Anthropic's Unknown Model Makes Major Breakthrough on the Riemann Hypothesis

One of mathematics' most complex and long-standing unsolved puzzles, the Riemann hypothesis has seen a major scientific breakthrough. According to ixbt.com, Anthropic's unreleased artificial intelligence model has made unexpected progress toward solving this difficult problem, significantly raising the lower bound for solutions in which the hypothesis holds. Techcrunch.com reports .

A $1 million prize is currently offered to anyone who finds a general proof of the Riemann hypothesis, which concerns the distribution of prime numbers, and it is still waiting for a winner. Although modern neural networks have not yet fully solved the problem, their ability to uncover new ideas in science and mathematics has intensified the debate.

Artificial Intelligence's Independent Scientific Research

The most astonishing aspect of this achievement is how it was obtained. An Anthropic employee with no advanced mathematical training tasked the model with making a serious attempt to prove the hypothesis. The system then worked independently for a day and a half, coordinating the entire process on its own.

During this period, the model tested a total of 650 different ideas for solving the problem. Sixty sub-agents worked in coordination with one another. According to the appendix to the scientific paper, two of these agents developed the main mathematical ideas, 13 contributed to the ideas, 30 attempted to find new solutions, 13 served as validators checking the correctness of the arguments, and the remaining two helped write the initial scientific paper.

Debate and Concerns in the Scientific Community

The results were confirmed by Anthropic's in-house mathematicians and formalized using the open-source Lean proof assistant. This is one of the latest examples of mathematical advances achieved with the involvement of large language models (LLMs). The fact that artificial intelligence models solved several Erdős problems this year is another clear example.

This pace of progress is causing both amazement and serious concern in mathematics. In a public declaration signed in June, prominent mathematicians expressed concern that artificial intelligence could undermine the core values of science. In particular, they emphasized that genuine mathematical proofs should be attributable to their authors, who must personally take responsibility for their correctness.

Nevertheless, experts remain divided over how new research methods should be approached. In a blog post responding to the published statement, Fields Medalist Timothy Gowers questioned whether artificial intelligence's impact could transform mathematics in a more complex and positive direction.