it's like a sort of public choice theory problem where the guy who has the decision over an efficiency improvement. which would be in the benefit or in the interest of the company at large.
It doesn't work in his interest or her interest, so he says no thanks. Yeah, no, I, I totally,
Steve Hsu: Just to give you an anecdote.
So we were talking to the guy who owns the entire customer support function for one of the major, you know, food delivery functions. Like, I don't want to say the name of the actual company, but you know, Uber Eats, DoorDash, that kind of thing. And he runs it, you know, a big chunk of it is run out of the Philippines because that's the lowest cost English speaking workforce that they can use.
And he just told us flat out. He said, well, my company is very technologically advanced because it's relatively, you know, wasn't a startup that long ago. And so we're kind of building out a lot of these systems ourselves. But if you just go to somebody who has my job at a more traditional company, a company that is like 30 years old instead of like 10 years old, that guy's just gonna, he's gonna listen to you, he's gonna take your meeting, he's probably got pressure from his CFO, because the CFO reads these articles in The Economist about AI and says, why aren't you delivering savings to me?
And that's the main pressure for the guy, he's got to like, act like he's exploring AI, but he doesn't want to buy the system from you, it's just a headache for him, it's a risk for him, and he's just going to stall you long enough so that he can retire and get another gig, basically. That's literally what we were told by the guy who owns the CX function for a huge organization.
So, I think, I feel like people who theorize about the deployment of AI in the economy just don't, don't have a granular picture of what, what it really looks like to the guy trying to sell the system.
Callum Williams: You know, I, that's really interesting the way you've outlined it. I mean, our, so the, the stuff that I've written that we've written about about AI has definitely been on the more bearish side of things. I mean, you know, there's this interesting kind of divergence that's taken place in, in, in the U.S. and other developed economies over recent years, which is that. You know, consumer adoption of tech has been very rapid. So, you know, you see those classic graphs about like the number, how long it took for a chat GPT to reach a hundred million users from like, it was like three days or whatever it was.
And how long it took threads to reach a hundred million users. It was like one day or whatever. So everyone's got an iPhone and all that kind of stuff. So there's this kind of sense that technology is moving faster than ever.
Um, but if you then look at what businesses are doing and, you know, businesses are the ones who drive productivity growth.
There has been this really interesting and very marked slowdown in productivity growth at the business level, really since the financial crisis, but even before where, you know, innovation has come along and just people are just not interested in them. And, you know, so, I mean, the OECD has got good data on things like how long did it take for half of American businesses that even have a website, right?
And it took ages. I mean, a website, right? It wasn't until like the 2000s, the mid 2000s or
Steve Hsu: Or fax machines is another similar story.
Callum Williams: Fact machines definitely hang around. Um. You know, I mean, it's, it's, it's much, much worse than Europe, but you know, there's many, many European countries where like, if you want to buy something off a website, like the thing is on the website, but you then have to like phone or email someone to like buy the thing, they'll send it to you and all this kind of stuff.
So, yeah, no, I, I think that's absolutely right. I mean, so the, the, the best data set, I think for this is from the Census Bureau. It's kind of gone weirdly under the radar, but they have a very similar frequency, I think it's every two weeks. Survey of like many, many businesses, thousands, possibly hundreds of thousands of businesses where one of the questions they ask is like, how many, like, are you meaningfully using AI to do something big in your organization?
Right? And it's gone from sort of like four to 5 percent in September 2023 to like 7 percent today. So basically not much of an increase, but you know, something it's, it's, you know, higher increase when you wait for employment. So the big companies that have more staff are more likely to take on AI.
So maybe you're getting to like 12 percent or 13 percent with some deployment of AI.
But you know, like I, I, even I have been surprised by how slow it's been. And it's one of the, it's a bit like the Russia Ukraine thing were like, what happened in March, 2023, when GPT 4 came out was like, people would say, adoption's gonna be slow, but just wait until 2024 and then it's really gonna take off. And then 2024 came along and then they say adoption was slow, but just wait until the, the, the fall of 2024 or early 2025 and like, you still haven't, you still haven't seen that. So, and of course you, you know, you, I mean this is even before getting to the question of like, is there an impact on jobs?
