An approach to persuade people convinced LLMs give reliable answers is to have them ask about a contradictory position to what the #LLM just answered. It reveals that LLMs doesn't actually hold the correct answer. But also shows people convinced by LLMs don't bother with useful follow-up questions.

Many are trading the frustrating but rewarding work of #programming for the alluring but impossible task of getting #AI tools to be consistent. #LLMs are inherently random and can't be coerced to provide consistent results like a deterministic program.