If you’re still looking to make the connection in your head between AI/deep-learning applications and actual creative output, consider this new science project from Nvidia Research Exhibit A. The software uses the deep-learning technique of generative adversarial networks, also known as GANs, to create surprisingly realistic imagery from the barest of sketches. With tongue in cheek, Nvidia calls it GauGAN. Watch the video above, released to coincide with Nvidia’s GPU Technology Conference in San Jose, to see it in action and stop by the Nvidia website for more details.