Is there an impact on productivity growth? No. There, there isn't. And I find it very odd. That there are a lot, I won't name anyone, but there are lots of people, you know, working at very prestigious universities as economists who are just very oddly unwilling to face this problem, even to discuss it. Very odd to me. I don't understand what's
Steve Hsu: Being the slowness of adoption or the slowness of, yeah, yeah.
Callum Williams: You know, there'll be some new invention and, you know, people will say this is going to change everything. And then in practice. I mean, you know, just nothing really happens, you know, eventually I'll be proving what I'm sure, but for now, um,
Steve Hsu: I think the place where I've seen some discussion of this in economic history, you know, people use this term technology diffusion and there they when they do look backs, they are aware of how long it took, like, to go from, like, oh, everybody has electric access to electrical power if they want it. But then, like, how do you reorganize your factory to really actually exploit this electricity?
And it could be decades, right? and people don't want to think AI is going to be like that. But, but my, my experience, it's not gonna be that long, but it is longer than people expect. I really like your characterization of, okay, in our daily lives, if we see something that's good, we adopt it right away.
No principal agent problem, right? Like,
Callum Williams: Right, right, exactly.
Steve Hsu: The bad email, the, the like tiresome emails that I get. I will just turn it on and it will answer the, you know, if it could actually do that, then I would, I would not have to answer these tiresome emails. I would just turn it on. But in a company setting, it's, you, you typically have more of a principal agent problem or more of like a kind of organ, bigger organizational problem in trying to deploy something.
Callum Williams: It's super interesting because you, you basically got, I think you've kind of got two variables, right? You've got, you've got a, you've got kind of an axis and a Y, X axis and a Y axis. And on one axis is like the complexity of the technology which, you know, I would, I would think that like compare, compared to something like the tractor in a farm or electricity or, like the factory system, some, like a sort of GPT model doesn't strike me as, you don't have to make like that many changes to adopt it in some way, right? To hold everything else constant, you don't have to make that many changes. I don't think you have to like to invest an enormous amount in a whole new factory or invest some enormous amount, the whole tractor.
But then the other axis is like the complexity of the firms that are adopting the technology. And I think if I'm just thinking out loud here, but you know, the average size of the U of us firms is obviously much larger now than it was in the days of electricity and days of the tractor and kind of ownership and decision making is much more diffuse than it used to be. It's much more bureaucratic.
So I wonder if that results in slower adoption of technology for a given complexity of that technology. Does that make sense?
Steve Hsu: Yeah, I, well, I think definitely the complexity of the firm because the number of people have to get their heads around it. And
Callum Williams: Yeah,
Steve Hsu: it. and actually, even though in a lot of cases where they are rolling out AI, they're like, we can't talk about this too much because the company doesn't want bad optics if it's they're clearly laying off a bunch of people.
Right? So it's sensitive in that way. But
Callum Williams: interesting.
Steve Hsu: One of the interesting things is that the call center business is very good because it's a little bit like you said, it's not that complex in the sense that like the connection that goes to the headset that the human is using to service incoming calls. I'm literally like, there's an API where I'm connecting, you know, the, the, the information that generates the sounds is some IP packets and my, my engineer can interface with their engineer and we can just like, make sure those packets go to the AI in the cloud instead of through that to the headset.
So, like, it's not that much work. They're not reengineering, like, some heavy machinery in the factory. As you were saying, it's like, well, literally, I'm plugging in the AI in place of the person. They still are worried because they know, like, what are the problems with a human? Like, oh, that lady was pretty good until she got addicted to meth, and then we fired her, and then this person comes in late, but, oh, this person's a hard worker, but doesn't adapt very well when we shift the scripts, you know, on the service, you know.
They, they kind of, it's, it's, it's a, it's a, it's a ball of management problems that they're very familiar with. And the people that are on top of those organizations, they've been doing that for 20 years, managing apes to do certain tasks. Now I bring them this totally alien thing, which they have no idea what the hell it is.
