The Mudcat Café TM
Thread #174300   Message #4228674
Posted By: Mr Red
17-Sep-25 - 05:51 AM
Thread Name: Artificial Intelligence - what could go wrong?
Subject: RE: Artificial Intelligence - what could go wrong?
Fer Gawds' sake (chose yer Gawds)

AI is not one thing. Deep Blue was configured around the game of Go, and reconfigured around protein folding (et al). And has tremendous potential for science.

AGI/GAI trawls the web for data and re-gurgitates it in a conversational context. And the web is 100% true isn't it? And just as the web has agendas marble throughout, so does AGI. Worse, now the average idiot will believe the results because they anthropomorphise. In the similar way they believe social media because it comes from a human (mostly?).

BUT

there is "model collapse" where people use AGI to write their text (or video) which is put on the web, which AGI trawls for................ Positive feedback (aka model collapse). Except collapse only refers to the validity of results, it will more likely break out into bubbles, where the "algorithm" is telling you what it thinks you want to know. How anti-social media developed is your clue. We thought it would spread the word, it did. But it localised the words (to like minds).

AGI will become de facto because people are lazy, and live busy lives (aka frippery). All aboard for the hell in a handcart.