I would not say I have had a "bad" experience with IA for I have always been conscious that it will unavoidably hallucinate, for once because it is reported permanently both by users and companies of origin, and fundamentally because I myself know that if we ourselves can hallucinate when making an investigation, it's simply impossible for an AI to not to do the same. Too many variables on the fly.

I did made some basic tests with some questions I knew the answer to, and while the models have made great progress very fast, it still falls way too short to be barely usable.

However, I do like to some times ask the AI about some subjects and problems because even when I know it will get it wrong, 50% of the time its slop gives me a clue on something, and 50% of that 50% of the time, it even gives me a clue indirectly, simply because the process of reviewing the slop sparks my mind into the right direction.

With that careful use, my experience has always been positive so far, although this procedure implies minimal use.