Modality and Technique: Collage vs. Computation

In terms of medium and method, Dada and AI art could not be more different on the surface. Dada works were hand-crafted (even if they involved found materials or chance). A Dada collage might include newspaper clippings, ticket stubs, photographs, and bits of string all pasted together – very much a physical cut-and-paste process. The human touch and material presence were evident, even if the composition ignored aesthetic conventions. In contrast, AI-generated art is created digitally, by a computer program running on silicon chips. The “materials” it uses are numbers (pixel values, algorithmic weights) and it “assembles” images in an abstract latent space rather than on a canvas. However, at a conceptual level, there is a strong analogy: both are a form of collage. Dada photomontage literally cut up existing images to create a new image; AI models figuratively cut up learned representations of millions of images to synthesize a new one. In Dada collage, the sources might be a 1916 newspaper photo of a politician and a magazine image of an axe, combined to make a statement. In AI, the sources are innumerable tiny fragments of visual data. For example, the AI might have “seen” thousands of horses, hundreds of astronauts, pictures of Mars, and so on. When asked for “astronaut on horseback”, it draws on all those examples, merging pieces of each (the general form of a horse from here, a style of space suit from there, color tones from elsewhere) into one composition. Neither the Dadaist nor the AI model necessarily respects the original context or intent of the source imagery. John Heartfield’s anti-Nazi Dada photomontages, for instance, repurposed images from newspapers in ways never intended by their photographers (a rough analog to how an AI might repurpose a living artist’s style in a new prompt image without that artist’s consent). Thus, both practices raise questions of appropriation. Dada was actually ahead of its time in this regard; it introduced the idea that borrowing and remixing existing media is a valid art technique, something we now see amplified in AI art (and digital remix culture generally). We might say that AI art is collage on a massive scale: automated and codified. A human Dada collagist physically selects and cuts source images, guided by subconscious or conscious decisions. A generative AI “selects” from visual patterns statistically, guided by the prompt and its training weights. The modality differs (analog versus digital, deliberate manual composition versus algorithmic generation), but the outcome is comparable. It’s a composite image that draws from prior works.

There is also a contrast in tactility and randomness. Dada collages often looked obviously assembled, with jagged edges and mismatched fragments that were part of the aesthetic. The process of chance (dropping paper, etc.) was often left visible as part of the art’s meaning (embracing imperfection and accident). AI images, by design, aim for coherence, often photorealism or a seamlessly painted look, and to the point that artists who use those styles (that were stolen) will get witch-hunted. The “joins” of the collage are invisible; the output is meant to look like a single, unified image. Any randomness in the generation process is hidden in the final product (except when the AI makes mistakes, like giving a figure too many fingers). In short, Dada’s modality was material deconstruction, whereas AI’s modality is mathematical reconstruction.