I think it's important to understand the humans _can_ do what LLMs do: predict next tokens from prior ones. But LLMs are only operating on text and humans are

I think it's important to understand the humans _can_ do what LLMs do: predict next tokens from prior ones. But LLMs are only operating on text and humans are only operating on <waves hands>