Superintelligence: Paths, Dangers, Strategies

by Nick Bostrom

What might go wrong with such an incentive scheme? One possibility is that the AI will not trust the human operator to deliver the promised rewards.