How AI automation can fulfill Thomas Piketty's predictions on rising economic inequality, and why highly progressive taxes on capital can help slow the spiral
Piketty was wrong about the past. He's probably right about the future. — 1. Introduction X: @briancalbrecht , @saikatc , @harryh , @jankulveit , @daniel_271828 , @krishnanrohit , @besttrousers , @jsevillamol , @mortenstostad , @robinhanson , @t_holden , @erikbryn , @daveg , @dwarkesh_sp , @stevehou , @deepdishenjoyer , @alltheyud , @elocinationn , @garrytan , @garrytan , @briancalbrecht , @pawtrammell , @pawtrammell , @yonashav , @briancalbrecht , @tszzl , @jackwiseman_ , @alltheyud , @digeconlab , @mylesudland , @ks_kulk , @garrytan , @drydenwtbrown , @dwarkesh_sp , @dwarkesh_sp , and @dwarkesh_sp . LinkedIn: Viktor Shvets Bluesky: @druce.ai and @tachikoma.elsewhereunbound … . Mastodon: @carnage4life@mas.to X: Brian Albrecht / @briancalbrecht : “If the economy grows 100x, wages must also grow 100x for labor's share to stay at 2/3. But prices are relative—so this means human labor becomes 100x more expensive compared to AI-produced goods... For labor share to hold steady as that ratio grows to 1,000x, then 10,000x, the @saikatc : Dwarkesh is one of the few people actually thinking deeply about the implications of runaway AI. We should be planning now for how to make sure a world with full automation, should it happen, leads to a good life for everyone, not just the capital holders. Unfortunately, if Harry Heymann / @harryh : @besttrousers 'eh. “capital's share of GDP goes to 100%” seems like an insane leap to me. Maybe this time is different, but I wouldn't bet on it. Jan Kulveit / @jankulveit : Great econ thinking as I'd expect from @pawtrammell Yet I have close to zero trust in the conclusions when read as futurism / directly applied to “how the world will look.” Several crucial considerations seem missing: CC1: Capital will likely end up owned by AIs, not humans. Daniel Eth / @daniel_271828 : Dwarkesh makes good points here, but his libertarian default thinking is showing. If powerful counties like the U.S. & China decide it's vitally important for there to be a global minimum tax, there will be. Weak-ass tax havens shouting “but muh sovereignty” won't stop it Rohit / @krishnanrohit : I find this an intriguing line of thought, it is hard to know what a galaxy spanning fully automated civilisation will be like. Many novels have tried to portray it, with varying degrees of seeming verisimilitude. But once you accept these assumptions, of a perfect labour Matt Darling / @besttrousers : Thought this was interesting (though noted some caveats in the replies). Jaime Sevilla / @jsevillamol : As usual from Phil, this analysis is exactly right. I've stared at the problem for years and the two most solid conclusions from human level AI+robots are a falling labor share and explosive inequality. Morten N. Støstad / @mortenstostad : Enjoyed this post by Dwarkesh and Philip. Some thoughts in a thread: Piketty did not argue that r>g means that wealth concentration will indefinitely increase. He argued that the 1900s was a period of abnormally low wealth ineq. because r was low, and g was high. Robin Hanson / @robinhanson : And the obvious solution is Robots-Took-Most-Jobs insurance. Why are so few interested? https://www.overcomingbias.com/ ... Tom Holden / @t_holden : No, AGI does not make wealth taxes a better idea. In a world in which capital is essentially the only input to production, taxing capital reduces the growth rate of the economy. Whereas at present capital taxes have only level effects. So if anything, capital taxes will become Erik Brynjolfsson / @erikbryn : Here's a terrific new post by @pawtrammell, an amazing postdoc at the @DigEconLab and the ever-insightful @dwarkesh_sp. Their argument in brief: in a world of powerful AI, capital, not labor, may be the main source of income. In turn, that implies that inequality will skyrocket, David Galbraith / @daveg : Dwarkesh is right, active redistribution is needed in AI and robotics era for the simple reason that capital owns both the means of production and the labour. This is why a new political framework has to come from a non-luddite Left. The current Right will exacerbate inequality Dwarkesh Patel / @dwarkesh_sp : @allTheYud We didn't mean to take a position on whether human property rights survive the singularity. Just that *if* owning capital still means anything, inequality explodes. If property rights get totally washed out, we've got bigger problems than inequality. Steve Hou / @stevehou : “With full automation, capital's share of GDP goes to 100% (since datacenters and solar panels and the robot factories that build all the above plus more robot factories are all “capital")." No, lots of issues with the arguments in this blog. Among the most basic ones is that it @deepdishenjoyer : funny that dwarkesh is getting shat on for saying something that's obviously true-this was the original gist of my “you have two years to accumulate capital until the music stops” meme. but now that ai equity has blown up the tech oligarchs are working overtime to squash dissent Eliezer Yudkowsky / @alltheyud : @dwarkesh_sp What is with this huge, bizarre, and unflagged presumption that property rights, as assigned by human legal systems, are inviolable laws of physics? That ASIs remotely care? You might as well write “I OWN YOU” on an index card in crayon, and wave it at the sea. Nicole / @elocinationn : @yonashav @dwarkesh_sp Aceomglu wrote about this and considered a ‘full automation’ scenario here https://www.aeaweb.org/... Economists weren't using the word ‘AGI’ but they were most definitely studying the effects of technological change under ‘automation’ research Garry Tan / @garrytan : If capital is mobile (are you really going to limit movement? that's an end to freedom!) then tax doesn't reliably stick to “the rich” It just results in less investment, which means fewer new challengers, lower productivity, and ultimately lower wages and weaker competition, a Garry Tan / @garrytan : This pro-redistribution essay describes mechanisms by which