Post #3368902

That’s a neat experiment for several reasons. It shows limits of LLM capabilities, the importance of training data, context sensitivity, very dramatically shows that LLMs should not be trusted with important tasks if not supervised and that their advice has to be taken critically.