A new study shows that AI is overconfident but doesn’t recalibrate as well as us human beings
Press enter or click to view image in full size
You’ve reported before on how AI is overconfident, haven’t you?
Yes, we probably all know that AI appears confident because of its language use: a previous study likened this to certain brain conditions whereby we humans can use language confidently but lose sense of meaning. We know this happens with AI.
We are also missing certain cues in contrast to humans. When we are less confident our voice changes, we may hesitate, or add a qualifier such as “I think”.
Well, we would think it could, but a new piece of research by Trent Cash and colleagues of Carnegie Mellon University shows it can’t. Unlike us human beings!
Isn’t recalibration and essential part of learning and the brain?
