The essential point is that machine can teach us by giving solutions to problem (or games) that we can't solve. As times goes by, the role of machines as teachers will need to be explored, and probably if such a system provides good results it will be standardized. If there is a new way to learn we need to reflect about how to apply it in the best way and how to explore all the juice we can extract from mit, far from being a pointless action, we need to frame it with the information we are getting from the improvement in ML. We need to develop the concepts and ideas about it because the reward could be immense.