Not that they're not, but I've been seeing automated blogspam for several years now. I'm Greek, and it often happens that when I'm searching for some specialised subject on the Greek-language web, I will sometimes stumble on pages that are clearly automatically translated aggregations of blog posts in some original language other than Greek (which is usually English, but not always).

So it's unclear to me, at this moment, how much the automated spam threat will increase because of GPT-4 and friends. We've already had plenty of that before we had publicly available LLMs.

Anyway I'm not convinced yet of the coming chatpocalypse. I think I'll just wait this one out and not try to make any predictions when there's so little to go on.

Edit: I'm more worried about scammers, scamming, than spammers, spamming, to be honest. There are plenty of people that don't have a clear understanding of how difficult it is to tell that one is having a conversation with a bot over the internet these days. People were already falling for bride scams over the 'net for a long time before LLMs. A scammer could automate that. And probably increase their revenue to boot (because they can target more people now).

Bride scam:

https://en.wikipedia.org/wiki/Bride_scam

This is massively overstated. The major costs of being a spammer are creating new accounts and staying off blacklists. Carrying on conversations or creating content are marginal costs, at best.

LLMs are no more valuable to a large spam operation than they are to any other business. This is like a 5%-10% profit increase, not 50%-100%. The key to that game is finding suckers. Language models help, but only so much.