aiTLAS

The images and text on this website have all been generated by artificial intelligence, particularly DeepAI for the text, and Stable Diffusion for the images. This allows us a unique opportunity to see how AI "sees" the world around us.

Consider this a parody of a real atlas, with generated images of maps, food, people, and my surprise favorite: some very unusual animals.

There is a lot of controversy about the material AI is trained on. Most of this revolves around copyright, but this art experiment is designed to explore how AI, particularly Stable Diffusion, renders people of different ethnicities. There is a clear American bias, but when it comes to outright racist visuals, the AI succeeds far more than it fails. However, I have left the bad examples in situ so developers can learn how to make AI more equitable.

AI is a moving train, and some of its flaws resulted in some really fun images, particularly animals.

Due to the speed at which AI is being developed, this is already out of date as new updated algorithms have been released in the time it took to render these thousands of images. This will likely result in some of the more problematic areas moot by the time you read this. Future AI will have less errors, which won't result in as many of the funny images you'll find in this site.