News for hackers. Stuff that matters.
Re:Experts starting to doubt AI ‘hallucinations’ will go awa... by mindcrime on Wednesday August 02, 2023 @06:09PM (#36978433)
As someone who also works professionally in this field, I'd be curious to know what you find problematic about my comment. I'll grant you that I may be over-simplifying a bit, but it seems appropriate in the context at hand.
Re:Experts starting to doubt AI ‘hallucinations’ will go awa... by heyitsguay on Wednesday August 02, 2023 @06:27PM (#36978627)
Actually you're right, that was my mistake. Current-gen autoregressive token predictor "hallucinations" probably can't be fixed, but yeah, the way forward will involve new architectures and I would place my bets that something like world models will be involved to ground generative output.
Re:Experts starting to doubt AI ‘hallucinations’ will go awa... by mindcrime on Wednesday August 02, 2023 @06:29PM (#36978653)
Ah OK. You had me scared there for a minute. :-)