OpenAI’s AI Proved Half a Millennium Problem. Experts Agree

Human Verified 0

Share to: , and

Golden swirling curves and glowing particles on a dark background

A committee of physicists led by Nobel laureate David Gross has validated an AI-generated proof that quantum Yang-Mills theory exists — the first half of one of the seven Millennium Prize Problems. The group, convened at UC Santa Barbara’s Kavli Institute for Theoretical Physics, published its verdict on 9 November and released the proof’s machine-checked code the next day. The second half of the problem, the mass gap, remains unsolved.

What the committee confirmed

The committee — Michael Douglas, Gross, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov — says the proof establishes the existence of pure Yang-Mills theory under the problem statements standard in physics and mathematics. OpenAI submitted the full Lean formalisation on 28 October, and the committee published that code and an assessment of its checks on 10 November.

Lean is a programming language that checks mathematical work step by step, so a proof written in it can be verified mechanically. OpenAI announced in late October that its models had produced the existence proof on 9 September, with the formalisation completed on 16 October. The proof runs to 1,147 pages — too long for any human to read through.

A physicist's desk with old books, a brass compass and a glowing glass sphere

The verdict covers only the existence half. Yang-Mills theory describes the strong nuclear force, and the Millennium Prize problem asks for two things: a proof that the theory exists rigorously, and a proof that it has a mass gap — a lowest energy for its particles. The mass gap remains open, so the $1 million prize is not in play.

Our Read

Read plainly, this is the strongest result yet for AI as a mathematician — and the committee’s words keep it in proportion. The physicists used OpenAI’s Astra model to digest the Lean code, and wrote: ‘We do not claim to fully understand the proof yet.’ A machine-checked proof is not an understood one, and the understanding is the community’s job.

The process also drew criticism. OpenAI approached Gross, who chose the other four members, and the committee’s approach runs against transparency recommendations from an OpenAI-affiliated advisory group. Some mathematicians call it gatekeeping, since the full 1,147-page proof has not been released — only the Lean code and the assessment.

A rival team is working on the same problem in the open. Led by a Northwestern mathematician and an Anthropic researcher, it has published its workflow and is preparing a human-readable paper for wider scrutiny. If the committee’s verdict holds up under that scrutiny, it will stand as one of the hardest results an AI system has produced.

Why It Matters

It matters because the Standard Model is the most tested theory in physics, yet its foundations have never been proved. Douglas said: no rigorous definition of the theory existed, and its properties were out of reach of mathematics. If this proof survives scrutiny, the gap closes — and the tool that closed it was an AI model.

The signal is for AI research. The same week, OpenAI warned its models are getting harder to monitor, while DeepMind’s AlphaGenome 2 learned to read DNA cell by cell. AI is producing results at the frontier of mathematics and biology — and the question is no longer whether models can do science, but how science checks their work.

Background

Yang-Mills existence and mass gap is one of the Millennium Prize Problems set by the Clay Mathematics Institute in 2000, each carrying a $1 million prize. The prize requires both halves solved; this proof addresses the existence half only. The Clay Institute has not commented on the committee’s verdict.

Source: the KITP committee’s statement and Scientific American’s report.

Share to: , and

Faheem R. Khattak

Senior automation engineer and AI marketing journo covering LLMs, algorithms, analytics, core updates, Bing search, and technical SEO.

0 Comments

Leave a Reply

Optional Take Tag:

🛡️ Spam protection: Honeypot + Time-Trap Active

This site uses Akismet to reduce spam. Learn how your comment data is processed.