Twenty Years to Draw: Harold Cohen's AARON
Long before text prompts, the painter Harold Cohen spent decades teaching a program called AARON to draw, then to paint.
In the late 1960s Harold Cohen was an established London painter whose work had represented Britain at the Venice Biennale. Then he took a post at the University of California, San Diego, where a graduate student named Jef Raskin introduced him to the campus mainframe and to FORTRAN. Cohen wanted to know how artists turn what they know into marks, and a machine that only does what it is told looked like the place to find out. By the early 1970s the software had a name: AARON, after the biblical brother who spoke for Moses.
Early AARON worked in black and white. A small robot "turtle" carrying a marker crawled across the paper, and in 1979 one drew live on the floor of the San Francisco Museum of Modern Art, next to a hundred-foot mural of AARON's forms colored by Cohen's own hand. Through the 1980s he gave it more to know: plants, tables, people. "It's taken me 20 years to teach AARON to draw. How can I possibly teach it to color before I die?" he said in 1989. By 1995, at the Computer Museum in Boston, a robotic arm was mixing its own paints and brushing them onto AARON's drawings.
Cohen called it a collaboration, and he never let it sign its own work. If it ever did, he told the Computer History Museum, "that would signal the end of all debate on the matter of its intelligence." He died in 2016, at 87.
In 2024 the Whitney Museum of American Art, which acquired AARON with funds from its Digital Art Committee, set plotters drawing live in the gallery for Harold Cohen: AARON, just as DALL-E, Midjourney and Stable Diffusion went mainstream. The Whitney calls it the earliest AI program for artmaking. It spent nearly half a century learning from one painter.
Sources