Some one recently told me that their definition of intelligence was data-efficient extrapolation & I think that definition is pretty good as it separates intelligence from knowledge, sentience, & sapience.
Cool. I don't know how data-efficient current LLMs are, in the sense that I don't know what they know and don't know. It certainly seems like in some domains, they can can extrapolate, but I don't know what's in their training data.
