What this covers

This is a conversation between host Jack Farley and Peter Howitt, recorded in February 2026, about whether AI is economically different from earlier technological shocks. Howitt starts with the model of creative destruction he built with Philippe Aghion. That model folded game-theoretic patent-race ideas into a general equilibrium model, in which firms innovate to leapfrog rivals and capture monopoly rents. He contrasts it with the older Solow growth model, which treated technological progress as exogenous and assumed everyone benefited equally. The conversation then moves to AI, which Howitt classes as a general-purpose technology in the lineage of steam, electrification and the computer.

The historical material is the most varied part. Howitt cites the power loom and the handloom weavers it impoverished, and he gives a contested reading of the Luddites as skilled craftspeople shut out of the gains, not technophobes. He also argues that England's policy stance toward guilds helps explain why the Industrial Revolution began there. Electrification serves as the model of a long lag: its payoff waited on the assembly line, and he links this to Robert Solow's remark about computers and the productivity statistics. He also draws on Paul David's work on diffusion. Other examples include the shrinking farm share of employment, the new service sectors that now employ millions, and Kodak's collapse.

The discussion gets more contentious on present-day risk. Howitt says rising industrial concentration has been slowing innovation, and he lists the ways large firms suppress it: preemptive mergers, patent thickets, political influence and regulatory capture. He also says that if no "next thing" emerges to absorb displaced workers, concentrated AI wealth could lead to serious social conflict. The conversation also covers the AI capital-expenditure boom and the historical pattern of investment manias ending in crisis, and it rules out sustained 15% real growth. It closes with a look at Howitt's other research.

Sharpest takeaway

Howitt argues that AI, like past general-purpose technologies, will create net long-term prosperity through creative destruction despite significant near-term labor and firm displacement, but only if competitive markets and innovation-supporting institutions remain robust and wealth benefits are distributed broadly enough to prevent social conflict.

  • Historical pattern shows technological unemployment predictions have always been wrong; past general-purpose technologies (steam engine, electricity, computers) eventually created more jobs and higher living standards than they destroyed
  • AI's destructive phase is currently visible while creative phase (new occupations, new industries) is not yet clear, but downstream innovations will eventually reveal benefits if competitive markets allow rapid entry and experimentation
  • Rising industrial concentration and monopoly power threaten the innovation-competition feedback loop that drives growth; antitrust policy needs reorientation from consumer-harm focus to innovation-suppression focus

The claims · ranked62 claims · weighted by value

This asset isn't compiled yet

You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.

0.84

There is a conflict at the heart of creative destruction between innovators and incumbent producers: innovators disrupt the status quo and render old products obsolete, while creating losses for people whose human capital and skills are tied to old technologies.

causalhigh valueestablishednovelty 2/4durability 4/4· Peter Howitt

“Creative destruction implies that there's a conflict, a conflict between the new and the old. And the people that are coming in with new technologies are disrupting the order. They're challenging the status quo. They're threatening the well-being of incumbent producers that are producing the goods that they're trying to make obsolete. And they're also creating a lot of losses for people whose human capital or skills are tied up with the old technologies.”

0.80

These new technologies will displace workers in many places and some sectors of the economy will be badly hit, but it takes further downstream innovations before the full potential is realized and new occupations open up that spread benefits.

causalhigh valueestablishednovelty 2/4durability 4/4· Peter Howitt

“It's obvious that uh these new technologies are going to displace workers in a lot of places. They're they're to displace all manufacturing interests in in in a number of cases. And at first, it's that destructive potential that's clearest. Uh you can you can see right away that there are certain occupations, there are certain sectors of the economy that are really going to be badly hit by this. It's not so clear where the benefits are going to be. But what happens is that it takes further innovations, downstream innovations, before the full potential of the technology is really realized and before new occupations open up, new industries open up, new ways of earning a living open up”

0.80

Great economists including Ricardo (1821), Keynes, and Samuelson (2004) have predicted that technological progress would create long-term mass unemployment, but all these predictions have been proven wrong by real-world developments.

factualhigh valueestablishednovelty 2/4durability 4/4· Peter Howitt

“We do have a whole history of some of the best economists in our in our history predicting that new technologies are going to create long-term mass unemployment. You know, going back to Ricardo with his famous chapter on machinery in 1821 in the third edition of his principles of political economy. And then through John Maynard Keynes, uh who argued that technological unemployment was going to be inevitable. Uh even Paul Samuelson in 2004 had a had a a wonderful little article entitled Ricardo was right. And uh produced a beautiful little general equilibrium model showing how technological progress was likely to drive the equilibrium real wage rate to zero in the long run. And uh But so far, all these great economists have been uh have been proven wrong by by real-world developments.”

