Hours after Anthropic claims a Claude proof of the Riemann hypothesis, Stanford releases an all-human competing draft
Anthropic mathematician Levent Alpöge announced Friday that he worked with Claude to solve the 167-year-old Riemann Hypothesis, posting a paper and code on his website. Alpöge said the method also resolves the Generalized Riemann Hypothesis (GRH) and could soon cover the Extended Riemann Hypothesis. Extending an idea from mathematician Stepanov, Claude found a way to mimic calculus derivatives on whole numbers by chaining the square root of X special transformations for a sequence of length X.
Because Claude spells out elementary steps, a printed version would top 250,000 pages. Alpöge therefore published the proof only in the Lean language across 14 data packages. Lean verifies every logical deduction, though experts must still confirm the starting problem was coded accurately, and Alpöge plans a human-readable version in the coming months.
Hours later, Stanford professor and American Mathematical Society president Ravi Vakil rushed out a competing, "all human, with no AI" GRH draft by Alessandro Masullo, previously vetted for months by Stanford number theorist Brian Conrad. Roughly eight hours later, Vakil updated his post to report that mathematicians and others, aided by different AI models, had already flagged problems in Masullo's draft that must be resolved before acceptance.
Continue in the app — vote & join in ➔