And like, even though I like, I turn it, I'm like, oh, turn it on for 1 percent of your calls and we'll send you the transcripts. And you'll see it's doing a fantastic job. It's better than your average. Humans. It's still like a huge shift in their thinking. And there's, it could be that intuitively, they kind of know there's a bunch of tail risks in the AI technology that they don't really know.
And they're totally different from the tail risks from their existing human staff. Like, they maybe are intuiting that, like, well, wait, there, maybe these guys are selling me something that looks really good, but it could go down. It could all go down one day. And then I'm just screwed. Right? And how do, how does someone who's not a technologist evaluate, like, those sort of tail risk scenarios, they have no idea, no idea how common or uncommon like, oh, what it could just get hacked. Right? And all of a sudden, like we have to turn off our entire customer service. Right? So there's just all these uncertainties that I think keep them from embracing.
The people that would embrace me would be like I'm a founder. I'm in a growth stage company, and we're selling so much of our product that we need a CX function, but I'm building it de novo, and I'm going to build it using the best tech. And I'm just going to, I'm just going to roll out AIs.
You know, it's mostly AIs and a few humans to handle the really complex cases.
And they're comfortable making that adaptation. But for existing companies, it's just, it's just a battle. It's a battle to get them to.
Callum Williams: Very interesting. And I guess that means that, again, thinking out loud here, but you, so what you're, what you, if you're, if you're bullish on AI, then what you're reliant on to a degree is like the flow of new companies coming onto the market. And so, but I guess the issue there is that the, the, the rate of entrepreneurship, despite rising a bit during the pandemic is still way down on its like average in the sixties and seventies and stuff.
So, that flow is lower than it would otherwise be, which means that that kind of creative destruction, that churn happens more slowly than you'd want it to.
Steve Hsu: Yeah, in fast food. So we actually operate in space. So Taco Bell has, is rolling out nationwide. Like, it's not hard. It wasn't hard for me in Michigan to find a Taco Bell outlet that had the AI. So you're in your drive through and you talk to the AI and you can order your food. It's pretty good.
Actually, the woman who used to, or the guy who used to take the order inside the kitchen with a headset on is still listening and can get pulled in. But mostly you can just give the whole order to the AI And the order goes into the kitchen and then they give it to you. So, Taco Bell is kind of the leader and all these other firms have.
They're like talking to their franchisees. It's a very complicated conversation because a franchisee might not want it. Or they're worried about it and to what degree can the parent company dictate it to the franchisee? Sometimes the franchisee is more tech forward and wants to roll a system out and the parent corporation won't let them.
So it's really what's the right way to say it? Inhomogeneous situation right now in food ordering. But the AIs are fully capable of doing food ordering better than humans at this stage. And so, but it's, it's going to take at least a few years before, you know, you see it everywhere. So,
Callum Williams: Absolutely. I mean, and then what about the question of like, does the question of what the customers expect or what the people think the customers will want, does that, does that come
Steve Hsu: Yeah, absolutely. It's uncertain, right? So you don't know, you don't know whether people are going to get really pissed off at you. Like, I hate AI. Like, how are you going to know? Like, if you put yourself in their shoes, the decision makers at Taco Bell, like there's the corporate and then there's also all the franchisees who all have different opinions of like, how is this going to play out?
Like, do, is this going to suddenly tank my sales? You know, or is it going to grow my sales? So there's a lot of decision making uncertainty with this stuff.
Callum Williams: I think also this gets to where, and this is quite speculative as well, but I think there's been this kind of shift in the West. And you can see this in some parts of the data where people are much more interested in avoiding big losses than making big gains.
Steve Hsu: Loss aversion.
Callum Williams: Massive loss aversion and it's got, it's got, well, worse or better, it's got more queues, I think, over recent, over recent decades, possibly because we've got richer, I don't know, possibly because, you know, social media means that any failure that you might have is like broadcast to the world or whatever, who knows,
Steve Hsu: right? Right. Yeah. So anyway, I think if you then switch to the magnificent 7.
Callum Williams: Yeah.
Steve Hsu: And you say, wow, what are these valuations based on?So we, you know, we have this huge I would say bubble in the mag 7. And if, if, Okay. You don't think it's a bubble. How's it going to pay off for you?
