Harold Cohen — AARON (first public drawings)

1973

Artists: Harold Cohen · Year: 1973 · Medium: Plotter drawings (later: robotic painting machine, then color pigment output) · Substrate: AARON — a C / Lisp program of Cohen's own design, evolving across 40+ years · Place: La Jolla, California, USA · Institution: University of California, San Diego Harold Cohen began writing the first version of AARON in 1971 and showed its first public drawings at the San Francisco Museum of Modern Art in 1973, as part of a residency at Stanford's Artificial Intelligence Laboratory (SAIL). The exhibition consisted of large drawings on paper executed live, on gallery walls, by a custom-built four-wheeled "turtle" drawing robot that carried a felt-tip marker and received its movement commands from a nearby PDP-10 running Cohen's AARON program. The drawings were bounded, irregular, organic contour compositions — shapes that hovered between representational and abstract, with internal hatching, open areas, and the visible wandering of a line that had made decisions as it went. AARON is the single most sustained project in the history of computer art: Cohen wrote and rewrote it for forty-three years, from 1971 until his death in 2016, producing successive versions that evolved from abstract line-drawing through representational figure-drawing to fully-colored paintings whose figures inhabited complex painted interior scenes. The 1973 version — pre-representational, pre-color, strictly abstract — is where everything starts. What distinguished AARON from its contemporaries was Cohen's specific claim. He was not generating random compositions from algorithmic constraints in the Stuttgart manner. He was attempting to encode, explicitly, the knowledge an artist deploys when drawing: rules about where to put a line, rules about when to break a region, rules about how to build figure-ground relationships. AARON was an expert system for drawing, in the specific AI-research sense of the late 1960s, and Cohen pursued the project alongside the mainstream AI community — he published in AI conferences, collaborated with Edward Feigenbaum at Stanford, and spent years at SAIL as a visiting researcher. The 1973 SFMOMA exhibition was a sensation in the Bay Area art world and a curiosity in the national press. It is one of the earliest cases in which a work of computer art was framed in an American museum not as a technology demonstration but as an artwork. Every subsequent generative-art practice that treats drawing as the primary medium — from Casey Reas's Process series to Tyler Hobbs's Fidenza — descends from Cohen's 1973 framing. Lore Cohen was a successful British figurative painter when he took a 1968 visiting professorship at UC San Diego; he had represented Britain at the Venice Biennale in 1966. What was supposed to be a one-year visit became a life in California, and the AI-department connections he made at UCSD led to the move to Stanford SAIL in 1971. He taught himself programming in LISP. The first turtle robot was hand-built by Cohen in his SAIL office with motor components salvaged from old photocopiers; a running joke at SAIL was that Cohen's robot was the only one in the lab that made art instead of trying to find coffee. Cohen and the turtle drew together, every working day, for over forty years — first in Cohen's Encinitas studio, later in his hillside studio near UC Santa Cruz. After his death, his widow Becky Cohen has continued to run retrospectives of the AARON work; the forty-three-year archive of programs is held at the Computer History Museum. Sources • McCorduck, Pamela — Aaron's Code: Meta-Art, Artificial Intelligence, and the Work of Harold Cohen (Freeman, 1991). • Computer History Museum — AARON program archive. • Cohen, Harold — "The Further Exploits of AARON, Painter" (Stanford Humanities Review, 1995).