moats harden (private-market access, intangible opacity, capital mobility, inheritance/trust commitment tech) and then jumps to redistribution, instead of treating open entry and anti-moat policy as the primary Brian Albrecht / @briancalbrecht : Quick Gemini graph. Output is area under the MPK curve. Capital gets paid RK. Workers get the rest, under black curve above supply of curve at R. “Robots building robots” means MPK declines less. Workers get smaller share. But any tax generates a BIGGER wage loss Philip Trammell / @pawtrammell : @garrytan To my mind, the main point of our section 2 is to dispute Piketty's long run series, and a main point of our section 3 is to make the case that open markets might not be sufficient to prevent extreme inequality. Philip Trammell / @pawtrammell : @BrianCAlbrecht We don't assume a perfectly elastic supply of capital at all. We acknowledge that taxing capital (/capital income) can discourage saving. The point is just that the capital share will rise to 1, so eventually it's basically the only thing to tax. Yo Shavit / @yonashav : I consider it a credit to @dwarkesh_sp's whole project that he's managed to elevate the status and legitimacy of AGI-related questions to the point where mainstream economists like Brian are now interested in deeply engaging on takeoff-automation-modeling and its policy Brian Albrecht / @briancalbrecht : Okay. Lots of thoughts on this. Maybe I'll write a full newsletter. But let me focus on one claim: that a progressive global capital tax will be “essentially the only way to prevent inequality from growing extreme.” We have the tools to think through this. Drumroll please... Roon / @tszzl : @dwarkesh_sp @pawtrammell suspect capital assets will be harder to maintain in the future. if you imagine the most powerful capital assets are rivalrous monopolistic goods like land or equity in Meta Land may be worth less if you can build new gleaming cities in the middle of Oklahoma, though it's no Jack Wiseman / @jackwiseman_ : @dwarkesh_sp @pawtrammell I expect inequality to remain essentially unchanged by AI (while the pie gets much bigger!) tl;dr people are going to continue wanting to buy things made by other people and technology will keep getting cheaper. e.g. are people going to watch sports in the future, go to Eliezer Yudkowsky / @alltheyud : @dwarkesh_sp @pawtrammell What is this hopeless delusion about this massive ASI power advantage never finishing up with a brief snap of military force and extermination? @digeconlab : “Though Piketty was wrong about the past, he will probably be right about the future.” Lab Postdoctoral fellow Phil Trammell revisits Thomas Piketty's “Capital in the Twenty-First Century” in a great read that looks ahead to the twenty-second: https://philiptrammell.substack.com/ ... Myles Udland / @mylesudland : The AI vibe is off heading into 2026 and it doesn't get much more off vibe than this https://philiptrammell.substack.com/ ... K Kulkarni / @ks_kulk : Very interesting argument on AI that suggests the future will be fundamentally different from the past. Throughout history, capital and labor have been complements. If the returns on capital got too high, wages would rise in response. This acted as a self-correcting mechanism, Garry Tan / @garrytan : Whoa Dwarkesh! Open markets are way better than redistribution. With ASI, the question isn't “will capital eat the world?” (Piketty's long run series are disputed, btw.) If entry is open, competition forces the gains to spread—prices fall, quality rises, everyone gets richer, no mass Communism needed. If government policy and regulation harden moats and pick winners, you get dynasties. Dryden / @drydenwtbrown : It's fascinating how many people in the valley are incapable of thinking clearly when they see an opportunity to sort an argument into the “capitalism vs socialism” bucket. Garry's “Woah!” chiding gives away the game. “Woah you're not allowed to think that!”, followed by a superficial argument that doesn't engage with Dwarkesh's core points. Dwarkesh Patel / @dwarkesh_sp : I've seen a lot of people misunderstand what we're saying. Our claim is that in a world of full automation, inequality will skyrocket (in favor of capital holders). People aren't thinking about the galaxies. The relative wealth differences in a thousand years—or a million—will Dwarkesh Patel / @dwarkesh_sp : New blog post w @pawtrammell : Capital in the 22nd Century Where we argue that while Piketty was wrong about the past, he's probably right about the future. Piketty argued that without strong redistribution of wealth, inequality will indefinitely increase. Historically, however, income inequality from capital accumulation has actually been self-correcting. Labor and capital are complements, so if you build up lots of capital, you'll lower its returns and raise wages (since labor now becomes the bottleneck). But once AI/robotics fully substitute for labor, this correction mechanism breaks... Dwarkesh Patel / @dwarkesh_sp : The ideas here are @pawtrammell's. I just had the pleasure of getting to talk them over with him. https://philiptrammell.substack.com/ ... LinkedIn: Viktor Shvets : Depressing but realistic preview of the future — In an absolutely brilliant exposition of the likely trends over the next several decades … Bluesky: @druce.ai : Capital in the 22nd Century: Thomas Piketty was wrong about the past. He's probably right about the future. @tachikoma.elsewhereunbound.com : i think economic perspectives on AGI are interesting artifacts, and probably useful as a starting point. but they often seem to fall short of grappling with AGI as autonomous agents, and seem to treat it like some form of capital, an engine or loom, the output of which is owned by a human. Mastodon: Dare Obasanjo / @carnage4life@mas.to : If we achieve AGI then our economic system, capitalism, will be broken. Capitalism assumes you need to pay workers to produce things and the more valuable their output, the more you have to pay them. — AGI implies we need fewer knowledge workers over time and they have little to no leverage as they can be replaced by AI. …