Is it because eventually there's so much utility generated by AI that the earnings of the Mag 7, you know, they actually Microsoft starts making shit tons of money because Taco Bell is using their AI on the back end. And all these agents are replacing people and stuff, but the time scale for that is not a few years, but 5 or 10 years and our investors are going to wait that out.
They're going to keep those valuations where they are while the, the, the actual revenues are very slow to materialize much slower than expected to materialize. Or is it none of what I'm saying is important because we're going to get to super intelligence and all of a sudden, like, you know, like the thing invents, like cure for cancer overnight for us. And that's where all rewards come from. Right?
Callum Williams: I mean, there's loss aversion there, right? Because if you're the company that doesn't massively invest in data centers and then your competitors manage to find the cure for cancer or whatever. Then you look like an idiot. So there's that. I mean, I think there's, then there's the kind of, there's the kind of Goldman Sachs argument from their head of equity research or something, which is like, you know, they're spending roughly a trillion dollars over the next four to five years on AI.
Like what's the trillion dollar problem they're solving. unclear, I think, as you say, because it's the point of the problem, but then you go back to the companies and, you know, the people who are bullish on the companies and they'll say, Yes, but if you look at CapEx relative to cash flows, you know, you know, information technology, that ratio went completely nuts in like the late 90s in the tech bubble, where they were investing hugely in CapEx, which has had actually no cash flows to back it up whatsoever.
The ratio now is kind of in line with the long run average because, you know, they've just basically since the pandemic, like they've just become even more money printing machines than they were before. So, in a sense, you could say that it might not be sustainable, but it might not be a big problem, I guess, would be the argument.
But I don't know. I mean, as you say, the market is still expecting massive, massive revenue growth at these companies for years and years and years. I mean, I think the, I think the, you know, these companies had a once in a generation shock, positive shock to revenue and earnings, which is basically the shift to working from home.
Because of the pandemic, like fingers crossed, you're not going to get that again. Like, you know, you're going from 5 percent of people working at home to 20%. That's a huge earnings positive shock to, to, to, to mag seven. Will that happen again? Probably not. So yeah, I mean, I just go back and forth on this question for sure.
Steve Hsu: Yeah. I don't know the answer. And yeah, you know, I'm right in the trenches because I'm one of the guys who's trying to generate revenue on the back end.
Callum Williams: Right.
Steve Hsu: At least before DeepSeek came along, that back end revenue would go back to OpenAI or Microsoft or whatever. Now, it might not.
Okay. But in fact, in fact we've checked, we can swap in these open source, these DeepSeek models, perfectly fine for what we were doing before, OpenAI or something like that. So that
Callum Williams: Mean, the Microsoft argument is, that their revenues, what you hear from the company is that, you know, revenues from Microsoft's AI business have increased much faster and to a much higher level, much more quickly than their revenues from Azure did. And yet they're spending about the same or only a little bit more on AI capex than they were on cloud capex.
And so there's no problem. Now, I mean, that might be true. Of course, it then raises the question. I mean, I would wonder if what counts as AI spending is much harder to measure, I think, than maybe what counts as your spending. I don't know if that's right, but you can imagine that like some lackey who's like an upper middle manager would be quite incentivized to count things as AI spending that might not be AI spending to make the numbers add up.
But I don't know.
Steve Hsu: Yeah, I mean, I think like the thing that we do where we're literally replacing human labor with a labor and we can really see like what the what the return on investment is these more like fuzzy things were like, oh, the your meeting got transcribed and you have a nice summary of the meeting. And it's like, wow, that is magic.
Like we could not have imagined years ago that that was possible, but like, what is the actual productivity gain from that? Like, did I actually get any productivity? Get, Oh, I have a nice summary of the meeting and I could look at the transcript if I want to. But I think, but that makes Microsoft some money because somebody.
Somebody is paying for that service. Right. Although then
Callum Williams: Sure, sure, sure.
Steve Hsu: But, but yeah, so the way the whole thing is going to settle out is still somewhat mysterious to me.
