Russell, Bostrom and the Risk of AI