Step through rounds between a forger and a detective. Each one improves because the other does, which is how the two networks in a GAN learn.
Each pushes the other to improve. Nobody had to define what a good forgery is — the competition works it out.
