Experts say verification would mark watershed moment for math and AI
A researcher at AI company Anthropic says the firm's chatbot Claude helped crack one of mathematics' greatest unsolved problems: the Riemann Hypothesis, which has resisted proof for more than 160 years. The claim has not yet been independently verified.
Levent Alpoge, an Anthropic researcher and mathematician, published a four-page document on his personal website Friday (local time) stating that he "believes" he and Claude have found a proof of the Riemann Hypothesis, according to AFP and the Financial Times.
The Riemann Hypothesis was first posed in 1859 by German mathematician Bernhard Riemann. Despite efforts by countless mathematicians over 167 years, it has never been formally proved.
Alpoge released proof materials written in Lean, a programming language used to verify mathematical arguments by computer. He said the work resolves not only the Riemann Hypothesis but also its broader extension, the Generalized Riemann Hypothesis.
The claim has not yet been subjected to rigorous peer review. Even so, mathematicians say that if the proof holds up, it would represent a historic achievement.
Brian Conrey, a leading authority on Riemann Hypothesis research, told AFP the development would be "a monumental moment in human history" and "the biggest thing that has happened in the history of mathematics." He noted, however, that he had not yet seen the paper itself and had only read what Alpoge posted on his website.
Kit Yates, a professor at the University of Bath in the United Kingdom, said the proof "will be subject to intense scrutiny" and that if it passes, it "will have enormous ramifications for both our understanding of mathematics and the capabilities of AI."
A central point of interest in the announcement is the role AI played in the proof. Alpoge said his collaboration with Claude produced more than 250,000 pages of work, and that in the later stages of the research, the AI surpassed his own understanding.
"Near the finish line, Claude flew and danced and sparkled far ahead of me," Alpoge said, adding that he found himself in the unusual position of "putting a complete argument into the world without fully understanding the surrounding mathematical background."
Alpoge said he announced the findings after consulting with Anthropic CEO Dario Amodei. He also acknowledged that the timing — coming the day after Anthropic's record-setting initial public offering on US stock markets — could invite skepticism.
AI companies have recently been locked in a race to solve long-standing mathematical problems. OpenAI announced in September that its AI model had developed a solution to the Navier-Stokes problem, another of the Millennium Prize Problems.
Alpoge, who had been conducting related research at the time, raised the possibility that OpenAI had used his data, a claim OpenAI CEO Sam Altman rebuffed as baseless plagiarism. Hundreds of mathematical papers OpenAI released last month also drew mixed assessments from experts, who said the results contained a blend of meaningful findings and errors.
Nestor Guillen, a visiting professor of mathematics at New York University, said he viewed it as a positive sign that Alpoge openly admitted he did not fully understand the proof, and expressed the view that AI can function as a tool to support research rather than a replacement for human researchers.
The Riemann Hypothesis plays a central role in explaining how prime numbers are distributed among the natural numbers. At its core, the hypothesis holds that all nontrivial zeros of the Riemann zeta function lie on a line in the complex plane where the real part equals one-half.
The Clay Mathematics Institute designated the Riemann Hypothesis as one of seven Millennium Prize Problems in 2000, offering $1 million for each solution. To be officially recognized, a proof must be published in a peer-reviewed journal and then withstand at least two years of scrutiny and broad acceptance within the mathematical community.
quq@heraldcorp.com