News for hackers. Stuff that matters.
Re:Twitter’s Lawyers Admit They’re Overwhelmed; 2000 Laid Of... by TeMPOraL on Monday June 19, 2023 @11:23AM (#36392105)
Well, if they're rational enough... we're in a situation analogous to two hypothetical superhuman AIs on a brink of conflict: both can honestly and accurately estimate their chances of winning, and thus skip the fighting and go straight to agreement that favors the winning side as much as their estimate says it would anyway.
Re:Twitter’s Lawyers Admit They’re Overwhelmed; 2000 Laid Of... by dylan604 on Monday June 19, 2023 @11:59AM (#36392570)
Which superhuman will be the first to suggest a nice game of chess instead?
Re:Twitter’s Lawyers Admit They’re Overwhelmed; 2000 Laid Of... by cyanydeez on Monday June 19, 2023 @05:12PM (#36396526)
Did you just assume no hallucinations?
Re:Twitter’s Lawyers Admit They’re Overwhelmed; 2000 Laid Of... by TeMPOraL on Monday June 19, 2023 @05:28PM (#36396712)
Yes. I'm not talking about GPT-4 on steroids, but a proper GAI.

FWIW, I find hallucinations of LLMs to not be unusual or that big of a deal, because if comparing to human mind, it's pretty clear to me that LLMs are the equivalent of the "inner voice", which - at least for me - tends to spit out things that "feel right", but aren't necessary factually correct. This suggests that LLMs alone won't make a GAI, but they may be an important component of it.