Capital in the 22nd Century

Every passage, on the record.

1. 00:00Reference

   It's called Capital in the 22nd Century by Philip Trammell and Anton Korinek. If you follow economics Twitter or certain rationalist circles you probably saw it, and you have to read the entire article.

2. 00:46Event · condensed

   It got a lot of attention, unbeknownst to me it had been making the rounds over the holidays.

3. 00:46Reference

   It makes a provocative claim, basically that artificial intelligence is going to create a level of wealth inequality that makes the Gilded age look like some sort of socialist utopia, and that without aggressive global taxation we're heading toward permanent techno-feudalism.

4. 00:46Reference

   That just sounds like word salad, like straight from what's, very fakis, what's that guy's name, the Greek finance minister. I think he wrote an entire book on techno-feudalism.

5. 01:18Claim

   Understanding why smart people, or supposedly smart people, believe wrong things is one of the most valuable exercises that you can do, and it's good to challenge the things that you believe to be true.

6. 01:18Quote

   Nothing gives me more conviction than going through what other people assume to be intelligent and just dismantling it in every facet.

7. 01:59Claim

   Papers like this are the reason that simple investment strategies don't work. The fact that credentialed economists at prestigious institutions can write something this fundamentally confused about the nature of wealth, and have it taken seriously by thousands of readers, tells you something very important about the state of economic understanding in the world.

8. 02:38Claim

   The glaring sign here is that alpha still exists in this space, in Bitcoin. You're going to be very wealthy if you're in it now, because just by understanding a few principles that puts you miles ahead of people who have absolutely lost the plot.

9. 02:38Claim

   If everyone understood what wealth actually is, or capital, or what knowledge does, then investment opportunities like the one we are pursuing here would be arbitraged away instantly.

10. 04:09Analogy

    It's significantly more beneficial to future you knowing where the Bitcoin price is going to be in 10 years versus if it was there tomorrow, because you have the runway to get all your ducks in a row and put them on the route, and the steamroller that is the fiat machine is just going to bowl them over for you.

11. 04:09Reference

    It's the classic Michael Saylor of it, significantly more beneficial to future you knowing where the Bitcoin price is going to be in 10 years.

12. 04:58Claim · condensed

    When smart people at Oxford are writing papers demonstrating their own confusion, that is a huge signal. They are doing you a favor, because it gives you a high fidelity map of the territory of error they have created for themselves.

13. 05:33Reference

    This paper builds on somebody named Thomas Piketty's supposedly famous work from about a decade ago. His big idea was that R is greater than G, the rate of return on capital tends to exceed the growth rate of the economy, so capital owners capture an increasing share of total wealth.

14. 06:35Reference

    If AI can do everything workers can do, and cheaper, labor share of income will eventually go to zero, all returns flow to the capital owners, and if R greater than G continues to hold, those capital owners compound their advantage forever.

15. 07:05Reference

    In the limit case where AI substitutes for everything, you get permanent dynasties. Whoever owns the capital at the moment of full automation locks in their position for eternity, their children inherit it, and it just never ends because there's no labor left to disrupt with.

16. 07:47Quote

    Let's tax them. Let's take their stuff. Let's take their stuff because we're scared.

17. 08:21Claim · condensed

    To their credit, they acknowledge how hard it is to do any of this, the global coordination to prevent capital flight is damn near impossible, it only takes one rogue nation deciding to attract wealthy people.

18. 08:51Claim

    This is a serious paper by serious people, the logic sort of works if you accept their starting assumptions. The giant glaring problem is that the starting assumptions are completely horseshit, and once you see the error at that foundation the whole structure collapses.

19. 09:34Analogy

    Imagine you're on a boat and you notice water coming in. The boat's taking on water, and if you don't do something it's going to sink. You're not going to be able to use politics to solve the hole in the side of your boat.

20. 10:05Analogy · condensed

    Someone writes a sophisticated paper analyzing the water distribution in the boat, modeling the flow rates between compartments, calculating that the people in the lower decks will drown first while the people in first class stay dry longer.

21. 10:36Analogy

    Their proposed solution is to redistribute the water more evenly among the passengers, take some from the lower classes and move it up, take some from first class and move it down, so that everyone drowns at the exact same rate. Equality for all.

22. 11:07Claim

    The problem is boats don't float because water is distributed fairly. Boats float because someone understood buoyancy and hull design and how to seal joints against pressure. The leak isn't a distribution problem, it's a knowledge problem.

23. 11:53Quote

    Wealth isn't some pile of stuff that just sits there growing. Wealth is the set of problems we know how to solve.

24. 11:53Idea

    A hundred years ago we couldn't cure bacterial infections, then somebody figured out antibiotics. That knowledge is wealth. We didn't have the ability to transform the physical world a certain way before, that arrangement of information they crafted became knowledge, and that knowledge is wealth.

