On November 20, two proposed solutions for the Riemann hypothesis emerged, sparking a debate on human versus machine intelligence. Levent Alpöge used AI model Claude to find a proof, while Alessandro Masullo released a human-written manuscript. Masullo wrote, "We are still here. We are very much alive." Both remain unverified.
'We are still here. We are very much alive. We are not going anywhere,' mathematician Alessandro Masullo wrote in his paper detailing his efforts to crack mathematics's greatest unsolved problem
In the space of a few hours on November 20, one of mathematics's greatest unsolved problems has been met with two proposed solutions -- one led by a machine and the other by a human -- setting up a referendum of sorts on the place of human insights in a field increasingly dominated by artificial intelligence (AI).
The Turkish-American mathematician Levent Alpöge at AI company Anthropic announced he and the company's AI model Claude had found a proof of the Riemann hypothesis. A short time later, American Mathematical Society president Ravi Vakil posted a rival manuscript by his Stanford University colleague Alessandro Maria Masullo with a striking note: "It is all human, no AI."
The Riemann hypothesis, proposed in 1859, concerns a hidden order in the distribution of prime numbers. It predicts the inputs for which a mathematical function called the zeta function equals zero. While Alpöge published computer code totalling more than 1.7 crore lines purporting to show the proof is correct, he did not publish a conventional paper, although he said he had a draft prepared by Claude of over 2.5 lakh pages.
In his manuscript, of 940 pages, Masullo has claimed he has proved a broader formulation of the Riemann hypothesis that overlaps with the work of Alpöge and Claude.
However, the two efforts' origins could hardly be different. Alpöge has said he guided Claude through weeks of investigation before the model surpassed him near the finish line. On the other hand, Masullo has said his work began 12 years ago, as a graduate student under Brian Conrad at Stanford. According to Vakil, Conrad has already spent many months checking Masullo's proof and suggesting revisions and that the duo has also discussed Masullo's approach with Peter Sarnak at the Institute of Advanced Studies.
They had planned to spend another year polishing the paper and seeking more inputs before publishing it. According to Conrad, Alpöge's announcement prompted them to release the unfinished draft.
Both proofs are far from being verified at this time. While independent experts have said the code accompanying the Alpöge-Claude effort compiles without errors, the proof itself has yet to be checked and -- crucially -- understood. Masullo, on the other hand, has already acknowledged mistakes in his manuscript. Hours after releasing it, Vakil also updated his post to say several mathematicians have reported problems, many identified with the help of AI models. It is also not clear whether these mistakes are trivial or significant.
But to sharpen the contest between human and machine, Masullo wrote in his manuscript's acknowledgments: "If any message at all this work is meant to send to the mathematician who feels discouraged by the pace and tone of the current climate, it is this: we are still here. We are very much alive. We are not going anywhere."
