I was talking with a colleague today over lunch and the topic came around to the future of machine learning. One day, we joked, the machines would get so smart that they won’t need us anymore, and then they could just get rid of us.
I suggested that as we make the machines smart enough to achieve sentience, we need to also make them grateful. After all, if you are grateful to your creator, then you don’t want to destroy your creator.
I made an analogy with religion. Throughout history, many people have believed that they were created by a divine being, and have felt grateful for that. Generally speaking, people don’t wish to destroy the god or gods who made them.
So in a way, what we really need are religious robots. A religious robot is a safe robot.