News for hackers. Stuff that matters.
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB by mplewis on Saturday December 21, 2024 @12:41AM (#42477767)
Humans can tell you how confident they are in something being right or wrong. An LLM has no internal model and cannot do such a thing.
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB by swiftcoder on Saturday December 21, 2024 @02:20AM (#42478037)
> Humans can tell you how confident they are in something being right or wrong

Humans are also very confidently wrong a considerable portion of the time. Particularly about anything outside their direct expertise

Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB by SketchySeaBeast on Saturday December 21, 2024 @11:00AM (#42480381)
People only being willing to say they are unsure some of the time is still better than LLMs. I suppose, given that everything is outside of their area of expertise, it's very human of them.
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB by daveguy on Saturday December 21, 2024 @11:00AM (#42480383)
That's still better than never being able to make an accurate confidence assessment. The fact that this is worse outside your expertise is a main reason why expertise is so valued in hiring decisions.