The company behind ChatGPT said it had cracked the Navier-Stokes problem, one of 7 Millennium Prize Problems published by the Clay Mathematics Institute to highlight some of the biggest unsolved puzzles in the field.

you are viewing a single comment's thread
view the rest of the comments
[–] 58 points 2 months ago (3 children)

No, OpenAI heard about a pair of researchers who were getting close to solving the problem and spent $22M on compute power to scoop them.

https://techcrunch.com/2026/09/08/openai-fought-dirty-on-career-making-math-problem-says-nyu-mathematician/

Even more concerningly, the researchers had been using OpenAI's codex model in their earlier work, meaning that OpenAI's model may have literally directly trained from and ripped off their work.

  • source
  • hideshow 3 child comments
  • [–] 33 points 2 months ago (2 children)

    Even more concerning they said they’d credit one of the authors if he wrote out a different author he collaborated with that worked at Anthropic, and refused to say they didn’t train the model on his active research.

    If you’re innocent you don’t do that.

  • source
  • parent
  • hideshow 2 child comments