Higher-order effects of LLM slop
I don't much enjoy LLM prose, at least when it's presented as human-generated. I've written about that, and since that post I've started to notice myself adjusting in indirect ways. Even when I don't suspect that something is AI-generated, there's a kind of punchy prose that I just don't want to be reading as much, and I'm doing more to avoid it. This means even less social media, and browsing different parts of newspapers and other news sources.
There's a sad side to this: there are any number of wonderful human writers I just don't enjoy quite as much. I've been a Jenny Holzer fan for well over 20 years now, but it's been a long time since I thought she worked in an under-explored niche. It makes me think of Impressionism after the rise of the camera. Either the technology is making some of the goals of the human work less relevant, or it's saturating some of our aesthetic receptors, or we want more diversity in what we consume, or some combination of those and other things.
The happy side is, of course, that tons of human writers appeal to me even more. I keep Wordsworth's Prelude bookmarked in my Kindle account so that I can read it as a sort of palate-cleanser. I'm now doing this more than once a day, and sometimes I feel quite desperate for it. To return to the Impressionism example, I bet that lots of humans start to write in distinctively non-AI ways. Many of us are already cutting back on em-dashes and certain kinds of contrastive devices, and this will happen at deeper levels also.
I'm unusual in this regard. I'm probably both unusually sensitive and unusually averse to LLM slop, and I definitely spend more time close to AI than the average person. (This is a good thing! I love being a programmer.) But I think my experience will become more normal: AI will keep spreading, and at least some people will revise habits for roughly the reasons I've laid out. I'll be eager to observe, and perhaps adopt some of, what they do.