DCGAN is initialized with random weights, so a random code plugged in to the network would produce a completely random impression. However, as you may think, the network has numerous parameters that we will tweak, plus the intention is to find a setting of these parameters that makes samples created from random codes look like the training facts.