Comment by emptiestplace

> Consequently, for an LLM to predict the next word in a sentence generated by a human, it must model these processes.

No.