When the first artificial general intelligence is built, we want it to be our friend.
If it is not, we are in big trouble. That’s because an AGI will quickly become ultraintelligent: Whatever its goals, improving its own intelligence is a good a way of achieving them, and it won’t stop until it is much smarter than humans.
An ultraintelligent robot which hates us is a bad thing: We can’t possibly outsmart it.