The ‘expectation gap’: 91 per cent of Anthropic employees believe that AI will make a Nobel Prize-worthy discovery by 2030, compared with 39 per cent of Americans

A survey carried out in August 2026 amongst a representative sample of American adults and Anthropic staff reveals agreement on what AI is capable of today, but a wide divergence of views on what lies ahead.

Anthropic economists, associated with Charles I. Jones and Anton Korinek, interviewed two populations in August 2026: a panel representing the adult population of the United States and employees of the company. On eight tasks ranging from routine e-mail to Nobel-level scientific discovery, the two groups rank difficulties almost in the same order and agree on the current state of capabilities. It's the pace of progression that separates them: 91% of staff believe that a Nobel-level discovery is possible by 2030, compared to 39% of the public, and according to the tasks, between 19 and 40% of Americans believe that AI will never succeed, compared to 2% at most at Anthropic.

An even clearer gap in one's own employment

The gap is widening when everyone talks about their work: on average, Anthropic employees believe that AI will be able to provide 75% of their current jobs in 2030, compared to 28% for all American workers. More than half of the staff even rely on at least 80%, a proportion that falls to 5% in the labour force. Perceived time savings also diverge: 26% of Americans who use AI at work say they complete a task in a quarter of the time or less, compared to 59% of Anthropic employees. Use plays out: regular users are much more optimistic than those who have never tried an AI tool.

Point of significant convergence: In both groups, the more it is expected that AI will absorb its own work, the more high unemployment is anticipated, and at an equal level of automation expected, unemployment forecasts are almost identical. The disagreement would therefore focus less on economic mechanisms than on the speed of technical progress.