Once you have a model that performs as well as a human in some domain, you can then use reinforcement learning to get it to perform better. Of course doing this is easier in some places than others.