AI Riemann Proof Claim Makes Waves; Verification Could Take Years

A mathematician at Anthropic says he has proved the Riemann hypothesis — one of the oldest unsolved problems in mathematics — using the company’s Claude chatbot. The claim made waves as leading mathematicians gathered in Shenzhen to mark the 200th anniversary of Bernhard Riemann’s birth.

Levent Alpöge, a mathematician at Anthropic, published on Friday what he says is a proof of the hypothesis, one of the seven Millennium Prize Problems, posed in 1859 — each carrying a US$1 million prize from the Clay Mathematics Institute. Instead of a conventional paper, he released about 17 million lines of code written in Lean, an open-source programming language that mechanically checks every step of an argument. No human has read it in full.

The claim landed as the Riemann bicentennial conference opened in Shenzhen. Shing-Tung Yau, a Fields medallist chairing the conference’s scientific committee, told The Beijing News that mathematics “has entered a new era”. James Maynard, an Oxford professor, told The New York Times he would be “98 or 99 per cent certain” the proof is correct. The Guardian reported on Saturday that two independent groups had verified the claim.

Others urge caution. Physicist Tony Padilla of the University of Nottingham told France 24’s Tech 24 that verification “can take years… It’s very early days”. Fields medallist Terence Tao wrote that “the digestion process will take some time”.

The timing — a day after Anthropic’s IPO likely priced above $78 a share — has drawn speculation. Alpöge called the timing “horrible” but said the cynical view of it “is false”. The sources reviewed record no on-the-record comment from Anthropic itself. Brian Conrey, a leading expert on the hypothesis who reviewed an earlier Claude paper for Anthropic, called the moment “the biggest thing that’s ever happened in mathematics”.

Why it matters: The Riemann hypothesis concerns the distribution of prime numbers — the whole numbers divisible only by one and themselves — and much of modern number theory assumes it is true. If the claim holds up, AI will have helped crack a 167-year-old problem. If the only copy of the argument is machine-checked code that no human has read, mathematics faces a new question: when is a proof a proof, and who gets to say?

The post AI Riemann Proof Claim Makes Waves; Verification Could Take Years appeared first on DailyAI.