DCGAN is initialized with random weights, so a random code plugged into your network would deliver a very random impression. On the other hand, while you might imagine, the network has a lot of parameters that we can tweak, plus the intention is to find a location of these parameters that makes samples produced from random codes appear like the tra