LLMs generate continuations. Those continuations are built from the text they are trained on. I see GP suggesting that because humans!= LLMs that "doing word p

LLMs generate continuations. Those continuations are built from the text they are trained on. I see GP suggesting that because humans!= LLMs that "doing word prediction" is not an exhaustive list of human behavior.