AIPE Atlas
AIPE Atlas
Keep them separate. The Transition is an empirical question about jobs, already generating data. The Endgame is a structural question about whether human labor has market value at all — where philosophy and political economy take over from labor statistics. Expand the rows to compare.
Which jobs grow and shrink, how fast, who is hit first, how to reskill and cushion the disruption.
Empirical, already arriving: payroll microdata, job postings, productivity studies, usage indices. Most mainstream economists live here.
Labor economists and institutions — Autor, Brynjolfsson, Acemoglu, IMF, OECD, WEF, McKinsey, PwC.
Care, trades, judgment, AI-complementary work, and Autor's historical regularity: new work created by the technology itself.
Disruption: entry-level collapse, broken career ladders, "so-so automation" that cuts labor cost without creating offsetting growth, widening inequality.
Augmentation rebuilds the middle class — AI as a "lever for expertise" that levels up novices (+34% for novice support agents; the biggest writing-quality gains at the bottom of the skill distribution).
Reskilling, job guarantees, universal basic services, capital and robot taxes. Stiglitz: not UBI yet — there's too much necessary human work to pay people not to do it.
Does human labor have any market value at all — and if not, through what channel do humans capture a share of the abundance?
Speculative and forecast-dependent. There is no data from a world that does not yet exist — so models, philosophy, political economy, and fiction carry the argument.
Korinek, Stiglitz, Harari, Bostrom, Varoufakis, the Intelligence Curse authors — and the AI-lab leaders' own essays.
Not market labor: ownership, redistribution, and non-economic life — relationships, play, craft, status, contemplation, care.
Wage collapse, concentration into a "new serfdom," the meaning crisis — comfort without ownership or power.
Broadly-shared abundance: cheap intelligence, cured disease, near-free energy — Banks's Culture as the shared utopian reference the lab leaders explicitly cite.
Citizen's dividends from sovereign wealth funds, universal basic capital, "universal basic compute," AGI governance — predistribution over after-the-fact transfers.
(1) Is human task-complexity bounded? — Korinek's ceiling question, possibly unresolvable in advance. (2) How fast does robotics master the unstructured physical world? — Autor bets it stays hard and expensive for a long time (he points to the slow, costly road of self-driving). The two horizons are separated mostly by how long the last 10% of human capability holds out.