4 ms·
Well, they're being trained with reinforcement learning now, so I, for one, am excited to see what kinds of unique extremely performant insanity they'll produce
by Philpax 2y ago
Well, they're being trained with reinforcement learning now, so I, for one, am excited to see what kinds of unique extremely performant insanity they'll produce in the future =D
- misrasaurabh1 2y agoWhat we discovered that LLMs aren't really the answer wrt performance, the whole system around verifying correct and performance is the key.