I can see how this issue will happen frequently, in diverse domains and abundantly. In other words, the prediction is that the most likely outcome will be a lo

I can see how this issue will happen frequently, in diverse domains and abundantly. In other words, the prediction is that the most likely outcome will be a lot of AI objects trained to be quite imbecile and will be optimal at that. The danger is that real people might be assumed to be guilty of things due to AI trained and automated imbecility. It's an ethical problem for the AI community and product designers.