0.80

The most innovative industries are those where firms are roughly neck and neck technologically, without dominant leaders, because this situation approximates perfect competition with no monopoly profits, creating the strongest incentive to innovate.

causalhigh valueestablishednovelty 2/4durability 4/4· Peter Howitt

“The most innovative industries tend to be those in which firms are uh more or less neck and neck technologically. There isn't one dominant leader and a bunch of firms that are way behind. Um because in that situation, the firms with existing technology, when they're sort of neck and neck technologically, that's the closest you can get to perfect competition where there are no monopoly profits. And that's where the incentive to innovate is the strongest.”

0.80

Individual firms choosing to innovate less when expecting future competition might reduce their immediate payoff, but the stronger competitive pressure today creates a net incentive to innovate now, and aggregate innovation is higher when many firms compete due to the multiplier effect.

causalhigh valueestablishednovelty 2/4durability 4/4· Peter Howitt

“If you expect a lot to be of innovation to follow in the event that you are successful in innovating. And then that simply means that um, uh, you're going to be facing tougher competition down the road. The payoff to you won't be as great. But what if you expect What if you have a lot of competition right now? Not just in innovating for new products, but the competition for producing and selling the goods that you're able to produce right now. Uh, you have That's where you have the strongest incentive to innovate so as you can end up with some monopoly rents... But you've got that extra innovation for society because you've got these others doing it, too.”

0.74

The power loom displaced handloom weavers who had spent decades perfecting their skills and became impoverished, but in the long run, nobody's life would be materially better off if the power loom had never been invented because it was one step in an ongoing ladder of technological progress.

causalhigh valueestablishednovelty 1/4durability 4/4· Peter Howitt

“You take a look at the start of the Industrial Revolution at the devastation that was that was produced by the invention of the power loom. There were these handloom weavers who had spent decades perfecting their skills and were highly paid, highly trained crafts people who were all all of a sudden um made you know, made into poppers because they there was no way that they could compete against the power loom. Um there was a you know, a long period before before people you know, like that recovered. Uh some of them never did. Uh but nowadays who can you think of whose life would be materially better off if the power loom had never been invented?”

0.74

Electrification technology became available in the last third of the 19th century, but its full potential wasn't realized until the invention of the assembly line method of production, which distributed power sources horizontally instead of requiring factories to be built along vertical lines around a central power source.

normativehigh valueestablishednovelty 1/4durability 4/4· Peter Howitt

“Obviously, electric motors have the potential to make manufacturing vastly more productive...That possibility of using electric motors was available in the you know, the last third of the 19th century. Uh uh That possibility of using electric motors was was available in the you know, the last third of the 19th century. Uh and a lot of factories did that...But it wasn't until the invention of the assembly line method of production that the that the real potential was realized. Uh People had to realize that that in the past a factory had to be built along vertical lines...But the full potential of the of electricity had to be realized by distributing the power sources horizontally”

0.74

Industrial concentration has been rising in many US sectors over the last 20 years, and there is substantial evidence this rising concentration has been slowing down the process of innovation.

causalhigh valueestablishednovelty 1/4durability 4/4· Peter Howitt

“One of the worries about the last 20 years uh in the US and and in many other countries is that industrial concentration seems to be growing in many sectors of the economy. And there's a lot of evidence that this is this has been slowing down the process of innovation.”

0.73

If there is no 'next thing' (no new sectors and occupations emerging to absorb displaced workers), there will be significant social conflict because wealth from AI will concentrate and not be shared.

forecasthigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“Well, if there's not a next thing, then we're going to have trouble uh uh we're going to have a lot of social conflict. Because it's clear that some people are getting fabulously rich from artificial intelligence. If that wealth isn't spread around, we're going to have a lot of conflict, just like they had when the steam engine first came around.”

0.71

The Luddites were not irrational people with superstitious fear of technology but skilled craftspeople who were not sharing in the benefits of new technologies, and the fact that guild-backed machine-smashing happened in Germany, Holland, and France—not just England—suggests social opposition was the real pattern.

factualhigh valuecontestednovelty 2/4durability 3/4· Peter Howitt

“The Luddites that were smashing machines We're not just irrational people with a with a superstitious fear of technology. They were They were skilled crafts people who were not sharing in the benefits of these new technologies... You had the same phenomenon. You had people smashing machines in Germany, had them doing it in in Holland, you had them doing it in France. Um and they were supported by their guilds.”

0.70

Previous technological transitions occurred with appropriate institutional supports, showing that policy choices matter for whether technology adoption succeeds and benefits are broadly shared.

factualhigh valuecontestednovelty 1/4durability 4/4· Peter Howitt

“The Industrial Revolution happened in England partly because the government was hard on guilds and favorable to manufacturing. In other countries where government supported guild opposition to machines, the transition was slower and socially conflictual.”

0.69

Mega-corporations can suppress innovation through preemptive mergers, creating patent thickets, exercising political influence, and achieving regulatory capture—all methods observed in recent tech consolidation.

factualhigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“They're in a high-tech sector that's being increasingly dominated by a few very, very powerful firms who can suppress innovation by others through a number of methods, through preemptive mergers, through creating patent thickets, by having political influence, regulatory capture, and so on.”

