It has to be messing with the heads of AI programmers everwhere. An AI app that has unfiltered access to all scientific knowledge is going to give answers that are sexist, racist, and classist by Progressive standards. So how do you make an AI program that is accurate about everything except a few forbidden conclusions?