25. 13:06Claim

    Robots, just like the pills, will be obsolete when someone figures out something else. The owners of the current robots have no special claim on future knowledge, they didn't own the knowledge that made their robots valuable in the first place, they just owned the physical instantiation of knowledge other people created.

26. 13:06Event · condensed

    There is a replication crisis happening now in science, studies getting published that can't be reproduced when other researchers try to verify them, because there's such a push to publish for grants and tenure that nobody bothers to verify any of it until it's already built on.

27. 14:20Claim

    Information is being generated but verification is failing. It's the same in finance, markets are distorted by monetary manipulation, so no one has any idea what anything is actually worth.

28. 14:20Analogy

    You're building this house where every contractor has a different measuring tape and every day the tape changes. Think about the amount of noise that goes into that process.

29. 15:08Idea · condensed

    All these large language models are just churning out plausible sounding text that nobody takes the time to verify, because you can't stop the slop. Signaling for actual knowledge is very very difficult because the slop is a tsunami that costs almost zero to generate.

30. 15:42Claim

    As information floods in it looks like a lot is happening, but knowledge doesn't grow because there's no error correction, because we're not verifying. Taxing the robots isn't going to fix this, redistributing the water isn't going to fix it either.

31. 16:14Idea

    Capital isn't a thing, capital is embodied knowledge. A factory is valuable because it embodies knowledge about manufacturing processes, supply chains, logistics. Without that knowledge that factory is just a pile of metal. The return on capital is really the return on the knowledge that is embedded in it, and knowledge doesn't compound automatically, it has to keep being true.

32. 17:24Analogy

    The railroad barons owned enormous amounts of capital, and by R greater than G their descendants should own everything by now, their capital should have compounded for 150 years.

33. 18:16Quote · condensed

    How come the Vanderbilts don't own Apple?

34. 18:16Claim

    You can't buy knowledge before it exists. The railroad barons understood land rights, logistics and political lobbying, none of that helped them understand quantum mechanics or transistor physics. Owning railroad track gave you no advantage when the semiconductor revolution happened.

35. 18:16Analogy

    It happened with whaling that switched to petroleum, telegraphs to telephones, mainframes to personal computers, every single transition required new knowledge and new knowledge created new winners.

36. 19:46Claim

    Railroad knowledge stopped being the relevant knowledge, the capital that embodied it stopped generating returns, the R in R greater than G collapsed for that specific capital even as it continued for people who own different capital.

37. 19:46Claim

    If you think wealth is stuff, you ask who owns the stuff. If you understand wealth is knowledge, you ask whether we're still creating knowledge. These lead to completely different analyses and completely different prescriptions, and one is true and one is not.

38. 20:26Reference · condensed

    The paper assumes that once AI can do everything humans can do, the pattern of startups disrupting incumbents will stop, because incumbents can generate the ideas too.

39. 20:26Idea

    This misunderstands what disruption is. Disruption isn't having ideas, it's new knowledge with reach that makes old knowledge obsolete. If AI generates genuine explanatory knowledge, that knowledge will have reach into domains nobody anticipated, and it will disrupt its owners' arrangements as readily as anyone else's.

40. 21:17Claim

    If AI is smart enough to replace all human labor, it's smart enough to create new knowledge, and new knowledge is inherently disruptive. The only way to get permanent dynasties is to stop knowledge creation entirely, and then the dynasty won't be permanent either, because the dynasty will be gone.

41. 21:17Reference · condensed

    They extrapolate trends in capital accumulation out to the 22nd century, where R greater than G combined with AI substitution yields predictable concentration.

42. 22:03Claim

    The future is not predictable because knowledge creation is not predictable. If we could predict what we'll know in 2100, we'd already know it. That's a logical impossibility, not some practical limitation.

43. 22:37Analogy

    The railroad folks couldn't predict semiconductors, and the semiconductor pioneers couldn't predict the internet. Each wave of knowledge created a future that was invisible from the perspective of the previous wave.

44. 22:37Reference

    Their response is that the solution is redistribution. Once you think of wealth as stuff that compounds for its owners, redistribution seems like the obvious answer.

45. 22:37Quote

    Tada, guess who's gonna be in charge of taking stuff, and probably me, because I came up with the idea and I'm super smart.

46. 23:15Claim

    Redistribution doesn't create any knowledge, it might even undermine it. Taxing capital discourages saving, and saving is deferred consumption, people choosing to solve problems for others now while consuming less themselves.

47. 23:15Claim

    Redistribution requires institutions that work, that can measure wealth accurately, collect taxes efficiently and distribute benefits fairly. If those institutions are degraded, if verification is failing everywhere, redistribution becomes another arena for corruption and extraction.

48. 24:20Reference

    Inside the paper they talk about the Gini coefficient as the metric that actually matters.

