Alright here's some explanation for everyone freaking out in the comments
This video, and images like it, were created with a neural network (I think Google made it.) This specific network is used for grouping images. It looks through a set of images and determines which ones have dogs or cats or birds or some other distinguishable feature and groups those images together. Over time the network learns what a dog looks like and when you reverse the process and ask the network to provide an image of a dog, it will give you its understanding of what a dog is. So (as I understand it) it takes images and tries to find features that look like dogs (or something else) and it exaggerates those features. The process can be repeated on the same image for a stronger effect. It can make some pretty horrifying things, but it can make some beautiful stuff as well.