If we cannot think at the rate of an AI system, how can we consider all the risk variables associated with its own development and learning apparatus. Narrow AI is wonderful, but many years from now (hopefully), we will need to approach challenges like AGI will great care.