DCGAN is initialized with random weights, so a random code plugged into the network would crank out a very random picture. Nevertheless, while you may think, the network has a lot of parameters that we could tweak, along with the aim is to find a location of these parameters that makes samples produced from random codes seem like the schooling info