Seven in ten Americans think AI will make it harder to find work; nearly a third fear for their own jobs. AI bosses themselves warn of disruption: Dario Amodei of Anthropic has warned of 10-20% unemployment, Bill Gates says in an AI world people will not be needed for "most things", and Sam Altman speaks of "disruption/significant transition".
Since 1300 GDP growth per person at the world's frontier economy has never exceeded about 2.5% a year (Robert Gordon of Northwestern University), capping the pace at which technology can destroy jobs. Farm employment in England has been falling steadily since the 16th century without ever collapsing. Mid-20th-century "white heat of technology" brought 2.5% annual American growth and twice today's level of job disruption, yet is remembered as an era of rising wages and unpolarised politics.
The Industrial Revolution's "Engels' pause" (1790-1840 wage stagnation) is often invoked by Silicon Valley but Nicholas Crafts argues it is "not a template" for AI: employment composition saw little churn until the 1850s, between 1760 and 1860 the British workforce ballooned from 4.5m to 12m, and stagnant real wages reflected rising food prices, not exploitation.
The Economist's analysis of National Association of Colleges and Employers data found that 2022-2024 graduates in the least AI-exposed quintile saw a 1.5 percentage-point drop in full-time employment, while those in the most exposed quintile (computer science, computer engineering, information science) suffered a 6.6 point drop. For the class of 2025, the full-time employment rate in heavily exposed fields fell from nearly 70% to 55%. Undergraduate enrolment in computer science fell 11% in 2025; in computer programming, 26%. Erik Brynjolfsson of Stanford found AI-exposed young workers' employment had fallen 16% relative to less-exposed peers. Economists at Google's Zanna Iscenko and Fabien Curto Millet found the trend predated ChatGPT.
Robot taxes, token taxes, higher capital taxes, lower labour taxes and partial nationalisation of AI firms have all been proposed. A South Korean presidential adviser's "national dividend" proposal—funded by excess tax revenue from the chip boom—briefly sent the local stockmarket down 5% before recovering. America's politicians murmur about distributing AI-company shares via "Trump accounts". Denmark's active labour-market policies are cited as a model for helping workers retrain.
One way to stop a runaway horse is to bet on him.