DCGAN is initialized with random weights, so a random code plugged into the network would deliver a totally random picture. On the other hand, as you might imagine, the network has an incredible number of parameters that we are able to tweak, and also the target is to find a placing of such parameters that makes samples created from random codes se