4 ms·

I think it's actually the opposite. Intuition is the only thing LLMs can do, which is why they are prone to hallucinating when they can't validate their intuiti

I think it's actually the opposite. Intuition is the only thing LLMs can do, which is why they are prone to hallucinating when they can't validate their intuition against reality. There is an idea that human intuition, expertise, and critical thinking are largely pattern recognition. When you encounter a situation, your brain gives you a plausible starting point, based on what it has experienced before. You then continue with explicit reasoning, which is slow and inefficient, and try to validate your ideas. The more relevant the patterns you have learned are to the situation, the more likely you reach a useful conclusion. LLMs are largely the same, except that they cannot learn from experience in normal usage. And except that they experience the world only through symbolic data, while the human brain has access to plenty of sensory data.

pegasus 13d ago
Kind of. First of all, reasoning LLMs can also do (a form of) reasoning. Second, yes you could say LLMs form a sort of intuition, but its domain is the space of human-produced text. It only translates to real world intuition to the degree that those intuitions make their way into the corpus the model has been trained on. You could say that, when it comes to anything other than textual prediction, their intuition is secondary, a reflection of a reflection, so it will always lag behind that of humans.