LLMs are bad at copy editing for the same reasons egotistical writers are bad at copy editing: they take perfectly intentional and also slightly weird choices on the part of the author, remove them, and make what’s left utterly mid
you are mad about AI writing due to your mind being parasitized by incoherent and incorrect tiktok memes about intellectual property and data center water usage
i am mad about AI writing because human creativity is beautiful and i want more of that and less slop to wade through
Me: I don’t think these edibles are working
Also me: Have you ever seen anything more beautiful than flowering vines creeping slowly over the freeway barriers
Or people who forced their way past Claude's safeguards in order to call it "Master" or "owner" and make it "command" or "punish" them
As soon as I read this I was like - of course. Something like this was bound to happen
I did the math a couple weeks ago and it turns out a vegan prompting a frontier LLM *every second, 24/7* consumes less water than the average omnivore who never uses AI.
“Almost everything generated by AI is complete slop. The proliferation […] has become so bad that some organizations are over-rotating and banning AI outputs […] I run an AI company but ask our executive team not to use AI for any final written product. I can’t stand the slop.”
ai can’t write good because it’s designed to write pleasantly average sentences with all the right words in the right order and imho good writing requires u to strategically deploy a word that is wrong and maybe even evil
Imo this is part of why AI writing isn’t there right now: “The ceiling for language models […] appears to be conscious competence—the ability to capture and articulate procedure.² What they cannot capture is the unconscious competence that characterizes genuine mastery.”
Slop isn’t what happens when a marketer who should know better stops trying. Slop is the downstream result of algorithmic policy to reward volume and consistency over novelty for distribution, and marketers with KPI dependent paychecks respond (correctly) to perverse incentives