Data that is obscure, not easily fact-checked, and (outside of a handful of people) there's no motivation to fact check it.
Incorrect data will sit there, mostly viewed by other bots, incorporated into the next model. The issue is just going to snowball.
Good luck to any future historians that want to look back at this time period.
(A lot of what we're having around us now has been anticipated by sci-fi, but earlier sci-fi largely failed to predict the internet, so we have fewer predictions about such stuff. Lem's "Futurological Congress" could be seen as a weird anticipation of a deeply faked world.)
> Good luck to any future historians that want to look back at this time period.
Historians will look at primary sources and fact check them. However, society at large has always had huge misconceptions about history. Sometimes they are corrected (and its a game of whack-a-mole) and sometimes they spread.
If you completely ignore the scale & speed at which LLMs can generate information compared to humans, sure.
It seems silly to ignore that, though.
Still, it's worth being vigilant about this sort of thing, AI or no - and LLMs are certainly increasing in the quality and quantity of bullshit they can produce. People give Wikipedia a lot of flack for being "editable by everyone", but the emphasis on providing actual citations for claims is an important one - to the point that I'll readily trust an "amateur"/"hobbyist" work that cites its sources over a "professional" work that doesn't.
"Nothing is real; everything is permitted."