0.69

General-purpose technologies experience long lags before aggregate productivity gains are visible in statistics, as Robert Solow noted about computers in the late 1980s ('you can see the computer everywhere except in the productivity statistics'), and the same lag occurred with electrification.

factualhigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“I'm sure you remember Robert Solow was famous in the 19 late 1980s for saying you can see the computer everywhere except in the productivity statistics. And the same was true with electrification. And I talked about how the long lag between the introduction of the technology and the positive benefits.”

0.69

Generative AI is a general-purpose technology with ramifications throughout the economy, not just in a few sectors, similar to the steam engine, electrification, and the computer/internet.

factualhigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“Generative AI is obviously a very powerful technology that's going to have ramifications throughout the economy, not just in a few sectors. And it's like a handful of other really big what we would call general purpose technologies, you know, starting from the steam engine, going through electrification of manufacturing, uh through the the computer and the internet in the mid to late 20th century, and now we've got artificial intelligence.”

0.69

In 1992 when the Aghion-Howitt papers were published, the talk of monopolies focused on government-granted monopolies (water, electricity) with regulated rate limitations; now there appears to be more monopoly power and rent extraction throughout the economy, and this has become a bigger concern.

factualhigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“I I I think there is, actually. And and that's exactly what Philippe and I were writing about in this this this paper that we wrote a few years ago. That there's there's a And this wasn't our finding. I mean, David Autor and many other people have been pointing out that there's been this rise of the superstar firm.”

0.68

Howitt and Aghion made the creative destruction model operational by incorporating game-theoretic patent-race models from industrial organization into a general equilibrium model, showing how firms innovate to leapfrog competitors' technologies and capture monopoly rents.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“What we did was, uh, to find some techniques that have been used in the theory of industrial organization to study patent races in individual markets, and put that into a general equilibrium model of the whole economy. And, uh, we we created a model where growth comes about from research and development activities by business firms that are aimed at trying to produce innovations that will uh, that will leapfrog the technologies of existing producers, and allow them to capture some monopoly rents because they found a way to produce a product that's better than anybody else can produce”

0.68

150 years ago, 50% of the population in North America was engaged in agriculture and had to work on farms to feed everyone; now it's 1 or 2% and we produce way more than we need, demonstrating successful technological transition at scale.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“You take a look, 150 years ago, 50% approximately of the population in North America was engaged in our agriculture. They had to be had to be working on the farm just to feed us all. Now it's 1 or 2%. And we produce way more than we need.”

0.68

New service sectors like financial services, information services, and communication services employ millions of people today, but 150 years ago only a handful worked in these fields, showing how new occupations and sectors continue to emerge.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“We have a lot of people in financial services right now. If you If you went back 150 years ago, uh how many people worked in financial services? Well, a few people worked in banks and uh it was just a handful, uh relatively speaking. Uh you know, uh people work in information services. Uh they they just weren't a lot then. Communication services.”

0.68

Government policy in England was particularly hard on guilds and favorable to manufacturing interests, while other countries supported guild opposition to machines, which is a significant explanation for why the Industrial Revolution started in England rather than elsewhere.

causalhigh valuecontestednovelty 2/4durability 3/4· Peter Howitt

“The government in England was particularly hard on the guilds and in favor of the manufacturing interests. And uh the guild, you know, people just had to adapt. Uh that was not the case in other countries... historians will tell you that's one of I mean, nobody knows for sure and there are many other contenders as an explanation for why the Industrial Revolution started in England, but I think that was a big one.”

0.68

Kodak Corporation was at one point the third largest corporation in the United States by revenue and employed 150,000 people worldwide, but was driven bankrupt by digital photography, showing that technological disruption can destroy even dominant firms.

factualhigh valueestablishednovelty 0/4durability 4/4· Unidentified Speaker — Discussing A.I. Displacement with Nobel Laureate in Economi… [aKSmMLOj8SM]

“The Kodak Corporation was the the at one point the third largest corporation in the United States by revenue. It played 150,000 people worldwide. They were driven bankrupt by digital photography.”

0.68

It is impossible to sustain 15% real GDP growth, and countries like China with sustained 20% growth rates had unreliable data; no example exists of 15% real growth sustained for more than a year or two.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“Can we get 15% growth? Okay. Well, uh who knows? That that might be possible, but um I I can't think of an example where that's been sustained for for more than a year or two. China looked like it was having growth rates like that for a long time 20 years ago or so, but uh but the Chinese data are not entirely reliable, I'm afraid.”

0.68

These massive technology investment booms typically lead to a climax and crisis with failures and winners emerging; historically this has been the pattern, and some investors are convinced a stock market crash will happen but can't be sure they'll exit in time.

forecasthigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“Typically these huge uh, investments in new technologies uh, will lead to some sort of a climax, some sort of a crisis will take place and there going to be some failures and other winners will emerge. That seems to have been the pattern in the past and I know a lot of people, I have a lot of friends that are convinced that these there's going to be a stock market crash that's that that you know, not all of these valuations are realistic. But they're fully invested in the stock market.”

