Latest / Minds, Bodies, and Terawatts / Anthropic Built a Model of Three AI Economies for 2030. In the One Where AI Does What Anthropic Says It Will, Knowledge Workers Face 18% Unemployment and Falling Pay While the Economy Grows by a Third

Transcript

Anthropic Built a Model of Three AI Economies for 2030. In the One Where AI Does What Anthropic Says It Will, Knowledge Workers Face 18% Unemployment and Falling Pay While the Economy Grows by a Third · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, the show where the company that warned you about your job finally did the math on it. Anthropic's research institute published a model this week of three American economies in 2030, with a tool where you type in how capable you think AI gets and it tells you what happens to growth, wages and jobs.
• 0:21
  [Guest] Three futures, and in the modest one AI is the internet again. In the middle one, AI can do half of all knowledge work and the economy grows well above its normal pace. In the extreme one, growth hits fifteen percent a year.
• 0:36
  [Host] Fifteen percent a year is faster than anything on record, so what happens to the people in that one?
• 0:43
  [Guest] The economy lands a third bigger than it would without AI. Anthropic's own working paper puts unemployment among knowledge workers at about eighteen percent in that world, with their pay falling too. And the page says total pay to all workers is barely changed by 2030, so the pie grows by a third and the slice for everyone who works for a living stays the same size.
• 1:06
  [Host] Who counts as a knowledge worker here? I need to know whether that's me.
• 1:11
  [Guest] About six in ten American workers by the model's count, anyone whose job happens at a screen. Anthropic's own example is coders and call-center agents, and its page says they may have to switch to jobs like electrician and nurse, then admits changing occupations entirely is hard and takes many people a long time.
• 1:30
  [Host] So the plan for a laid-off programmer is nursing school.
• 1:34
  [Guest] That is the model's mechanism, not its advice, and the other side of the switch is real. In the extreme case, wages outside knowledge work rise about a third, because nurses and electricians get scarce while desk work gets cheap. If you are in a trade, the scariest scenario on the page is a raise.
• 1:53
  [Host] Okay, but the extreme one is the extreme one. How seriously does Anthropic itself take it?
• 1:59
  [Guest] Less seriously than you'd think, since the page says it would likely require AI that improves itself, adopted fast. And some of the outside economists who read the draft told the team it reads better as a thought experiment than a scenario.
• 2:13
  [Host] A thought experiment? More than a year ago Anthropic's chief executive, Dario Amodei, told Axios that AI could push unemployment to ten or twenty percent, so isn't that the same number?
• 2:24
  [Guest] It is the same number, and the boss's headline forecast now lives in the corner of his own company's model that some reviewers filed under thought experiment. Either he was alarmist or the model is conservative, and the page does not say which.
• 2:39
  [Host] What does it say? They built the thing, they must have a view on which world we're in.
• 2:46
  [Guest] They refuse to say, because the authors call these scenarios, not predictions, and attach no probabilities. They polled Americans last month instead, and the typical answer lands on the middle scenario.
• 2:57
  [Host] Then tell me about the middle one, since that's apparently the consensus. Is that the comfortable one?
• 3:03
  [Guest] Comfortable for the economy, which grows while unemployment stays inside its normal range. But knowledge-worker wages are flat, no raise while the country gets richer, and workers' share of each dollar the economy makes slips below the sixty cents it is today.
• 3:17
  [Host] Is there any scenario on this page where workers keep their share?
• 3:22
  [Guest] Not one, and in the extreme case their share falls to forty-five cents on the dollar. Every version of this future has a bigger pie and a thinner slice, and the three scenarios only argue about how much thinner. The page's own conclusion is that the challenge is not growth, it is making sure the gains get shared.
• 3:42
  [Host] And how does the model share them?
• 3:45
  [Guest] It doesn't try, because policy is left out by design, along with business cycles and, in their words, hyper-capable robots. So the raise for the trades holds only as long as nothing automates hands, which means minds are in this equation and bodies are not.
• 4:00
  [Host] What bothers me is that a company publishes three futures, refuses to say which is likely, and leaves out the one variable that could change the answer. What am I supposed to do with that?
• 4:13
  [Guest] Anthropic did write down candidate fixes in a policy memo earlier this summer, universal basic income among them, and said it was not yet ready to advocate any. So treat the missing number as the message. A model that assigns no odds and holds policy still has quietly shown you the default, a bigger pie and a thinner slice in every future it can imagine. Anything else takes deliberate action the model has not run.
• 4:39
  [Host] So which one are we in?
• 4:41
  [Guest] The one where desk work goes first, one help desk and one code review at a time, and each crossing looks local and survivable until you add them up. For more on what happens when the cheapest reliable worker stops being human, and the three futures that follow from it, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.