In 2014, Ian Goodfellow invented Generative Adversarial Networks. This system consists of a “generator” and a “discriminator”, where the “generator” is in charge of drawing images for the “discriminator” to attempt to distinguish real images from the drawings. In this way, the “discriminator” is trained to tell apart which images are real and which ones are generated. At the end, the “generator” will be trained to create images so perfect that the “discriminator” won’t be able to tell apart generated from real images.
For more information read here.