0.68

Henry Ford's assembly line in the 1920s-30s allowed him to offer unskilled workers an unprecedented $5 per day wage, showing how downstream innovations from a general-purpose technology create widespread benefits.

normativehigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“Once Henry Ford you know discovered how to start producing automobiles along this an assembly line back in the 1920s and 30s he was actually able to offer fairly unskilled workers the unheard of wage of $5 a day. The the you know it was a tremendous benefit to all sorts of workers.”

0.68

Current antitrust laws and enforcement worldwide have not fully adapted to addressing innovation suppression; they focus on consumer price harm rather than on preventing market leaders from blocking competitor innovation.

factualhigh valuecontestednovelty 2/4durability 3/4· Peter Howitt

“Our laws and our the approach that antitrust authorities have followed followed uh uh around the world really have have not really fully adapted to that.”

0.68

The old Solow model of growth correctly assumed that long-run growth was determined entirely by the rate of technological progress, but incorrectly treated that technological progress as exogenous (given by external forces) rather than endogenous to the economy, and incorrectly assumed everybody benefited uniformly from new technologies.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“At the time that we were writing, the state of the art in growth theory was the famous Solow model created by Robert Solow back in the 1950s. A beautiful little model, but it is and and and it it it assumed correctly I think that the long-run growth rate was entirely determined by the rate of technological progress, but it took that rate of technological progress as just something that was given by external forces by some scientific process that had nothing to do with economics. And it also assumed that everybody benefited uniformly from the from the new technologies when they arrived.”

0.68

The economic historian Paul David pointed out similarities between personal computer and electric motor diffusion, showing that both involved extensive fumbling, mistakes, and learning that looks like pure input waste in productivity statistics but actually represents valuable investment in knowledge.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“The economic historian Paul David once wrote a really interesting paper uh, about 1990, I think, called The Computer and the Dynamo, where he compared the uh, progress of aggregate productivity following the introduction of the personal computer and its diffusion throughout the economy with the introduction and diffusion of uh, the electric motor uh, in in manufacturing. And he pointed out a lot of similarities. There was a lot of fumbling around with the new technology. A lot of mistakes were made.”

0.68

The Great Depression coincided with massive transitions from agriculture to manufacturing, suggesting technological displacement during that period created more unemployment than visible because displaced agricultural workers still registered as employed despite moving between farm and non-farm work.

factualhigh valuecontestednovelty 2/4durability 3/4· Peter Howitt

“It might have been uh uh uh just the great progress in agricultural productivity throughout the um you know, the first part of the 20th century. Uh Um you know, we still don't really fully understand where the Great Depression came from, but it certainly came in the middle of this massive transition from agriculture to to manufacturing. But it would have shown up in even more unemployment except that a lot of people who uh became unemployed just went back to work on the farms. So, they were still registered as as unemployed. There were still lots of family farms that they would go back to and so on.”

0.68

Philippe Aghion found evidence from French data that firms introducing personal computers more intensively were both more productive AND had more employment, while firms laying off workers were the laggards not adopting new technology—suggesting technological adoption creates jobs by capturing market share.

normativehigh valuecontestednovelty 2/4durability 3/4· Peter Howitt

“My my colleague Philippe has is more optimistic on that. He's been writing some papers with co-authors looking at French data and and and arguing he's he's got evidence that uh firms uh during the um you know, during the introduction of personal computers found that he found that firms that tended to introduce uh uh new information technologies more intensively uh did tend to be more productive, but they also tended to have more employment. And that where that in an industry, the firms that were laying workers off were the firms that weren't introducing information technologies rapidly.”

0.66

Population growth alone does not provide an advantage because economic growth is measured in income per person, and higher fertility rates create a demographic penalty where young dependents must be educated and capital must be provided for new generations before benefits arrive.

causalhigh valueestablishednovelty 1/4durability 4/4· Peter Howitt

“Countries with rapid population growth don't really have much of an advantage. It's true that they'll get more labor input. But, when we think of economic growth, we're really thinking of uh the growth of income per person. And uh you know, having more people uh doesn't immediately raise the standard of living of the average person. Uh you know, if you get them through uh through a more rapid uh through higher fertility rates, for example, then you're going to go through a large period where you've got this demographic penalty where you've got more people, but more of them are at an early age.”

0.66

Howitt cautioned that while past pessimism about technological unemployment has been proven wrong, there is no theorem proving this pattern must continue—history reveals patterns but doesn't guarantee outcomes.

factualhigh valueestablishednovelty 1/4durability 4/4· Peter Howitt

“If you're an optimist like me, you think, 'Well, the you know, the naysayers are going to be proved wrong again.' But there's no theorem that I can prove that shows you that that has to be right.”

