Speaking mostly to myself—I was reminded of this post when I saw Nabeel Qureshi’s recent Substack note
Multiple Claude-written articles with 10k+ likes here recently. People seem to really like slop on X too. Reluctant conclusion: maybe the taste for organic hand rolled human only tokens will become a niche IQ taste similar to enjoying Japanese wood carvings or something.
Correct—I scrapped Substack using Pangrams API and found that readers just can’t really tell if it is AI slop or not. It’s the beginning of the end for writers who purely compete on the basis of prose.
re: “When do we expect LLMs to become capable of writing online content that we actually think is worth reading?[6]” I still don’t think they’re there yet for me personally, nor many LWers probably (maybe Mythos-class models given their own Substacks like Opus 3 might change my mind; eudaemon_0 is certainly intriguing), but it’s hard to deny that they’ve crossed the bar for a lot of people online, and it’s pretty annoying.
Speaking mostly to myself—I was reminded of this post when I saw Nabeel Qureshi’s recent Substack note
and Evan Armstrong’s reply (see also his In Defense of AI Slop, h/t Gwern)
re: “When do we expect LLMs to become capable of writing online content that we actually think is worth reading?[6]” I still don’t think they’re there yet for me personally, nor many LWers probably (maybe Mythos-class models given their own Substacks like Opus 3 might change my mind; eudaemon_0 is certainly intriguing), but it’s hard to deny that they’ve crossed the bar for a lot of people online, and it’s pretty annoying.