4 ms·
I think serious math research progress should come in 1-2 years. It basically only depends on how hard informal verification is, because training data should be
by Davidzheng 11mo ago
I think serious math research progress should come in 1-2 years. It basically only depends on how hard informal verification is, because training data should be not a problem and if informal verification is easy you can throw RL compute at it until it improves.
- trenchgun 11mo agoLLMs are already a powerful tool for serious math researchers, just not at the level of "fire and forget", where they would completely replace mathematicians.