People have been having fun with the text-to-image generators lately. Enter a description, and the AI churns out believable and sometimes detailed images that match the input. The reason these systems work is because the models were trained on a lot of data, in the form of images. Andy Baio and Simon Willison made a tool to browse a subset of this data behind the recently released Stable Diffusion.
Images behind the generated images from Stable Diffusion