Callum Williams: Yes. I mean, there is that sort of like joke about AI that, you know, person one gets notes and asks for a ChatGPT presentation and person two takes a PowerPoint presentation and asks AI to turn it into notes.
Steve Hsu: Yeah. So
Callum Williams: there is this interesting,
Steve Hsu: Yeah, I think there's a lot of non productive churn that's going on right now, actually.
Callum Williams: There is this kind of theme that you get, that is explored again by some historians and I think may also apply to AI, which is this idea that it's, it's sort of a bad thing for productivity when the costs of producing content go down a lot.
So, you know that I mean, I think these are very simplistic correlations, but like the relationship between the length of the U.S. tax code and the ease of writing stuff out on a computer and printing it out, like.
Steve Hsu: Yes.
Callum Williams: There's something there. And so you can imagine this world where a huge amount of AI slop is being sent around both internally within companies and externally.
Steve Hsu: I will tell you a story which you might enjoy, which was a shocking experience to me. Okay, so another startup that I'm involved in, which is in genomics, and it uses AI, but a different kind of AI than gen AI, but we were looking at a proposal from the Saudi government to build a lab there. Okay, and so the CEO asked me not to. I'm not in the day to day operations of the company, but I'm on the board.
So I'm one of the founders. And the CEO asked this very junior person on her business team to, like, look into some of the details about like, oh, you know, like, how hard is it to hire people in Saudi? Can you fire the, you know, just basically, what's it like to operate there? Right? And so I go into this meeting she's like, the connection to the Saudis was like through one of my connections. So, I was kind of peripherally involved in this. So, I got pulled into this meeting, right? So I got pulled into this meeting and this young lady is giving a presentation to us about what she learned about researching just what it's like to operate a business in Saudi.
KSA and I're listening and I'm like, you know, she's really not saying very much is, you know, does she really know what she's talking about? Like, and then I realized the whole presentation was AI slop. The way this woman had done her research was she probably just asked chat GPT the questions that the CEO wanted answers to.
And, you know, those were important questions. You should go online and find out what, you know, what's minimum salary, what's employment law, what, what do like leases look like, you know, what are the regulations on laboratory, sir, you know, like, there were real questions to be answered, but she got superficial shitty AI slop answers and I was about to like throw my, you know, my, my glass through the screen of my TV. Cause I wasted like half an hour listening to this. I was like, cause it took me a while to figure out that I was listening to AI slop, and then I realized like, I'm not getting any deep information from this. It's just like slop.
And then I realized she just got it from GPT. So
Callum Williams: Yeah. Happens all the time.
Steve Hsu: Negative productivity.
Callum Williams: Totally. Yeah, absolutely. Yeah, no, I think that's a real risk. I mean, I think the other, the other sort of relate more subtle risk, I think, which is, I'm very convinced by is this kind of, it's a sort of adaption of a Paul Graham argument, which is like, when you get somebody else to do your writing and researching for you, the process of thinking stops.
Steve Hsu: Yes.
Callum Williams: I, and I think that, I really think that's quite a profound insight. Because there are, I mean, there are, I mean, I know this myself from being a journalist. Like, there are some things where the only way to really learn about something is to try and write it out, you know, in a clear way. And you really discover very quickly what you don't know.
And if you've got someone to do it for you.
Steve Hsu: Yeah, if you formulate your own arguments and you're critically trying to lay it out, you know, readers are going to hold you to account. You're like, that's a different kind of processing than just like, cut and paste from what they said, right? Which is, I mean, there was a nice presentation.
There was a PowerPoint there, like, you know,
Callum Williams: Yes. Yes. Yes. Yes.
Steve Hsu: Wait a minute, but this, this is not answering our questions like we need to know this, the answer to this, this and this just before we decide to green light this right? And like, you didn't answer my questions, right? So
Callum Williams: Yes. Yes. I think it also very much raises the sort of the alpha to you know involved in, in finding tech, you know, texts that are maybe not out of print or not readily available online, you know, things that are, you have to go to like a library for, or you have to get from a book or something, you know, there's really interesting ideas or, you know, old documentaries on the BBC that like no one's seen for years and years.
