DCGAN is initialized with random weights, so a random code plugged into your network would crank out a totally random image. However, while you might imagine, the network has many parameters that we can tweak, as well as objective is to find a setting of such parameters which makes samples created from random codes appear like the instruction data.