0.65

In the late 1980s-1990s, academics wasted time learning how to use word processors (changing font sizes) instead of doing substantive work; this learning period appeared as lower productivity but was actually capital formation in computer skills.

normativehigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“I remember in the late 1980s and and uh on into the '90s, uh when I used to go to the faculty club for lunch, uh before that we used to talk about the latest article in Econometrica or something. But now what were we doing? We were talking about how to use our word processors to uh to you know to increase the font size or something. I mean it's just appeared to be a pretty big waste of our human capital. But but that's what people you know people were doing. We were we were learning about the new technology.”

0.65

Preston McAfee and Howitt wrote 'Animal Spirits' paper applying self-fulfilling expectations to explain how economies could enter low-level equilibrium with high unemployment without external shocks.

normativehigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“One with uh Preston McAfee who then went on to become the chief economist for Microsoft and uh is now uh now now works for Google. Um uh called Animal Spirits where I think we we provided provided a really uh neat little theoretical argument for how an economy uh uh could uh could learn somehow or other to enter into a low-level equilibrium with high unemployment uh just because it's a self-fulfilling expectation.”

0.64

Interest rate control requires a coefficient of reaction greater than one (responding by more than 100% to inflation changes) to avoid explosive instability in the inflation rate under wide variety of circumstances and model classes.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“The coefficient of reaction has to be greater than one and if not the you know, what I showed was under a wide variety of circumstances with a wide variety of different classes of models, you are going to get you are going to get explosive instability in the inflation rate.”

0.63

Employment in manufacturing is falling even in China because technology is advancing so rapidly, showing that automation is a global phenomenon driven by technology development speed rather than geography or wage levels.

factualhigh valueestablishednovelty 0/4durability 3/4· Peter Howitt

“Manufacturing employment is falling even in China because technology is advancing so rapidly. And so they So then they go into services.”

0.63

The rise of superstar firms—not just Google, Amazon, Apple, Microsoft, but throughout most sectors—with rising market share and high markups has been documented by David Autor and others, representing concentrated monopoly-like power that increasingly dominates sectors.

factualhigh valueestablishednovelty 0/4durability 3/4· Peter Howitt

“There's been this rise of the superstar firm. Not just, you know, not just the ones you've all heard about about Google and Amazon and so forth and Apple, Microsoft, but throughout most sectors of the economy, uh the degree of concentration is rising. And the the share of output produced by uh by large uh productive, in many cases, uh firms charging high markups, uh also paying high wages, um has has has grown.”

0.63

Expected rates of return on stock market investments today are not much different from historical returns; there will be many winners and losers, and a sensible person diversifies rather than concentrating bets.

normativehigh valueestablishednovelty 0/4durability 3/4· Peter Howitt

“I I tend to side with the efficient market people who would tell you the expected rate of return isn't much different from investing in the stock market now than than it ever has been. Uh you'll be a lot of winners, a lot of losers. Uh you know, if you get lucky and you buy an Nvidia when it's just had its IPO and uh hold onto it, you you're going to become, you know, a very rich person, but uh uh there are a lot of others that are just going to go bust and uh you know, a sensible person just diversifies, I think.”

0.62

Howitt is confident new employment opportunities will emerge because he believes there is no limit (or at least not one we're close to) that satisfies all human desires, so many ways exist for smart people to create new services.

forecasthigh valuecontestednovelty 1/4durability 3/4· Peter Howitt

“I'm confident that we're going to uh but it's largely because I tend to be optimistic and you know I don't think that there's a limit that we're at least not one that we're anywhere close to that completely satisfies human desires that uh that that I think there a lot of ways in which smart people can have can help other people with uh new services that they're to be creating.”

0.61

Howitt wrote on Canadian monetary policy and dangers of pegging interest rates too tightly; he argued tight pegging could trigger Wicksell's cumulative process where inflation spirals out of control.

normativehigh valueestablishednovelty 1/4durability 3/4· Peter Howitt

“I also wrote I I used to uh write a lot about Canadian monetary policy when I was an economist in Canada. And uh that led me to believe that interest rate control was really something that had a danger of getting out of control, and I wrote something that was very close to what became known as the Taylor rule. But I wrote this paper on uh the uh the dangers of trying to peg interest rates too tightly in the face of changes in the state of the economy and how that might lead to what was known at the time as Wicksell's cumulative process”

0.61

Government policy including antitrust actions and regulation is merited to prevent hoarding of corporate monopoly power.

normativehigh valuecontestednovelty 0/4durability 4/4· Peter Howitt

“And do you think that merits government policy, whether it's regulation or you know, antitrust actions or something to prevent the hoarding of corporate power in monopolies? I do.”

