DCGAN is initialized with random weights, so a random code plugged in to the network would deliver a very random image. Having said that, as you may think, the network has a lot of parameters that we can tweak, and also the goal is to find a environment of such parameters that makes samples created from random codes appear to be the training detail