IMO "confabulation" is more accurate than "hallucination" because the former indicates a lack of intent. Given that LLMs are not sentient, they lack intention. At most, they are reflexively responding to a reward function that optimizes towards producing text roughly resembling the pattern of their training data, but that's different from intent.
davemwilburn@infosec.exchange
@davemwilburn@infosec.exchange
Indlæg
-
I've never been opposed to the word "hallucinating" for describing how AI makes mistakes ... until now. -
This is so deeply embarrassing and shocking at the same time.It's OK, Brian. It's not like we have any historic precedent for what happens to a dehumanized population in a country ruled by a corrupt authoritarian dictator whose government is intensively engaged in dragging them away into concentration camps while fighting illegal wars of conquest.
-
*gently grabs the cheeks of all programmers to stare deeply into their eyes*Sorry, best I can do is an excruciating and meandering 29-minute long YouTube video sponsored by Nord VPN.