News for hackers. Stuff that matters.

A comment on Hallucination is inevitable: An innate limitation of large language models.

Re:Hallucination is inevitable: An innate limitation of larg... by DinaCoder99 on Sunday February 25, 2024 @02:16PM (#39503809)
"Hallucination" implies perception of non-real things, not generation of phrases that map poorly to reality (or are simply incoherent). It seems like a really bad term for this phenomenon.
[ Parent | on HN ]
Re:Hallucination is inevitable: An innate limitation of larg... by bonzaidrinkingb on Sunday February 25, 2024 @02:35PM (#39503998)
We use "confabulation".

It is a feature, not a bug.

"confabulation" could be "solved" when LLMs realize they are uncertain on a reply and making things up. But that requires the LLM saying "I don't know" and rewarding that more than a wrong guess. That requires a change in loss functions and not even sure if all users desire that.

[ Parent | on HN ]
Re:Hallucination is inevitable: An innate limitation of larg... by TheBlight on Sunday February 25, 2024 @02:21PM (#39503865)
"Bullsh***ing" seems more apt.
[ Parent | on HN ]
Re:Hallucination is inevitable: An innate limitation of larg... by earthwalker99 on Sunday February 25, 2024 @02:23PM (#39503886)
It makes being wrong sound impressive and mysterious, so it's here to stay.
[ Parent | on HN ]