Updated
Updated · OpenAI · Oct 26
OpenAI Unveils 3 AI-Generated Discoveries, Adopts Split Release Strategy
Updated
Updated · OpenAI · Oct 26

OpenAI Unveils 3 AI-Generated Discoveries, Adopts Split Release Strategy

3 articles · Updated · OpenAI · Oct 26

Summary

  • OpenAI said its models produced three new results in mathematics, computer science and physics, and it is releasing them under different community-led plans rather than a single disclosure model.
  • Three strategies reflect field-specific advice: the math proof is public now, the computer science result comes with an external collaboration and immediate public artifacts, and the physics materials are being held back while outside experts write them up.
  • In mathematics, OpenAI said its system proved the local smoothing conjecture in all dimensions—implying the Kakeya and Stein restriction conjectures—and posted both a natural-language writeup and a Lean formalization.
  • In computer science, OpenAI said its system disproved the Exponential Time Hypothesis, a widely used assumption in hardness results, with Russell Impagliazzo, Ramamohan Paturi and Ryan Williams leading open verification and OpenAI funding a conference.
  • In physics, a group led by Nobel laureate David Gross will shepherd claimed existence proofs for 4D Yang-Mills, QCD and a unified completion of the Standard Model, as OpenAI says it will forgo academic credit and back independent scrutiny with $5 million in broader research support.

Insights

Why did OpenAI release its math and ETH proofs publicly but withhold the QCD result while physicists prepare their own explanation?
If Lean verifies an AI proof of local smoothing, what will mathematicians still need before accepting a claimed all-dimensions breakthrough?
What would a real refutation of ETH change across algorithms and fine-grained complexity, and how can experts tell it is not a formalized loophole?