The machine-learning model behind Refik Anadol’s installation Unsupervised, shown at the Museum of Modern Art from November 2022, was trained on 138,151 pieces from MoMA’s digitized collection - spanning more than two centuries of modern and contemporary art. Anadol used StyleGAN2 with adaptive discriminator augmentation to learn from this archive, then let the model continuously generate new, never-repeating imagery on a large media wall in the museum’s lobby.