In this video, Dawkins acknowledges the possibilities of hard takeoff and open-ended recursive self-improvement in Artificial Intelligence.