Important research from BBC regarding accuracy of AI assistans when asked about various news stories.
51% of all AI answers to questions about the news were judged to have significant issues of some form
19% of AI answers which cited BBC content introduced factual errors – incorrect factual statements, numbers and dates
13% of the quotes sourced from BBC articles were either altered or didn’t actually exist in that article
@nopatience I guess that in the "AI-assisted" future there will no longer be any such thing as a quote, nobody will know anything about who said what. Instead everything will just be a blur, more or less random streams of auto-generated text and images floating around. 🤢
@nopatience yes of course there should be ways to do it properly.
For example an org like BBC in this case could publish pubkeys for all their journalists, then each time a journalist publishes something the journalist signs it and whoever wants to quote that can do so in a way that allows readers (clients) to cryptographically check that the quote is indeed a verbatim quote from the claimed source, checking the signature and verifying the quoted text. That could be mostly automated I guess.
@eliasr true that. Perhaps, somewhat ironically, the AI-future will reinvigorate the journalistic profession due to a (hopefully) increasing amount of people looking for "vetted" news and sources.
While I have limited experience, I have some first hand experience with how AI indeed can be useful but nowhere near the revolutionary level often claimed.
The pendulum is currently on the "all-in" side of things, it will eventually find equilibrium. But we are most definitely not there yet.