DCGAN is initialized with random weights, so a random code plugged to the network would deliver a totally random impression. Nevertheless, as you may think, the network has countless parameters that we can tweak, plus the target is to locate a location of such parameters that makes samples generated from random codes look like the schooling details