Like the kind of insights you can get from that, I think, the value of kind of taste, I suppose, will, I think, go up quite a lot when there's so much sort of consensus information that's so easy to find.
Steve Hsu: Yeah. I think the danger is a lot of people don't like you and I have some idea of what's going on when they train these AIs, but a lot of people have no idea. So they don't realize it's giving them consensus slop back. So imagine you're not, you're not super high on the cognitive scale. You're just kind of an average person.
Okay. And you're being told to research something and you don't realize what you're getting back is some average slop, right? Consensus slop, not to not to. Yeah. Anyway, I just think a lot of junior people at sort of ordinary firms could get caught up in that without knowing it.
Callum Williams: I, I think they already are. I mean, I hear lots of anecdotes about precisely this, where like some senior person is like, what the hell have you just produced? They're like you, like you with this, with this person for the sound effect.
Steve Hsu: This is a crazy experience. It was a shock to me.
All right, so we're running low on time. Last question. We've talked before about the concept of stagnation and productivity growth, technological innovation. Our friend Tyler Cowen and Patrick Collison are active in researching this area. It's called Progress Studies, and it studies, you know, how do economies and societies really progress? What are the main factors that control the rate? How does technology diffuse once it's invented into the actual practical applications?
What's something interesting? I think that you and I talked about the last time we saw each other. I think Tyler has thrown in the towel because he's now saying like, well, you know what? I was wrong because all the time I was complaining about the slow rate of progress and stagnation. While I was complaining, they built this AI and I love this AI. And now, like, I can see like, you know. There was progress. I just wasn't aware of it. And now I think progress is gonna be super fast because of AI. What do you think of Tyler's point of view?
Callum Williams: Well, my personal view would be that, that, that that Tyler Cowen is way too bullish on, on AI, but that's, you know, that's sort of one thing. But I think that, you know, the point that, you know, he had made with Collison, but also Peter Thiel has argued for a long time, is that like, once you take a step back from what is, what's, what's Thiel's formulations, like, like bits and atoms or something, once you step away from tech, broadly defined, which is bits, I think, and look at atoms, which is like real stuff in the economy, then I think it is pretty, like, obvious that there has been some slowdown, quite marked slowdown, actually, in in progress or in or in kind of the quality of life broadly defined.
I mean, if you think about things like, you know, travel speeds around most countries are falling, having been increasing for a long time, certainly in the UK, you know, during the 19th century, 20th century, you know, sort of average miles per hour of a trip was going up and up and up. And that has plateaued and now it's falling.
Um, you know, you look at things like, I mean, there's, there's crazy things like how quickly it takes to put up a building, you know, that takes ages now, you know, San Francisco put up the Golden Gate Bridge in three years. The Empire State Building went up in a year. Like that just would never happen now.
You know, things are things like, I mean, I guess this is a bit closer to bits, but like, you know, the quality of a telephone call, it sounds incredibly analog, but like telephone calls still have terrible audio quality. And they just haven't really improved in many years. Like I was, it was interesting because, you know, when Peter Thiel was on the Joe Rogan podcast a few months ago, you know, I think Rogan put to him what I think is the typical objection to his ideas, which is like, aren't we surrounded by amazing technology all the time?
And I was quite surprised by how like Peter Thiel was oddly quite sort of like, he wasn't able to respond very quickly to that objection. I mean, the one he ended up doing, I think, was that we haven't, like, landed on the moon since 1971. Which, again, is a great example, like, why haven't we landed on the moon?
It's nuts, right? so I think there's loads of examples, and I don't think that, like, AI, which obviously isn't, I'm not trying to say AI isn't amazing, because obviously it's completely amazing, I use it all the time, but, like, the notion that AI by itself undoes all of that stagnation strikes me as just completely wrong, you know?
Steve Hsu: Yeah. My view on this is that the basic observations about, like, average velocities of travel have not changed and. Time to put up a building, you know, has increased a lot. I think those observations are all true. some, maybe for the buildings, it is bureaucracy and red tape. It's maybe not a technology thing, but for some stuff like, oh, could we fly around at 5000 miles per hour instead of 500 miles per hour?
