< Back to all clusters
[TECHNOLOGY] · United States · 17 sources

started · updated

OpenAI releases hundreds of AI-generated mathematical manuscripts

OpenAI has released a massive repository of 719 to 722 mathematical manuscripts generated by an unreleased, internal frontier AI model. The collection is organized into 372 families of results covering diverse fields such as algebraic geometry, number theory, and theoretical computer science.

Among the claimed breakthroughs are advancements related to the Riemann hypothesis, specifically a demonstration of the 'quasi-Riemann hypothesis', and the Birch–Swinnerton-Dyer conjecture. While the scale of the release is unprecedented, it has sparked significant controversy within the scientific community. Many mathematicians have expressed concern over the lack of transparency, as the underlying model remains proprietary and inaccessible for independent verification.

Critics, including the Association for Human Mathematics and prominent mathematicians like Terence Tao, have criticized the method of release. They argue that the simultaneous dumping of hundreds of unverified papers bypasses traditional peer-review processes and threatens the collaborative nature of mathematical research. Some researchers have described the event as a 'mathocalypse' or a 'demonstration of power' rather than a scientific contribution.

OpenAI noted that approximately 42% of the top-line results have been formalized using the Lean programming language to allow for computer-assisted logical verification. However, the company has already had to retract or revise several manuscripts due to errors, such as sign mistakes in algebraic geometry proofs.

Entities

Alex Kontorovich · Association for Human Mathematics · Clay Mathematics Institute · GitHub · Institute for Advanced Study · Martin Bridson · OpenAI · Princeton · Sam Altman · Terence Tao

Claims

What the coverage asserts, and how many sources carry each claim.

Sources