Ah, ok, 27% -> 19%, an 8% YoY drop, and the numbers were obtained using identical methodologies across multiple years. Yeah, I agree, seems to check out.
Still, there are definitely a shitload more bots than there were before.
Are there more bots or does it just seem that way because there are fewer people? I don't ask to be snarky, I'm legitimately curious what the numbers were and are.
If people set up their own instances with 100% LLM bots how would we even know they are bots to defederate? The AIs are too realistic. Only giveaway would be the ones dumb enough to make accounts post/comment 24/7 with seconds between each comment.
Right now, it's still possible to realize texts are written by llms, due to style and mistakes, but it can become a serious problem in the near future, and we are kind of ignoring it, because no one is preparing in any way.
It will take someone really smart to figure this out without requiring some form of identity confirmation (which would be horrible imo), like using credit cards or (even worse) government digital IDs (I live in a country where we have a gov ID app which can be used to sign documents and confirm ID online but if I had to use it on Lemmy or any other social media I'd just fuck off).
Wasn't like one of his focuses going to be getting rid of bots? Did they just not do anything and the bots just invaded or did the bots just get better?
I think the focus on bots was primarily used in an attempt to back out of the deal. It was never an issue, especially if it inflated his ego by pumping up his account.