Five years ago, the top results for this question would have been from Quora and its ilk - chock full of uninformed idiots "answering" questions for clout or whatever; training an LLM on reams of that garbage just cuts out the middleman.
I'm not defending them - I hate'em too!
What I was saying was that for questions like these, the plagiarism machines were turned loose on a corpus that was _already_ garbage due to a weird business model - and the result is, if anything, even worse.
