Kat Caribeaux

Artificial Intelligentsia

“Language has been given to [people] so that [they] may make Surrealist use of it.”

Andre Breton, Surrealist Manifesto, 1924.

One hundred years ago in the wake of the First World War, a group of artists coalesced in Paris seeking a free and unencumbered access to the
imagination and subconscious. Under the name “Surrealists,” these artists used exercises in dream analysis and free association to guide their arts practice. In one such exercise called Exquisite Corpse, participants would make a drawing on a piece of paper guided by their subconscious, then fold the paper to hide their sketch and pass it to another artist who would continue the “automatic” drawing. Unfolded to reveal the composite drawing, Exquisite Corpse was not only a collaboration between artists, but with chance, serendipity, and happenstance.

Today, OpenAI’s image generation platform DALL-E2 takes its namesake from the the most popularly recognized artist associated with Surrealism:
Salvador Dali. DALL-E2 and other AI image generators are built on a diffusion model– a type of generative model trained on large collections of image data that obscures images with noise, then attempts to recover image data by learning to reverse the process. Much like the Surrealists who took the diffuse data of their own dreams and reassembled them into uncanny and haunting works of art, DALL-E2 uses its diffusion model to generate original images based on text prompts supplied by users.

Artificial Intelligentsia invites the creative writers at the helm of Northwestern’s literary magazine Helicon to engage in their own version of Exquisite Corpse with DALL-E2. Exchanging prose, prompts, and images to create new works, our participating writers explore the advantages, surprises, pitfalls, and challenges of AI image generation as a creative tool of the future.