Superintelligence: Paths, Dangers, Strategies

by Nick Bostrom

We have already cautioned against anthropomorphizing the capabilities of a superintelligent AI. This warning should be extended to pertain to its motivations as well.