Reactions from ranked influencers
3 posts
@binarybits Claude Code alone used over 50. It also uses regular expressions, conditionals, loop etc. It’s a completely different paradigm from the all the feed-forward single pass neural network.
Gary mustve been really able to come up with counterexample of JC, but no, he simply did not use enough APIs.
this is confused. it’s not LLMs that are good at math; it’s LLMs merged with (neuro)symbolic tools that are good at math. amazing that people feel free to clutch their pearls without understanding even the basic technical details. https://twitter.com/austen/status/2079237340451442918
Gary is onto something here. LLMs would be extremely valuable when it comes to math if only they could, in some magical way, have access to a “(neuro)symbolic tool” that could compute. If only there were some machine that could do a lot of computing really fast.
this is confused. it’s not LLMs that are good at math; it’s LLMs merged with (neuro)symbolic tools that are good at math. amazing that people feel free to clutch their pearls without understanding even the basic technical details. https://twitter.com/austen/status/2079237340451442918