Anthropogenic Staffer Claims Solution to Mathematical Riemann Hypothesis

Stay informed with free updates

An Anthropic employee has claimed to have solved one of the hardest and oldest math problems, in what would be the most dramatic advance yet in AI’s transformation of the field.

Levent Alpöge said he worked with Anthropic’s Claude model to develop a proof of the Riemann hypothesis, a riddle about prime numbers whose “shadows have guided mathematicians for 150 years.”

The purported proof, which has yet to be independently verified, would be extraordinary, even compared to a recent spate of AI-assisted mathematical advances. Alpöge, himself a mathematician, presented it as a landmark moment in an article published on his website on Friday.

“It is natural to think that this is an end, in my case a loss of a main thread,” he writes. “But what I have seen in the course of progress has amazed me so much that I believe this is the dawn in science of something much greater.”

Mark Chaplain, president of the London Mathematical Society, said the announcement was a “historic development”.

“This opens the floodgates to new levels of human understanding of mathematics and its applications,” he said.

Other experts said it would take time to assess the evidence, but if confirmed the implications would be profound.

“An announcement of this magnitude will come under scrutiny,” said Kit Yates, professor of mathematical biology and public engagement at the University of Bath. “Nevertheless, if Alpöge’s argument survives this scrutiny, the consequences would be enormous, both for mathematics and for our understanding of what AI can do.”

The hypothesis has baffled mathematicians since it was formulated by German GFB Riemann in 1859. It is a central and famous problem in number theory, which concerns the properties of integers.

Alpöge has yet to reveal a formal account of its claimed solution with a prose explanation. Instead, he published his results in Lean, a programming language specializing in verifying and expressing mathematical proofs.

He said his AI-powered work exceeded 250,000 pages, and he concluded that it would “obviously (be) inappropriate to attempt to publish a massive, unreadable parody of ‘slop’.”

He credited Claude as the main driving force behind the final push toward a proof.

“Claude and I have had a great time together over the past few months, but this time, Claude far beat me to the finish line,” he wrote. “No, Claude flew and danced and sparkled near that finish line.”

He noted that he now found himself in a “rather extraordinary situation that this age has made possible.”

“I present to the world a complete argument before I have mastered a vast neighborhood around its mathematics,” he writes.

The Riemann hypothesis is one of the problems posed 26 years ago as a Millennium Prize challenge with a million dollar reward for each.

OpenAI, the creator of ChatGPT, announced in September that it had developed a solution for another Millennium Prize topic known as the Navier-Stokes problem. Alpöge had worked on the Navier-Stokes challenge with a mathematician from New York University, Tristan Buckmaster. The duo claimed that data from their work could have informed OpenAI, since they used the company’s tools. Sam Altman, chief executive of OpenAI, rejected what he called “unfounded accusations of plagiarism.”

Advances in AI have stunned mathematicians and sparked fear, criticism and fears about the future of human activity in this area. OpenAI sparked a row last month by unveiling hundreds of papers, including what it claimed was evidence of some long-standing problems.

Experts said the hoard mixed impressive works with others that contained errors or were illegible.

The Association for Human Mathematics, a group committed to protecting the field from the “threat of artificial intelligence,” urged mathematicians to stop working with OpenAI. The company defended its actions and pledged $5 million to a math-focused foundation.

Gn bussni

Logo