DCGAN is initialized with random weights, so a random code plugged into the network would generate a completely random picture. Nevertheless, when you might imagine, the network has millions of parameters that we could tweak, as well as the objective is to locate a location of these parameters which makes samples created from random codes seem like