Interesting blog post! From a 10,000 ft overview is this method utilizing the conditional posterior's to incorporate uncertainty information into the model GAN

Interesting blog post! From a 10,000 ft overview is this method utilizing the conditional posterior's to incorporate uncertainty information into the model GAN model? If so how does it differ from using drop-out's to approximate Bayesian behavior ala https://arxiv.org/pdf/1506.02142.pdf https://arxiv.org/pdf/1506.02142.pdf ?