I am not making a comment about the usefulness of LLMs to manipulate text. If anything, my view is that to the extent that LLMs are seen as intelligent it's pre

I am not making a comment about the usefulness of LLMs to manipulate text. If anything, my view is that to the extent that LLMs are seen as intelligent it's precisely because they manipulate text in ways humans find useful, not because it follows from some philosophy about intelligence or rationality. Maybe I'm being too opinionated but I think we should stop dressing up explanations of large language models in misleading terminology. I'd prefer instead to talk about the actual technology and reason from there.