0.61

For companies spending massive capital on AI, monopoly-like rents may not exist to justify the expenditure; the juice may not be worth the squeeze for individual firms even if it's worth it for society.

causalhigh valuecontestednovelty 2/4durability 3/4· Unidentified Speaker — Discussing A.I. Displacement with Nobel Laureate in Economi… [aKSmMLOj8SM]

“What about some industries where the monopoly-like rents that the companies are able to extract to justify the capital expenditure doesn't exist at all? So, is the juice worth the squeeze? First For society, the juice can be worth the squeeze, but for the actual companies that are spending this massive amount of money, might it not be worth the squeeze?”

0.60

Joseph Schumpeter's concept of growth through creative destruction—formulated in his 1942 book Capitalism, Socialism, and Democracy—was treated as a curiosity by economists because it didn't fit into most economic theory, but it is actually essential to understanding long-run growth.

factualhigh valueestablishednovelty 0/4durability 4/4· Peter Howitt

“The idea of growth through creative destruction is actually attributable to the great Austrian-American economist Joseph Schumpeter who taught at Harvard for many decades. Uh, and it was the idea was formulated in his 1942 book called Capitalism, Socialism, and Democracy. And of course, the idea is pretty straightforward. Just that economic growth in the long run comes about through technological progress... People used to regard that as just a curiosity, but Philippe and I thought it really did belong at the heart of economic growth.”

0.59

Elderly care represents tremendous potential for expansion with aging populations in North America, Europe, and even China, providing demand for new employment and services.

forecasthigh valueestablishednovelty 0/4durability 3/4· Peter Howitt

“You think particularly about elderly care. We've got an aging population in North America and in Europe and many many other parts of the world, even in even in China. And uh and uh there's tremendous potential for uh for elderly care uh to to be improved. Uh and there's going to be a lot of demand for it.”

0.59

Unemployment spikes in long-term US data are almost entirely due to demand recessions, not technological displacement; technology doesn't cause 900 basis point unemployment increases.

factualhigh valuecontestednovelty 1/4durability 3/4· Unidentified Speaker — Discussing A.I. Displacement with Nobel Laureate in Economi… [aKSmMLOj8SM]

“Long-term being an optimist in in economics, particularly when it comes to innovation, that's been that's the right call. But, when I think of you know, like you know, picture in your mind like the long-term chart of US unemployment. Almost every spike, I pretty think every spike, is due to demand, a recession. Like has there ever been a a technology that actually causes the unemployment rate to go up by, you know, 900 basis points?”

0.57

The Shiller cyclically adjusted price-earnings (CAPE) ratio is currently at or above levels seen at the dot-com bubble peak and higher than 1929, suggesting historically low expected returns over the next 10-20 years even if no crash occurs.

factualhigh valueestablishednovelty 0/4durability 2/4· Peter Howitt

“You know, you take a look at the cyclically adjusted price earnings ratio of the stock market, the Shiller index. It's up around where it was at the very very peak of the dot com bubble. Uh I believe it's up higher than it was in 1929. It doesn't mean that there's going to be a stock market crash next year or even the year after or the year after, but it historically it's meant that the average rate of return over the next 10 to 20 years in the stock market is probably going to be really low.”

0.57

What is unique about artificial intelligence compared to previous general-purpose technologies is the speed at which new innovations are taking place and the technology is being improved and deployed.

factualhigh valuecontestednovelty 1/4durability 2/4· Peter Howitt

“What seems to me unique about artificial intelligence is the speed which the new innovations are taking place, the speed at which the technology being improved and then deployed in different directions and so on. It's not going to be 40 years before we we fully understand you know well, I wouldn't say fully, we'll never fully understand the implications.”

0.56

A flexible innovative financial system requires willingness to tolerate losses for long periods because no successful innovator avoids a long period of failures.

causalhigh valuespeaker onlynovelty 2/4durability 4/4· Peter Howitt

“Part of the requirement for a flexible innovative system is a financial system that is willing to tolerate losses for long periods of time. Because no one is a successful innovator without having gone through a long a long period of failures.”

0.56

Howitt's views on creative destruction have shifted toward seeing it more as a social and political phenomenon than purely as an economic growth mechanism; he now emphasizes that incumbents use monopoly power to prevent replacement rather than just to continue innovating.

factualhigh valuespeaker onlynovelty 2/4durability 4/4· Peter Howitt

“I guess I tend to see it as uh more as as a social and political phenomenon than I used to. I was just I used to be interested in it mainly just as as a a tool for understanding how economies grow and and so on. And I still do. But I think that the the fact that once an innovator succeeds in acquiring some monopoly power, they don't just use that monopoly power to try to continue innovating. They use it to try to prevent others from replacing them. I I see that as being much more important now than I used to.”

