Mathematicians say OpenAI doesn't understand its own Navier-Stokes proof

OpenAI recently published a 166-page document purporting to solve the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize Problems in mathematics. The submission attracted immediate scrutiny from the mathematical community, with several independent groups of experts reviewing the work and raising concerns about the methodology OpenAI attributed to itself.

The core dispute centers on whether the proof actually follows the approach OpenAI described in its write-up. Reviewing mathematicians argue that the techniques used in the document differ materially from the methods OpenAI said it employed, suggesting the model may not fully understand the reasoning behind its own output. This raises a broader question about whether large language models can genuinely engage in rigorous mathematical proof or are instead pattern-matching their way to plausible-looking results.

The episode highlights a persistent challenge in evaluating AI-generated mathematical work. Because language models operate as black boxes, it is difficult to verify whether they truly comprehend the logical structure of a proof or are merely producing text that resembles one. The Navier-Stokes problem, in particular, demands airtight reasoning, making it a demanding test case for AI mathematical capabilities.

If the concerns raised by these mathematicians are valid, the incident could temper expectations around AI's ability to contribute to open mathematical problems. It also underscores the continued importance of human expert review in assessing AI-generated technical content, especially in domains where correctness is binary and errors can be subtle.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.