49. 24:20Claim

    Absolute capability matters more than relative inequality. The poorest person today has access to antibiotics, refrigeration, telecommunications and clean water that the richest person in 1800 couldn't buy at any price. Relative inequality between now and then is incomparable because the categories have completely shifted.

50. 24:20Claim

    If knowledge creation continues, absolute capabilities keep expanding and the floor keeps rising. But if knowledge creation stops, the floor stops rising and relative inequality becomes zero-sum conflict over a fixed pie.

51. 25:51Idea

    This is that quality constraint that I always talk about. Information becomes knowledge when it's stabilized by constraints that make falsification expensive while keeping verification cheap.

52. 25:51Analogy · condensed

    The go-to example is DNA, it has multiple independent repair mechanisms that catch errors. Scientific theories have this property too, but only when they're subjected to replication and independent testing.

53. 26:34Claim · condensed

    When constraint quality is high, knowledge accumulates and our ability to transform the physical world grows. When it's low, information floods in but nothing sticks, there's no persistence.

54. 27:21Claim · condensed

    The replication crisis in science is a constraint quality crisis. Same with financial markets. AI systems generating unverifiable text are going to suffer from the same thing.

55. 27:55Quote

    You can't redistribute your way out of a knowledge crisis. You can't tax your way to truth.

56. 27:55Claim

    If you maintain high constraint quality, knowledge keeps growing, problems keep getting solved, absolute capabilities keep improving, and distributional questions become less zero-sum because the entire pie keeps expanding.

57. 28:28Idea

    This entire train of thought is Kabuki theater. They're proposing elaborate policy solutions to a problem they've completely misidentified, effort spent researching optimal tax rates and modeling international coordination, when they could have spent the entire time asking what wealth is, what capital is, what creates it, what maintains it.

58. 29:26Idea

    This confusion is the milkshake. Alpha exists when you understand something the market hasn't priced in. If everybody understood that wealth is knowledge and capital embodies knowledge, the arbitrage between what is true and what most people think is true would disappear.

59. 30:48Claim · condensed

    The authors worry about who's going to own the robots and propose taxes to redistribute from capital owners to everybody else, but they never ask what if there was a form of capital that anyone could own, that didn't require permission, couldn't be debased or easily confiscated, and didn't depend on institutions.

60. 31:21Idea

    Bitcoin is a system where verification is cheap and falsification is prohibitively expensive. Anybody running a node can verify all of the transactions, that's high constraint quality by design.

61. 31:21Claim · condensed

    Ordinary people can buy Bitcoin on their phone today and no one can stop them, so the concern about being locked out of capital ownership doesn't apply the way the paper assumes.

62. 32:14Claim

    Capital is constantly being extracted through monetary manipulation, but Bitcoin can't be debased and it can't be stopped, it moves at the speed of information. There's no way to stop capital flight in the Bitcoin space.

63. 32:14Claim

    Every time a paper like this gets traction, the gap between people who understand and people who don't gets wider, the confusion gets reinforced and the wrong policies get proposed, and that means there's an opportunity to opt out of the confusion and own something with high constraint quality.

64. 33:52Claim

    If you get wealth right, distributional questions become irrelevant because the pie keeps growing and new knowledge keeps disrupting old arrangements, so dynasties cannot persist. If you get it wrong, no amount of clever policy is going to save you, you're redistributing water on a sinking boat, playing Kabuki theater while the actual problem goes unaddressed.

65. 33:52Claim

    We're at a critical moment where constraint quality is under pressure everywhere, and the flaw is that this paper treats it like a distributional crisis when in fact it's an epistemological crisis. The question isn't who gets what share, the question is whether we can still tell what's true, whether we can filter the signal from the noise.

66. 35:15Prediction

    Where these guys see a system that can only be prevented by global coordination that even they admit is nearly impossible, I see a future where knowledge keeps getting created, where new knowledge keeps disrupting old arrangements, and where the floor for everyone keeps rising even if the ceiling rises faster.

67. 35:15Prediction · condensed

    Their scenario is not true because Bitcoin exists. If it didn't exist, that future would be a real possibility. But because it exists there is no way to turn it off or stop it, people will find it and shield themselves from all of this.

68. 35:15Analogy

    If they shield themselves long enough from all of this, it's going to look like they're getting significantly more wealthy, and everyone around them will say, what is Bob doing, why is Bob's life so much better and mine sucks.

69. 36:14Claim

    That one thing requires you to do nothing, it's super easy, anyone can do it with any amount of value they wish to input, from anywhere at any time, and no one can stop them.

70. 36:14Claim

    All we have to focus on is whether we're maintaining the conditions which are ideal for knowledge creation. That's the bet we're making, that's where all the effort should go, and it just so happens that Bitcoin allows for the ultimate ecosystem for knowledge generation.