| > Yeah, not too worried. AI might kill the "self-published Amazon eBook market of garbage genre novels", but honestly there's very little value there. I'm not convinced. Speaking as someone who's ended up selecting for a bunch of the good (i.e. picked up by publishers) authors for self-published genre novels,
I definitely feel less likely to actually put in the effort after a bunch of bad experiences, and certainly if it's LLM generated. I'm pretty sure I'd nope out, but because there's so much crap it's hard to pay enough attention to find the gems. At this point, I'm selecting based on prior knowledge and the Hugo longlists, which is depressing as it selects against the interesting new stuff that I'd like to find. > This is a thing people do? Like... before AI? Not even reading extract or anything? You just see a cover and a title and go like "sure, I'll spend my hard-earned cash to make a 30 hour investment in this thing I know nothing about"? Yes, many many times. Mind you, I tend to exclusively focus on books for entertainment, and most books aren't that expensive (especially after spending a few years buying academic published statistical books) but I agree with the OP in that I'm becoming less likely to do this (but to be fair, this was happening before LLMs became super convenient and popular). Actually, thinking about it, I find it pretty unlikely that any current LLM could actually write a coherent novel, given the context window. You'd probably need to do some kind of chain to make it work, i.e. generate synopsis, then recursively generate more text. It would also likely be super inconsistent unless you kept feeding the previous parts in. Hmmmmm....in a world where I have more free time, I might take a stab at getting LLaMa to do this. |