Anthropic Researcher Claims AI Solved Riemann Hypothesis
An employee of the artificial intelligence company Anthropic claimed Friday to have solved the Riemann hypothesis, a 166-year-old problem that carries a $1 million prize, using the firm’s own Claude model to produce a proof he says exceeds 250,000 pages.
The researcher, Levent Alpöge, a Turkish-American mathematician employed by Anthropic, posted an account of the work on his personal website but has not released the proof itself. No independent mathematician has verified it. The Clay Mathematics Institute, which listed the hypothesis among its seven Millennium Problems in 2000, has not commented on the claim.
Alpöge did not specify which version of Claude he used. Anthropic did not respond to requests for comment from Agence France-Presse or 20minutes.fr.
The Riemann hypothesis, formulated by Bernhard Riemann in 1859, concerns the distribution of prime numbers. It underpins a vast body of number theory and modern cryptography. A verified solution would trigger the Clay Institute’s $1 million award and likely reshape fields dependent on the behavior of primes.
Brian Conrey, a leading authority on the hypothesis and director of the American Institute of Mathematics, called the announcement “a monumental moment in the history of humanity” in comments to Agence France-Presse, though he added he had not yet examined the proof. “It’s the most important thing we’ve ever seen in mathematics,” Conrey said. “Some people aren’t going to agree, but it’s something major.”
Nestor Guillen, a professor at New York University, told AFP the result itself would not have immediate practical impact, “but the understanding it brings will have many, many applications in the coming months and years.”
Guillen contrasted Alpöge’s approach with the September announcement that an OpenAI model had solved the Navier-Stokes equations, another Millennium Problem, in under four days. The Riemann proof, Guillen said, is “the result of a researcher assisted by an AI model, and not the pure product of an artificial intelligence alone,” making it “more respectful of the method and scientific practices.”
Eliezer Yudkowsky, a researcher known for warning of existential risks from advanced AI, wrote on X that the Riemann hypothesis was “the challenge most often cited as a warning” about AI capabilities.
Hours after Alpöge’s announcement, Ravi Vakil, a mathematics professor at Stanford University, reported that a postdoctoral researcher at Stanford had independently submitted a separate proof of the hypothesis.
Neither the Stanford researcher nor Alpöge responded to AFP’s requests for comment on the competing claim.
Alpöge previously collaborated with Australian mathematician Tristan Buckmaster on the Navier-Stokes problem. When OpenAI announced its solution in September, Buckmaster posted publicly questioning similarities between the AI’s approach and his own work.
OpenAI said in early October that its models had solved 377 distinct mathematical problems.
The mathematical community now faces two unverified claims to the same prize problem. The Clay Institute’s formal review process, which typically requires publication in a refereed journal and a two-year waiting period, has not begun for either submission.