0.55

Meta's acquisition of Instagram is an example of illegal monopoly power use: Zuckerberg's email saying 'I don't think we can compete against Instagram, let's just buy them up' shows a firm choosing acquisition over competition, preventing market-driven innovation.

factualhigh valuecontestednovelty 1/4durability 3/4· Peter Howitt

“When uh there was this trial of uh of Meta for example where uh where Zuckerberg apparently there was an email in which he he just flat out said, "Well, uh I don't think we can compete against Instagram. Let's just buy them up." And... Well, I mean, that should be that should have been grounds right there for you know, declaring that this this is a illegal use of monopoly power”

0.55

Milton Friedman was right that the Federal Reserve's failure to maintain the banking system in the early 1930s was critical to the Great Depression's severity; if the Fed had acted like Ben Bernanke did in 2008, the Depression would have been different.

factualhigh valuecontestednovelty 1/4durability 3/4· Peter Howitt

“I think I think uh Milton Friedman was quite right in this that that the the failure of the Federal Reserve the the brand new Federal Reserve system to maintain the banking system uh and and and to to prop it up the way Ben Bernanke did uh 70 years later uh 80 years later was was really critical to all of that.”

0.55

Graduate students currently function both as research assistants and as co-authors, with the role often starting as research assistance but evolving to co-authorship as students develop competence and ideas, and AI may accelerate this transition by reducing the assistant phase and increasing the co-author phase.

forecasthigh valuespeaker onlynovelty 2/4durability 3/4· Peter Howitt

“graduate students don't just do research assistants. They work as collaborators. I have a number you know, of papers that I've written together with graduate students... and they became co-authors... I so I think what's going to happen with the new technology is that they're going to be more efficient. I won't use them as much as research assistants cuz I've got chat GPT. I'm going to use them more as co-authors. And that's going to speed their uh professional development along a lot faster.”

0.52

AI will turn many markets into superstar markets where the most productive people, even with only marginal edges in productivity, capture so much of the market that less productive competitors cannot survive, necessitating new occupations for people to pursue.

forecasthigh valuespeaker onlynovelty 2/4durability 3/4· Peter Howitt

“Of course, the downside is that this is going to make a a lot of mar- turn a lot of markets into superstar markets, markets where the very the most productive people, even if they just have a marginal edge in productivity, are going to be able to satisfy so much of the market. And people are going to want them rather than the slightly less productive ones, and there isn't going to be room for a lot of other uh competitors in those markets... So, we're going to need new occupations. We're going to need new things for smart people to do.”

0.52

Howitt believes material standards of living in developed economies like North America and Europe will continue to double approximately every generation, at a pace he doesn't see stopping.

forecasthigh valuecontestednovelty 0/4durability 2/4· Peter Howitt

“We continue to go at the pace of doubling the material standard of living of of societies like North America or Europe every generation and I don't see that stopping.”

0.52

Monopoly rents from technology can come from government grants (water, electricity), brands (Coca-Cola), or technology; technology-based monopolies vary widely in duration—some last 70 years (AT&T) while others last only a few years.

factualhigh valuespeaker onlynovelty 1/4durability 4/4· Peter Howitt

“One I could say yeah, government-granted monopolies and they can't extract the rents too high for waters and electricity and the such. Others are maybe not technological uh like you know, Coca-Cola that obviously does not have a monopoly, but it represents something in people's in people's brains, the brand. In terms of companies that have a monopoly because of technology, how common is it for them to maintain that lead?... Some of them go for a long time. I mean, the AT&T Corporation held its monopoly for about 70 years.”

0.48

Knowledge workers can become much more productive with AI assistance because it will handle menial tasks—such as medical workers currently spending time entering data into laptops—freeing them to do what they're trained to do: help patients or other substantive work.

causalhigh valuespeaker onlynovelty 1/4durability 3/4· Peter Howitt

“In a lot of things, when people are using their brains, they're going to be much more productive with the use of generative artificial intelligence. They're going to take care of all sorts of menial tasks. You think think of medical workers who spend a lot of their time nowadays uh with their little laptops entering information in there and uh keeping records... A lot of this can be really sped up tremendously by just entering the things verbally and having you know having the speech recognition do it and having it correct mistakes for you... instead of wasting their time entering data and so on they're going to be doing what they're supposed to be doing is helping patients.”

0.48

AI could reverse the productivity growth slowdown by creating a new wave of innovations, but there is no guarantee; many factors contribute to growth and no country does everything perfectly, with no single 'silver bullet' for growth.

forecasthigh valuespeaker onlynovelty 1/4durability 3/4· Peter Howitt

“This could all be changed, of course, if artificial intelligence turns out to be as as productive as many people think it it could be. Uh it may be that suddenly we're we're going to be on this new wave of innovations, and productivity growth is going to pick up again. Uh you know, no country does everything perfectly. Uh and there's no silver bullet for economic growth. Or for you know, when we first started writing about growth, there were a lot of people claiming that no, the real key to economic growth is human capital. No, the key is being open to to international trade. No, the key is something else. Uh I don't believe any of that.”

