DCGAN is initialized with random weights, so a random code plugged into your network would produce a very random image. On the other hand, while you may think, the network has millions of parameters that we can easily tweak, plus the objective is to find a placing of these parameters that makes samples generated from random codes appear to be the