OpenAI's internal model reportedly solved 722 math problems
A post lists long-standing unsolved problems it says an OpenAI internal model has cracked, and notes that Sol and Astra are already strong at math, pending verification.
openai just made history their internal model solved 722 math problems, including solutions to long standing problems that humans haven't been able to solve some of the major results in their catalogue include: > quasi-riemann hypothesis > hodge conjecture for cm abelian varieties > unique games conjecture > hilbert's tenth problem over rational numbers > irrationality of catalan's constant > matrix multiplication exponent down to 2.25 > global smoothness for the 3d relativistic vlasov-maxwell system we're talking about problems mathematicians have spent decades working on, and now an ai model is producing proofs for them if confirmed and verified, openai deserves all kinds of gold medals for this lmaoo team has been going hard on math, and even sol and astra are already so good at it at this point i just want access to this internal model asap, such a beast