I think of language as a way to determining the generation of mental representations, which are causally interactive code. LLMs demonstrate the ability to interact with reality, eg write meaningful computer programs, because the text tokens they analyze are not meaningless.

→ View original post on X — @plinz,