Elon Musk: “AI will beat all fields… maybe 2028, at the latest”

Elon Musk does not deal in maybes — except when the maybe is a year and the claim is total. Told that a well-known AI researcher’s forecast for the future of science was bold, Musk decided it was timid instead, and answered with a single line that folds the whole future of human expertise into two dates.

“AI will beat all fields by the end of next year or maybe 2028, at the latest”

He wrote it in reply to François Chollet on X, after Chollet resurfaced an old thread arguing that nearly every scientific discipline — physics, chemistry, biology, archaeology — was quietly becoming a branch of computer science. Chollet said the once-mocked idea now looked “obvious by the day.” Musk called that reading conservative.

Read it slowly and the pressure is not on the year. It’s on the verb. Chollet described AI becoming essential — a tool no serious lab can work without. Musk swapped that for beat: not assisting the best physicist or oncologist, but outperforming them, across every discipline at once. Those are not the same claim scaled up. One is about adoption, which is already happening. The other is about sustained, verifiable superiority in open-ended research, which no benchmark yet measures and no system has shown. “Mastering a field” has no test you can fail, which is precisely what makes it a comfortable thing to promise.

It is also a promise Musk keeps re-dating. In 2017 he put human-level AI somewhere around 2030 to 2040. By 2024 it was the end of 2025. At Davos in January he moved it to 2026 or 2027. Now it’s 2027 to 2028. The horizon slides forward at almost exactly the speed of time, which is the tell of a forecast built to be unfalsifiable in the present: close enough — “next year” — to land like a commitment, far enough — “2028” — to outlive being wrong. Musk has earned the benefit of the doubt on hard targets before; reusable rockets and mass-market electric cars were also called fantasy. But those were dates he hit. This is a date he keeps needing to reset.

The sharper contradiction is internal. Weeks earlier, when Anthropic’s Dario Amodei made the case for slowing AI down — third-party evaluation, cooperation between labs and governments, room to test between releases — Musk’s reply was two words: “Dario is right.” So the same person is underwriting the acceleration and warning about it in the same breath, forecasting total mastery inside two years while agreeing that the safe move is to go slower. Both can be sincere. Together they describe someone for whom a prediction is not a measurement but a forcing function — a deadline announced loudly enough that the industry, and the capital, feel obliged to chase it.

That is the part worth taking seriously, more than the number. If Musk is right, the institutions built around scarce human expertise — universities, licensing, the slow apprenticeship of a career — have about two years to reckon with becoming optional. If he’s wrong, the cost isn’t his; it’s paid by everyone who reorganized a budget, a syllabus or a hiring plan around a verb that was doing more work than the evidence could carry.

The one field AI has demonstrably conquered so far is the market for confident deadlines. On that, at least, 2028 is already looking conservative.

Tags: artificial intelligence, Anthropic, Elon Musk, Dario Amodei, AI predictions, François Chollet