Starting around 1973, the British painter Harold Cohen developed AARON, a computer program that generated original drawings without a human deciding each line. AARON was not a tool that executed a picture Cohen had already designed; it embodied rules about how to make a drawing and then produced its own compositions within that knowledge.
AARON was built in the symbolic, knowledge-based tradition rather than the data-driven one. Cohen hand-coded what an image needs to read as a coherent scene - how to close a contour, how to arrange figures, how to imply space - and the program applied those rules to produce endless distinct drawings. Early on, output was rendered by a “turtle” robot that crawled across paper on the gallery floor; a 1979 exhibition showed the machine drawing live in front of visitors. Over the following years Cohen extended AARON from line drawing to autonomous coloring, a step he reached in the mid-1990s.
AARON stands at the start of the computer-art lineage that runs through later generative systems. Where modern text-to-image models learn statistics from millions of images, AARON encoded an artist’s expertise by hand - the same symbolic approach that defined expert systems of its era. Cohen worked on the program for roughly forty years, making it one of the longest-running art projects in computing.
Why business readers should care: AARON shows that “the computer made the art” is a fifty-year-old idea, and that the question of authorship - who owns what a program produces - long predates the current debate.