Here's the thing. In interpersomal communication we listen to feedback to understand if a person is upset, harmed, confused or misinterprets.
If an unsupervised AI is not programmed to interpret feedback it can continue to harm people indefinitely.
#digitalethics