Particle.news
Download on the App Store

Altman Acknowledges Mathematicians’ Concerns While OpenAI Says It Will Keep Publishing AI Proofs

The company’s decision to release claimed breakthroughs raises urgent questions about how results are checked, credited and rewarded.

Overview

  • Altman told the Cheeky Pint podcast on Thursday that he feels empathy for upset mathematicians but that OpenAI will continue to publish fundamental discoveries and that the company holds more unreleased results.
  • OpenAI has posted hundreds of AI-generated manuscripts and published a formal Navier–Stokes report that the company says required roughly $10 million of work, a claim that has drawn intense technical scrutiny.
  • Many mathematicians have criticized OpenAI’s rollout for poor attribution and thin explanations, and 28 Fields Medal winners published an open letter while OpenAI withdrew from a Caltech math hackathon.
  • Independent reviewers say verification will take months because OpenAI has not released its internal model, full prompts, or complete provenance even though it provided a small number of machine-checked formalizations.
  • Institutions including the Clay Mathematics Institute and the Institute for Advanced Study are assessing prize relevance and verification procedures and the dispute could reshape norms about credit, careers and access in mathematics.