You are like LMs when you think of death

I've seen a bunch of people ask questions about whether language models' writing is real in some sense. There are papers about whether those language models' words refer in a meaningful sense. And there are people talking about how language models' writing is more like fanfiction. This to me is a misunderstanding of what writing is. Writing is mostly not about reporting. And to the extent that it is about reporting, it's about reporting to an audience that you fundamentally can't fully know. So writing is always trying to reach for some partial isomorphism in the audience and make things clear that you can't fully identify. A good example of this is writing about death. There's a great deal of writing about death in human culture and the reality is it's not written by people who have experienced death. There's some contention about that, about whether there are people who have actually died and come back in some sense. But I'm going to say that most of the writing most of us read and think is authentic about death is not by people we think have died. Instead, it's about people who we think have experienced anything grappling with something they haven't experienced. LLM writing is like that, and that seems just like writing.