Exhibition Overview – Harold Cohen: AARON
Harold Cohen: AARON examines the practices and experimental programming of artist Harold Cohen, a pioneer in computer-based art. The exhibition looks at the evolution of his AARON software, the earliest artificial intelligence software for artmaking and one of the longest-running contemporary art projects. Fascinated by the computer’s power and potential as an artmaking tool, Cohen devoted his life to pushing the boundaries of AI’s possibilities and understanding what makes an image evocative. Throughout Cohen’s decades-long creative collaboration with AARON, the software underwent iterations that expanded its capabilities. Cohen focused on exploring AI as an instrument that translates artistic knowledge and process into code while creating with autonomy. The artist and his work are a crucial bridge between the worlds of art and engineering.
Harold Cohen first conceived his artmaking program at the University of California at San Diego in the late 1960s. From 1973–75, he further developed it at Stanford University’s Artificial Intelligence Lab and named it AARON in 1973. When working through AARON’s development, Cohen drew inspiration from children’s drawing processes, which typically begin with linking shapes and lines before connecting forms to represent an object or person. A 1973 visit to Chalfant Valley, California also influenced Cohen’s thought process. There, he saw prehistoric petroglyphs that furthered his grasp on how a system of marks can translate into an image with meaning. Cohen created AARON as an artmaking tool with the intention of exploring the nature of artistic representation. The program’s name alludes to the biblical figure Aaron, who was anointed speaker and mediator for his brother Moses, much like AARON was Cohen’s collaborator. Cohen’s background as a painter provided him with formal knowledge of artistic principles like color, line, composition, shape, and dimension. He equipped the software with compositional rules and comprehension of artistic paradigms and drawing strategies. For example, Cohen made AARON understand that it must begin drawing in the foreground before moving to the background. Cohen also seeded knowledge of the outside world and specific objects into AARON’s code, which became accessible in the software’s long-term memory. Evolving in several phases, AARON’s proficiency in these knowledge areas became the criteria for measuring the success of the program’s drawings, which were abstract in the early years.
The Figurative Phase
As AARON’s capabilities progressed, Cohen’s code started focusing on figures, often pairing them with plants or flowers in the compositions. Cohen was always striving to expand access to original artworks at a low price for a larger number of people. A reflection of these efforts was the distribution of a screensaver version of AARON, produced in collaboration with computer scientist Raymond Kurzweil’s CyberArt Technologies (KCAT). The KCAT software was released in 2001 and became one of the most widely known versions of AARON’s output.
Fundamentally different from the processes of currently popular AI software, AARON is a procedural system that mimics human decision-making to create images. Today’s AI image creation tools rely on algorithms making associations between images and text descriptions and, based on a user’s text prompts, generate their output from large data sets of existing images. In contrast, AARON determines its output with rules and instructions for completing tasks based on knowledge stored in its memory. Cohen sought to program a painter’s cognitive processes into his algorithms, providing AARON with an understanding of the relationship between line and form within a composition and the ability to imitate an abstract painter’s approach to representation in an image.