DCGAN is initialized with random weights, so a random code plugged in to the network would generate a very random impression. Even so, while you might imagine, the network has many parameters that we can easily tweak, plus the purpose is to find a location of those parameters that makes samples created from random codes appear like the instruction