OpenAI claims scientific advances, seeks independent review

OpenAI says its AI models have made three major breakthroughs in mathematics, computer science and theoretical physics. The claims include disproving the Exponential Time Hypothesis (ETH), a long-standing assumption used to assess computational difficulty. The findings require independent examination before their significance can be established.

The company says its model produced the ETH disproof on October 16 and formalised it using Lean the following day. Lean checks whether proofs follow formal mathematical rules, but verification still requires researchers to examine exactly what has been established. OpenAI has published the explanation and formal proof on GitHub, with computer scientists Russell Impagliazzo, Ramamohan Paturi and Ryan Williams leading a group reviewing the result.

In mathematics, OpenAI claims its models proved the local smoothing conjecture in every dimension, which it says would also resolve the Kakeya conjecture and Stein’s restriction conjecture in all dimensions. In physics, the company reports existence proofs for certain quantum field theories, including Quantum Chromodynamics, alongside a result concerning pure Yang-Mills theory.

OpenAI says external experts are organising groups to assess the findings and explain their implications. It has also pledged funding for conferences and research programmes to support scientific scrutiny.