Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
AI-doomer fan fic FTW
Yesterday, following the recent pattern, the shares of various companies that might be vulnerable to AI disintermediation fell — software, asset management and, joining the party for the first time, banks. The most popular explanation for the renewed jitters was a blog post from Citrini Research about how AI might get a lot of high earners fired and tank the economy. From the FT:
Investors have recently seized on social media rumours and incremental developments by small AI companies to justify further selling, with a widely circulated blog post by Citrini Research over the weekend describing how AI could hypothetically push the US unemployment rate above 10 per cent by 2028, proving the latest catalyst.
The most important thing about the post is not what it says. It is that the stock market has reached the point where blog posts cause significant stock moves, or at least where people think that they do. It is often true that markets sell off for reasons that are obscure or all but unknowable, and people like me have to feel around for an explanation. This may be one of those cases. Either way, the Citrini fuss is further evidence that we are in an expensive market that is looking for an excuse to fall, for reasons that are probably wider than just AI.
But what about the post caught the market’s attention? It takes the perspective of several years in the future, from which it reflects on a recession and financial crisis that have already happened. “What follows is a scenario, not a prediction . . . The sole intent of this piece is modeling a scenario that’s been relatively underexplored,” it begins. But the opposite is true. The post took off because it is an extreme version of one of the most explored scenarios of all time: robots turning on their creators. It’s been explored to death at least since Karel Čapek introduced the word robot and, at the very same moment, predicted that robots might cause a lot of trouble. That was over a hundred years ago. More than a century before that, some weavers in Nottingham, under the banner of the fictional General Ned Ludd, explored the theme by smashing up some textile mills.
On to the post’s argument. The central scenario is that AI destroys jobs and income, but does not replace them, dragging the economy into recession and markets into a crisis:
The owners of compute saw their wealth explode as labor costs vanished. Meanwhile, real wage growth collapsed . . . white-collar workers lost jobs to machines and were forced into lower-paying roles . . . When cracks began appearing in the consumer economy, economic pundits popularized the phrase “Ghost GDP”: output that shows up in the national accounts but never circulates through the real economy . . . The velocity of money flatlined. The human-centric consumer economy, 70 per cent of GDP at the time, withered . . . AI capabilities improved, companies needed fewer workers, white collar layoffs increased, displaced workers spent less, margin pressure pushed firms to invest more in AI, AI capabilities improved . . . It was a negative feedback loop with no natural brake . . .
The irony of this was that the AI infrastructure complex kept performing even as the economy it was disrupting began deteriorating. NVDA was still posting record revenues. TSM was still running at 95 per cent+ utilization. The hyperscalers were still spending $150-200 billion per quarter on data center capex. Economies that were purely convex to this trend, like Taiwan and Korea, outperformed massively.
Something struck me as odd about this picture, although (with my non-existent formal education in economics) I struggled to articulate what it was. It feels like financial commentary that considers only one side of a trade (“stocks went down as sellers outnumbered buyers”, etc). Citrini pictures a world of massively increasing productivity accompanied by a collapse in consumption. Does that make sense?
Joseph Steinberg, an economist at the University of Toronto, helped me flesh out my intuition that something is amiss. “The first part of the argument I hope my economics students would flag is the bit about ‘ghost GDP’,” he told me. What does it mean for output to “show up in the national accounts but never circulate through in the real economy”? If GDP is rising — all those robots out there making stuff, faster and faster — then something on the other side of the national account identity has to be rising, too. The possibilities are consumption, investment, government spending or net exports. In the Citrini scenario, consumption is falling, fast. So is government spending rising (on the basis of taxing or borrowing from who, exactly)? Or exports — to other countries undergoing the same crisis?
None of that makes loads of sense, so that leaves investment. But investment, Steinberg points out, is only sustainable on the basis of future consumption; “it only makes sense to invest in AI if there is income to buy these things the AI is generating”.
It seems to me there has to be a distributional element to the argument that Citrini does not fully flesh out. Income and consumption rises, but this happens somewhere that doesn’t benefit your average US white-collar worker. Maybe it is all in the “economies that were purely convex to this trend, like Taiwan and Korea”. Or maybe it all goes to a few, very annoying gagillionaires in Silicon Valley. In this scenario, we all end up bringing very, very expensive lattes to Peter Thiel. Or perhaps Thiel’s wealth is seized by the government and redistributed. But it’s hard to see how the Thiels of the world stay rich for long in the world Citrini envisions, because their AI investments stop paying off when consumption collapses.
None of this is to say that I believe that the AI revolution, like past technological regime changes, must in the end create more jobs and making all of us richer. I don’t. What I am claiming is that the Citrini account of the risks seems incoherent (perhaps readers with a better grasp of macroeconomics can make better sense of it?). The post went viral not because it elucidates the present situation, but because it speaks to ancient fears.
One good read
“A fire chief may have the authority to close roads during an emergency but that doesn’t imply that the fire chief has the authority to impose road tolls.”
FT Unhedged podcast
Can’t get enough of Unhedged? Listen to our new podcast, for a 15-minute dive into the latest markets news and financial headlines, twice a week. Catch up on past editions of the newsletter here.