There is no skepticism. LLMs are fundamentally lossy and as a result they’ll always give some wrong result/response somewhere. If they are connected to a data source, this can reduce the error rate but not eliminate it.

I use LLMs but only for things that I have a good understanding of.