how (and why!) to make slop
slop doesn't happen in a vacuum.
[pause for a moment and take a look at that sentence; that's a great sentence]
slop whether llmish or otherwise is a structural issue. the count of monte christo didn't need to be a million pages long. in fact it very much improves with a severe redaction. but dumas was getting paid on length and publishing in serial. ergo--length.
you can blame the readers all you want for going to the trough and lapping this stuff up but if your hopes of doing some good in the world depends on a new kind of person who prefers kale to mcdonalds boy howdy do i have some bad news for you.
llm slop can only really exist in the shortform video space and linkedin slash facebook and there's a structural reason for that (and a technological one). facebook has an audience of pretrained idiots who can't tell reality from nonreality and an algorithm that exclusively exists to get you to consume more algorithm by making you (primarily) angry (it's very unpleasant!). linkedin has an audience of folks who use text mostly as a business display (like a mating display but instead of mating there's m&a) and frankly already wrote like llms before llms existed. shortform video is built to dripfeed dopamine so who care is the video is real or not--it doesn't have to be right! it just has to be short! (i'm pretty sure this is a john mulaney quote?)
ai generated content doesn't really work outside of these domains because the other domains don't fit structurally or monitarily. when there's no algorithmic and monetary lubrication it doesn't work because there's no audience and/or no compensation. also because an ai podcast is stupid?