Perhaps that was not the right word to use. I meant to say that artificial intelligence that can write its own software will have an exponentially increasing effect. It is a little scary to think about, but AI cannot understand abstracts. It can spit out something that one may quantify as abstract in an image or text but the meaning behind it will always be decided by humans. Robots cannot understand subjective reasoning. They can try to emulate it based on input but grasping the concept is not possible. I would almost argue that AI that is programmed to do one thing and one thing only would be more dangerous than one that accepts input to return a result.