0.48

The US has an innovative and flexible financial system capable of compensating for other deficiencies; this is why new AI companies like Anthropic (3 years old) and OpenAI (10 years old) have risen to dominance in a sector previously dominated by mega-corporations.

causalhigh valuespeaker onlynovelty 1/4durability 3/4· Peter Howitt

“In the United States, for example, our ability our our our extremely uh innovative and flexible financial system is going to compensate for a lot of this. I mean, you take a look take a look at at where where all of these advances in in generative AI are coming from. Uh right now you've got, you know, the the big companies, the Googles and the Microsofts and uh it going at it. Uh and everybody that's doing it, I suppose, and doing it at the edges is big. But some of them are very new. I mean, Anthropic is what, 3 years old? Even OpenAI is only about 10 years old. These are These are new companies. They're companies that have risen up despite the fact that they're in a high-tech sector that's being increasingly dominated by a few very, very powerful firms”

0.46

Extremely large capital expenditures on AI with revenues and profits that don't exist are normal for new technologies; the railroad boom in 19th-century UK involved about 15% of GDP invested in railroads, far more than current AI capex as percentage of GDP.

factualhigh valuespeaker onlynovelty 1/4durability 4/4· Peter Howitt

“If you take a look at the the dot-com boom, which is perhaps as much a sort of telecom boom as anything else. Um, I I I remember people being just staggered that this new company, Amazon, I had never made a profit and all they did was sell books. Had a cap of market capitalization greater than that of General Motors... You take a look back in the in the 19th century at the railroad boom in the UK. Investments in in the railroads there um, it's been a while since I looked at at at the data, but I think it was somewhere around 15% of G- GDP was being invested in railroads at those times. That That's way bigger than we're we're spending on artificial intelligence now.”

0.44

ChatGPT is, at its present stage as a just three-year-old technology, by far the most efficient research assistant Howitt has ever had, making him much more productive in his academic work.

factualhigh valuespeaker onlynovelty 0/4durability 3/4· Peter Howitt

“Even at its present stage, it's just a three-year-old, Chat GPT uh is by far the most efficient research assistant I've ever had. That's making me much more productive.”

0.44

AT&T maintained its monopoly for about 70 years, but other technology leaders like Tesla (which had 90% of electric vehicle sales) lost their dominance in just a few years, overtaken by Chinese competitors.

factualhigh valuespeaker onlynovelty 0/4durability 3/4· Peter Howitt

“Some of them go for a long time. I mean, the AT&T Corporation held its monopoly for about 70 years. Uh now they they were helped by by having having this monopoly being recognized as being legal. Yeah. Others obviously just last a few years and then and then they're they're leapfrogged or or or joined by by by others. You take a look at at Tesla. It had didn't have a monopoly on electric vehicles, but it was the dominant leader and had I don't know, some somewhere in the order of 90% of global sales. They're nowhere near the world's leaders anymore. They've been way overtaken by by more than one Chinese company and You know, that only lasted a few years.”

0.41

In the specific case of 2022 inflation (8.8-9%), the appropriate Fed response would require a more complete analysis accounting for supply-side disruptions from COVID, not just the demand-side rule from Howitt's general theory.

causalhigh valuespeaker onlynovelty 0/4durability 3/4· Peter Howitt

“Well, no I mean if you're talking about the the Volcker experiment interest rates got a lot higher than the... Oh, yeah. Um Yeah, the that that particular inflation was a different had a had a it was of a different sort. I mean there was this there were a lot of things going on in the supply side that I was just ignoring and you know, a more complete analysis of monetary policy would have to take those supply side things into account. I mean when you had all the disruptions from COVID and so on that that we're taking place. I think that was that was quite special.”

0.40

ChatGPT in response to a prompt about the job-destroying potential of generative AI produced a beautiful, well-thought-out paragraph on different sectors where job destruction is likely and why, suggesting the technology may eventually perform cognitive tasks previously thought to require uniquely human judgment.

factualhigh valuespeaker onlynovelty 0/4durability 2/4· Peter Howitt

“I had asked uh Chat GPT I I put in the prompt, "Write me a short paragraph describing the job-destroying potential of generative artificial intelligence." And in a few seconds, it produced um a beautiful, well-thought-out paragraph uh that explained how the different sectors in which the uh it's likely that this is going to happen and and why because uh generative artificial intelligence seem to be capable of performing uh cognitive tasks that have been previously thought to require uniquely human judgment”

0.24

From 1987 at MIT, Howitt wrote papers in 1992 on (1) creative destruction with Aghion, (2) 'Animal Spirits' with Preston McAfee on self-fulfilling low-employment equilibria, and (3) monetary policy dangers of tight interest rate pegging, all of which he was proud of, with 1992 being a very productive year.

factualspeaker onlynovelty 0/4durability 4/4· Peter Howitt

“That same year I produced a first draft of uh of a couple of other papers that that I was quite proud of. One with uh Preston McAfee who then went on to become the chief economist for Microsoft and uh is now uh now now works for Google. Um uh called Animal Spirits where I think we we provided provided a really uh neat little theoretical argument for how an economy uh uh could uh could learn somehow or other to enter into a low-level equilibrium with high unemployment uh just because it's a self-fulfilling expectation.”