What's going on when the recipient of a graphic composition is a machine ? The recognition of shapes by machines is not new (look at OCR typographies or QR codes) but at a point where are recognized not only forms, but also the relations that can exist between these forms. The machine begins to understand the composition, it can describe what it sees by natural sentences: "a man riding a bike down a street next to a traffic light " ect... Will the machine be able to describe graphic design? And to do what? Let us examine this new capacity of the machine to see the world by proposing graphic systems to be seen.