An unreleased Anthropic model made progress on one of math's biggest unsolved problems | TechCrunch

An unreleased artificial intelligence model developed by Anthropic has achieved a notable advancement in addressing the Riemann hypothesis, a famous and long-unresolved mathematical problem concerning prime numbers that carries a million-dollar bounty.

Triggered by a simple prompt from a staff member who lacked advanced mathematical training, the system autonomously organized and executed its research over a period of thirty-six hours. The system deployed a complex network of sixty specialized sub-agents to coordinate tasks, evaluate hundreds of distinct concepts, and process tens of millions of tokens, ultimately expanding the proven range of the hypothesis. Company mathematicians later verified the correctness of the achievement using Lean, an open-source computer-assisted proof tool.

This achievement is part of a broader trend where large language models are increasingly used to solve complex theoretical mathematics, following several other high-profile mathematical conjectures resolved by AI systems earlier this year. While these developments have generated significant excitement, they have also sparked deep philosophical debates within the global academic community. Some scholars express concern over the potential dilution of traditional academic authorship and personal accountability when automated systems generate proofs. Conversely, other prominent mathematicians suggest that shifting away from human-centric discovery could usher in a highly productive new era for scientific exploration where individual attribution is no longer the primary focus of mathematical progress.