DCGAN is initialized with random weights, so a random code plugged to the network would deliver a very random impression. On the other hand, as you may think, the network has many parameters that we will tweak, and also the purpose is to find a placing of those parameters that makes samples ge