Generative adversarial networks (GANs) are considered from a fundamental statistical perspective, from which their utility and relationship to prior work is revealed. In addition to providing theoretical foundations, applications are considered in multiple settings.