
21 hours
Wait, did it prove the RH or disprove it? A proof would be impressive, though not necessarily as indicative of intelligence as Sam Altman would want you to think; a disproof could just be a counterexample, though, which could be generated by < 100 lines of code and enough compute. Given the amount of compute available to LLMs, that’s not really impressive at all. They could literally have just found the highest number anyone has confirmed doesn’t disprove the RH and then checked each higher number in a loop.
Trying to get computers to think: A+ idea, no notes.
Trying to get computers to think by destroying the environment, spying on everyone, stealing people’s work, destroying rare books, and making it so no one can buy a computer: I have a few criticisms.