As the paper notes, this is a different perspective on generative adversarial networks:

  Goodfellow et al. (2014) proposed generative adversarial nets
  (GANs). GANs, while independently invented, are essentially
  based on the same idea as Turing Learning.