Abstract

The article analyzes the reasons why the images generated by deep learning algorithms (such as DeepDream, the Generative Adversarial Networks), or the recent text-to-image models (such as DALL-E 2, Stable Diffusion, and Midjourney) are often described as “dreamsµ, “visionsµ and “hallucinationsµ of the “machineµ, and considered as “surrealµ or “surrealistµ. By examining a recent work by the artist Grégory Chatonsky and the images generated by DeepDream, the author shows that in some cases, rather than being merely superficial, the connections between AI-generated images and Surrealism are grounded in some specific analogies having to do with the relations between words and images, with the psychological phenomena of apophenia and pareidolia, and with the tradition of images hidden within images.