DCGAN is initialized with random weights, so a random code plugged into your network would create a very random image. On the other hand, while you may think, the network has many parameters that we are able to tweak, and the aim is to locate a setting of such parameters which makes samples produced from random codes look like the teaching informat