Harold Cohen

Harold Cohen is a British artist who has lived in California since 1968. His major interest has always been color. However, for several decades he also concentrated on the representation problem: How little effort in lines does it take to generate an image that people can easily recognize as representing something specific? In the early 1960s, Cohen had already been successful as a painter—convincingly demonstrated by his participation in the Venice Biennale 1966, as one of five representatives of the UK. In 1968 he came into close contact with the computer’s capabilities.

He moved to San Diego to work as a visiting professor in fine art at the University of California, San Diego. He soon met Jeff Raskin, who showed him what flowcharts could be good for. This episode ended when Raskin gave him a Ditran (Diagnostic Fortran) manual. Cohen continued to work on his own and began investigating methods of Artificial Intelligence for applications in fine art. In 1970 he met Edward Feigenbaum, who invited him to the AI Lab at Stanford University. There Cohen informally learned the programming language SAIL, which at the AI Lab was the most used for memory-intensive and real-time applications. Cohen needed this level of performance for the real-time exhibitions he was preparing.

During the 1970s, he constructed computer-controlled drawing machines. They controlled a “Turtle” moving on the floor. Different from the MIT Turtle Cohen’s device moved by spinning two wheels at different speeds, which resulted in very smooth curves. The Turtle, via a sonar navigation system, even knew where it was on the drawing—a feature necessary to control the device on an area of three by nine meters.

Cohen’s program, now called AARON, made its first public appearances at the Los Angeles County Museum in 1972, and the La Jolla Museum in 1973. For these occasions AARON controlled hand-built flatbed plotters. The Turtle first went on display at documenta VI at Kassel, Germany in 1977.

Cohen developed AARON into an enormously powerful and complex rule-based system. In the 1990s, the drawing machines were replaced by painting machines (now in the permanent collection of the Museum of Computing History in Mountainview, CA.). Typical subject matters were portraits of “imagined” people, as well as still-lifes.

In 2002, he abandoned the painting machines in favor of wide-format printers. Although, color had been central to his work all along it was not until 1985/86 that he began to see how a program without vision could become an expert colorist. The resultant rule-based system became intractable by the end of 2005, when he replaced it by an algorithm, which AARON has been using ever since.