DCGAN is initialized with random weights, so a random code plugged into the network would create a completely random graphic. Nonetheless, while you might imagine, the network has numerous parameters that we will tweak, and the purpose is to find a setting of these parameters which makes samples created from random codes appear like the teaching