![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Full headline, blah, Dreamwidth subject field sucks, yadda: "ChatGPT use declines as users complain about ‘dumber’ answers, and the reason might be AI’s biggest threat for the future"
The gist: It's more a threat to the AIs themselves, rather than to humanity or whatever. AIs are apparently "cannibalizing" themselves now. In other words, with the rapid proliferation of subpar AI-created content starting to pollute the Internet now, and then that content, along with everything else, being scraped to train future iterations of AI, that all seems to be fucking up the AIs.
Seems to me like yet another reason in a big list of reasons why it was a terrible idea to hook up these LLM/AI systems to pretty much every part of the Internet in the aimless, careless way it has been done over the past year or so. Or, at the very least, it's another reason why it's a terrible idea to just scrape the entire Internet, willy nilly, to train one's LLM/AI system.
Looks like even AI doesn't like consuming cruddy AI-crap.
The gist: It's more a threat to the AIs themselves, rather than to humanity or whatever. AIs are apparently "cannibalizing" themselves now. In other words, with the rapid proliferation of subpar AI-created content starting to pollute the Internet now, and then that content, along with everything else, being scraped to train future iterations of AI, that all seems to be fucking up the AIs.
Seems to me like yet another reason in a big list of reasons why it was a terrible idea to hook up these LLM/AI systems to pretty much every part of the Internet in the aimless, careless way it has been done over the past year or so. Or, at the very least, it's another reason why it's a terrible idea to just scrape the entire Internet, willy nilly, to train one's LLM/AI system.
Looks like even AI doesn't like consuming cruddy AI-crap.
