Some thoughts from my recent “LLM diary” (https://chrisdone.com/posts/llms/):
I’ve been recently thinking that the whole phrasing around interacting with LLMs has become increasingly anthropomorphic [in my circles], and thought of a clever mental hack to avoid it: instead of “talking to Claude,” say “talking to the computer,” and instead of “AI generated” or “LLM generated,” simply say “computer generated.” I find it thoroughly cuts through the illusion that we are doing anything more. But YMMV.
