Who knew you could win gold in the International Math Olympiad without truly reasoning?
Also I learned from a16z that current AI is different because it's "probabilistic". Who knew probabilism was the secret ingredient that could get gold in IMO?
Ok here’s my real take
Liron Shapira
Liron Shapira19.7. klo 19.15
Current hypothesis: Yeah LLMs are missing a key aspect of brainlike intelligence (learning from small input & robustly fixing their mistakes, or something)… And using LLMs to power chain-of-thought reasoning is a fragile hack… But fragile hacks will get to bootstrapping RSI.
34,95K