I feel like we can get some content out of an AI focused thread.
Yesterday OpenAI announced they'd solved a millennium math problem.
Quick context, 26 years ago, the Clay Mathematics Institute announced a million dollar prize if you solve one of 7 unsolved math problems. Very old, very difficult problems, obviously. Only one has been solved (interesting story, solved by a reclusive mathematician who declined the money and a fields medal and disappeared).
Anywho, OpenAI got wind that someone was on the verge of solving the Navier-Stokes equation (fluid dynamics) with the help of their frontier model. OpenAI researchers then spent millions in compute running 10,000 models running an unreleased GPT build. After 88 hours, it solved it. Except...
The NYU professor that was on the verge of cracking it was using GPT as an assist and there are questions as to whether openAI knowingly or unknowingly used some of his work or if his work unwittingly helped train the model that solved it.
OpenAI obviously doesn't care about the money. They blew through the prize money in compute costs. And they've already gotten the publicity.
XI Wiki




