What Happens When A.I. Stops Doing What Humans Want?

“Alignment” is the science of teaching A.I. to do what is in line with human preferences, ethics and judgment. But, at times, the systems have gone rogue.