Artist Trevor Paglen and AI researcher Kate Crawford have launched ImageNetRoulette, a viral digital art project that uses artificial intelligence to label user-uploaded photos. The project, which is part of their "Training Humans" exhibition at Fondazione Prada, gained massive social media traction by generating often offensive or bizarre classifications for users. By exposing the problematic labels—ranging from "mediatrix" to racial slurs and criminal accusations—the creators aim to reveal the deep-seated systemic biases embedded in the ImageNet database, one of the world's most influential AI training sets.
This initiative matters because it demystifies the perceived neutrality of artificial intelligence and facial recognition technologies. As these systems are increasingly integrated into social institutions—affecting hiring, policing, and surveillance—Paglen and Crawford’s work highlights the danger of allowing flawed, human-trained algorithms to classify and judge individuals. The project serves as a critical intervention in the conversation surrounding digital ethics and the political implications of computer vision.