There are real kinds of physics limitations. So, there's a low hanging fruit story. Where we, when it comes to the atoms, we picked a lot of the low hanging fruit, and now the rate of progress is just slower. One thing I would say in that context, though, is that if instead of looking at macroscopic things, like how fast we fly around or our buildings, if you go the other direction, look at nano technology, like what we, what we can engineer on a microchip, that has improved by factors of a million.
So, so we, we did get factors of millions relative to, say, the computer that they used on the Apollo. we did get that, and that is why we have AI now. So, so we did, our mastery of the physical world really shifted toward the quantum world, toward the nano world, but we did master it. And it's kind of invisible to most people, because most people don't really understand what a semiconductor, you know, physics is, or something like that.
But, but that's, that's integral to, like, wow, this AI thing is passing the Turing test, right? And that wouldn't happen if we just didn't have, like, factors of a million in computer advance.
Callum Williams: Totally agree. But I think that's, that's the, that's the atoms v bits. Point of,
Steve Hsu: Well, but I think where I would, I would kind of like berate Peter a little bit on this is like your bits, we can do the bits because literally on the atoms
Callum Williams: Okay, fine. Yes, yes.
Steve Hsu: mastered the atoms. We didn't master how to make planes that can fly 5, 000 miles an hour, but. We did master the atoms, and that's why we're able to move the bits around for you.
Callum Williams: Yes.
Steve Hsu: Yeah.
Callum Williams: I guess I, my only counter to that, I suppose, would be to say that I think the reg, the regulatory system in which technology operates is as important as the tech in a sort of allhouse equal sense. So, I mean, you know, I mean, to, to use the example of the airplane speed. I mean, I, my understanding is that like average air or maximum airplane speed in the U.S. has been fixed for many decades, and we would very easily be able to go 20 to 30 percent faster, as indeed we did with Concord, but the regulations were such that we're not, we're not allowed to do that. So I, you know, the government plays a big part too.
Steve Hsu: There's a kind of hard thing, you know, because if you're supersonic, there's all kinds of bad things, like sonic booms and the, and, and the fuel usage grows like the velocity squared. So there are reasons why we kept out at a certain level, and we just don't want to be flying around at 5, 000 miles an hour, because then you get a plasma layer in front of your front of your plane.
And it's a real problem for the pilot. So it's just, there's, there's actual hidden things here, which really come from physics that people just aren't aware of for macro.
Callum Williams: Totally agree. I just think we're way, way, way short of that, like, frontier, you know, like the, the advent of the TSA in 2002 or whatever it was like. That has surely, surely, surely caused net social losses in the billions to hundreds of
Steve Hsu: Yeah, I would differentiate between, like, some bureaucratic or social system. Things are holding us back and like physics sometimes are holding us back.
And I think one of the big stories, which I think people are slowly coming around to because of the big data centers. The power story is we basically left this whole nuclear thing on the table. So if you, if you talk to physicists in the fifties and you read science fiction from the fifties, they all assumed we were going to be, have basically unlimited energy and actually be using nuclear powered spaceships and stuff very soon. Cause we, we could have done it, but people were just so irrationally scared of nuclear energy for the last 50, 60 years that we didn't do any of it. And now we're kind of reconsidering our irrational fear of nuclear energy. And we can now look back on well, how many people in France got killed by nuclear energy? How many people in Japan got killed by nuclear energy? Almost none. Right? So now we can, we both have a very long look back period for safety and it's becoming more of an acute issue to generate clean energy.
So, I think the whole, but historians will look back and go, wow, these, these apes got to the nuclear frontier, and then they just didn't improve it for 50 or 60 years. And then they started, right? So,
Callum Williams: Very interesting.
Steve Hsu: Yeah. So, Callum, it's been great chatting with you. You know, I, you know, maybe we can do a regular series where we just come in and you, you enlighten us on specifically economic topics. That would be
Callum Williams: I'd love that, because there's a lot of, you know, it's one of those things that gets really bad coverage, and especially online, it can be really bad, so that sounds great.
Steve Hsu: Yeah. Great. All right. Well, thanks a lot.
Callum Williams: Alright, thanks Steve.
