3 ms·
> LLM doesn't learn incrementally from previous encounters This. Lack of any way to incorporate previous experience seems like the main problem. Humans are oft
by tliltocatl 1y ago
> LLM doesn't learn incrementally from previous encounters
This. Lack of any way to incorporate previous experience seems like the main problem. Humans are often confidently wrong as well - and avoiding being confidently wrong is actually something one must learn rather than an innate capability. But humans wouldn't repeat same mistake indefinitely.
- ACCount37 1y agoYou can gather feedback from inference and funnel that back into model training. It's just very, very hard to do that without shooting yourself in the foot. The feedback you get is incredibly entangled, and disentangling it to get at the signals that would be beneficial for training is nowhere near a solved task. Even OpenAI has managed to fuck up there - by accidentally training 4o to be a fully bootlickmaxxed synthetic sycophant. Then they struggled to fix that for a while, and only made good progress at that with GPT-5.