[rvz]

Well it doesn’t surprise me since I have been saying this for a while that these LLMs hallucinate nonsense to the point where you end up triple checking whatever it outputs.

LLMs thrive in applications that involve creativity and non-serious applications mostly around fantasy or creative writing. Anyone using them seriously outside of summarization for high risk use cases is going to be very disappointed.

[nickpeterson]

Perhaps the outcome is we get better at actually checking things, not a terrible result.

[geoduck14]

I recommend LLM users leverage the RAG technique