3 ms·

It depends on the AI. ChatGPT's higher models (o1-pro/o3/o4-mini-high) have some kind of limited capability to detect errors in the user's thinking, and have re

It depends on the AI. ChatGPT's higher models (o1-pro/o3/o4-mini-high) have some kind of limited capability to detect errors in the user's thinking, and have relatively few hallucinations.

UltraSane 1y ago
I've had fun debates about things like p-zombies with Gemini 2.5 Pro
energy123 1y ago
o3 have twice the hallucinations of o1 according to their own hallucination benchmark