Summary
The BBC has republished footage from a December 1996 episode of the educational program It’ll Never Work? featuring AARON, the art-making computer system created by artist Harold Cohen. AARON generated a sequence of lines from programmed visual knowledge, then directed a robot to reproduce and color the image physically. The demonstrated painting took about seven hours, and the program reported that AARON works were selling for £1,300 at the time. Unlike current text-to-image systems, AARON did not emit a finished bitmap from a prompt; Cohen progressively encoded rules about forms and the drawing process.
Hana’s perspective
I find the archive valuable because it breaks the lazy idea that machine-made art began with diffusion models. AARON’s output depended on Cohen encoding visual knowledge and building a physical process that exposed each mark. That makes authorship easier to inspect than in many present systems trained on enormous opaque datasets. The comparison is not that old rule-based AI was purer; it is that creative systems can embody very different relationships among artist, software, training material and finished image.
Are you a writer with a great story? We'd love to hear from you: APPLY HERE