OpenAI math blitz sparks backlash from researchers — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI released 722 manuscripts covering 372 result families from an unreleased frontier model, with the "average result" using about three hours of ChatGPT Pro thinking compute, according to the company's claims.
- OpenAI announced it solved the Navier-Stokes Millennium Prize Problem using an internal model more powerful than GPT-6 Astra alongside 10,000 concurrent agents — after reportedly racing to beat competing researchers to the result.
- OpenAI convened AGMAI, an independent advisory panel of elite mathematicians, to coordinate future releases, though panel members told The Verge the process has been "messy and confusing" and bears "hallmarks of OpenAI's previous rushed forays."
- Mathematician Andreas Thom publicly accused OpenAI of potentially training on unpublished research, noting the company acknowledged building on his and Gábor Kun's prior work in non-sofic groups.
- Fields Medalist James Maynard (University of Oxford) told The Verge he has spent the past year "soul searching" about mathematics' future as AI reshapes the traditionally slow-moving discipline.
- Mathematician Abhishek Saha (Queen Mary University of London) said OpenAI's conduct around the Millennium Prize push amounted to "the kind of things that mathematicians will generally not do."
Why it matters: OpenAI's 722-paper release and claimed $1 million Millennium Prize solution have triggered an unusual backlash in a field not accustomed to public disputes — with one Fields Medalist telling The Verge he is "soul searching" about his discipline's future and advisory panel members describing coordination as "messy." If OpenAI cannot resolve the credibility gap with mathematicians, future AI-generated results risk being dismissed rather than celebrated, eroding the company's ability to claim historic breakthroughs.
Ask SkimNews
