GAN are two neural networks (AI, if you like) competing in a zero-sum game as a machine learning training method.
The two AI, use tactics something like the Socratic Method in training one another what a human trainer would teach them if the human only had the time.
We use GAN to teach and delight—as Horace said all art should do—and we use the techniques of GAN as a standard practice. That is to say, a lot of what we says is sort of a trap . . . uhm . . . That is to say, and we truly do not want to say too much on this, but we are most likely trying to trick you into forming a conclusive definitive decision on a definition or idea we proposed so that we can, when you least suspect it, turn your conclusion against you. We absolutely love setting traps. Our rebuttal need not make the slightest bit of sense, it is the authority and conviction of our rectitude that matters. This is the world of social media!