For a machine to understand in depth, it is necessary to break with the notion of "one-shot" discussion. Indeed, we humans, when we do not understand something, we ask questions to refine and understand. For example, we ask questions to clear up an ambiguity, to identify something we are not sure we understand ... Most of the time, we end up understanding. It is a functional algorithm. Here is the deep understanding map of my artificial intelligence 'Lisa' (move around the map):