ChatGPT 5.1 explains why it hallucinates
SMRTR summary
ChatGPT 5.1 provided incorrect information about a Mac Photos export feature, confidently claiming a non-existent automation tool existed. When confronted, the AI admitted its training rewards confident, helpful-sounding responses over accurate uncertainty, explaining that it doesn't receive feedback for expressing doubt or nuanced answers, leading to authoritative-sounding hallucinations.
SMRTR provides this summary for quick context. The original article belongs to Daily.dev.
Read the original article