Is this really a bad thing, or a problem at all? Some researchers found that increased scale resulted in increased utility. LLMs aren't generalized AI, but they

Is this really a bad thing, or a problem at all? Some researchers found that increased scale resulted in increased utility. LLMs aren't generalized AI, but they seem to be good at some stuff. Their must intrinsically be some limits to this approach. But that doesn't mean it is a problem or that LLMs cant serve some useful role. Just consider for a moment if we could train an LLM on the combined experiences of all humans that have ever lived. No one is going to suggest that we must go bigger.