I hadn't read the whole thing when I wrote my first comment.

Yesterday, someone tried to pass off a Haiku as AI generated. It was too good. I accused the human of editing to make it passable.

OpenAI can write a poem - thematically, it's not terrible (and I'll admit, it does as well on a villanelle as most of us poets do, but it's not great). The problem with OpenAI poetry is that the poor thing has absolutely no concept of meter. It is so-so on rhyme, but can avoid that simply by not attempting rhyming verse.

Nevertheless, one simple word-change and I knew the human had edited.

Unfortunately, it's good enough already to "pass" for 90% of readers who only skim. And factually? Most of what it generates is utter nonsense. But people will spread it, post the content on blogs, call it their own - and GOD do I hope those people are held liable for the misinformation they post, because it could be dangerous. The only people who see the warnings are the ones working with the generator - not the ones reading the crap they post.