Until you can explain the entire logic path from input to output, even if the ai makes a logical mistake… Then you can’t trust the data.

I haven’t seen any results on another important requirement: An ai that can 'forget’or discard information that is bad.