This is why AI makes many of us a little nervous, feed it a biased dataset, and well, we have a psychopath.
“In one way, we shouldn’t be surprised – our AI systems have already proven to be fraught with human biases. But this experiment highlights the need for an ethical approach to AI. With poor datasets, you cannot rely on the decisions an AI system makes – and you can alter its perception of the world.”
Granted this experiment went out of it’s way to create Norman, but that’s kind of the point. Someone will do something similar, either on purpose, or by accident, by training AI on some kind of biased data, or a set of really dark human behavior that it “learns” to be normal. What then?