It would be fascinating if we could get an AI that can only reply with conclusions based on empirical evidence. I feel that it would probably be called leftist, since more leftist viewpoints seem to be science based than rightist ones (especially in the US).
It would also confuse the heck out of people, because it will change its answers based on new evidence.
It would also confuse the heck out of people, because it will change its answers based on new evidence.