LLMs Hallucinate but Get It Right

maybe a crazy idea, but what if LLMs hallucinate 100% of the time, and just happen to be right 80% of the time? This is exactly what's going on, not even an approximation :-)

A LinkedIn comment thread jokes that large language models hallucinate constantly but still happen to produce correct answers surprisingly often. The reply claims this is exactly how LLMs work, making fun of AI reliability and confident inaccuracies. EN: A LinkedIn comment thread jokes that large language models hallucinate constantly but still happen to produce correct answers surprisingly often. The reply claims this is exactly how LLMs work, making fun of AI reliability and confident inaccuracies.

• @ithueti / 15898 2025-05-06