Are you learning new things with AI in completely new domains, or still within your domain? If you are within your domain, you still have your experience to filter out the inaccuracies. Try with something very different.

Just couple of days ago I spotted a significant but subtle error in a somewhat straightforward summarisation task I was trying with Llama. It could have easily slipped past me if I were not careful.