A Short History of Generative Art

Generative art in its computer form started in the early 1960s and went public in 1965, when Georg Nees showed plotter drawings in Stuttgart in February and A. Michael Noll and Bela Julesz showed computer-generated pictures in New York in April. The ideas behind it are older. Artists were handing decisions to chance about fifty years before they had a computer to run their rules.

This history goes further than the short origin story in our guide, What Is Generative Art?. It follows the practice from Dada chance procedures through the 1960s plotter pioneers, Harold Cohen's AARON, the artist-built software of the 2000s and long-form editions on blockchains, to the live work being made now. Every date is sourced, with the full list at the end.

Key takeaways

• The first public exhibitions of computer-generated art took place in 1965: Stuttgart in February, New York in April and Stuttgart again in November.
• Chance and rules came first. Marcel Duchamp's dropped threads (1913-14) and Jean Arp's chance collages (1916-17) predate computer art by about fifty years.
• The word "generative" entered the field through Stuttgart, in Georg Nees's doctorate on generative computer graphics, supervised by the philosopher Max Bense.
• Harold Cohen conceived AARON in the late 1960s and developed it until 2016. The Whitney Museum of American Art calls it the earliest AI program for artmaking.
• Programming tools made for artists followed: Design By Numbers (1999), Processing (2001), openFrameworks (begun around 2004-05) and p5.js (2013).
• Art Blocks launched in November 2020. On it, a hash created when a collector mints is fed into the artist's script to produce that collector's output.

When did generative art start?

Computer-generated art was first shown in public in 1965. The Digital Art Museum (DAM) timeline lists three shows that year: Georg Nees at the Studiengalerie of the Technische Hochschule Stuttgart from February 5 to 19, Noll and Julesz at the Howard Wise Gallery in New York in April, and Frieder Nake with Nees at Galerie Niedlich in Stuttgart from November 5 to 26 (DAM).

The images are slightly older than the shows. The V&A dates Noll's Gaussian-Quadratic to 1962-63 and cites a Bell Labs memo of August 28, 1962, in which Noll described "an exploratory series of design-producing experiments" made with an IBM 7090 computer and a microfilm plotter (V&A).

There were machine images before that. The British Computer Society notes that Ben Laposky made abstract images in the 1950s by photographing the signal on an oscilloscope, and that he is often credited with the first graphic image generated by a machine (BCS). Those were analog, so this history starts with the digital work of 1965.

What came before computer-generated art?

Chance procedures came first, and instruction-based art developed alongside the computer. In 1913-14 Marcel Duchamp dropped a one-meter thread from a height of one meter onto a painted canvas, repeated it twice more, and kept the three curves as 3 Standard Stoppages, now at MoMA. MoMA's label quotes him saying he no longer wanted to do anything with his hands (MoMA).

In 1916-17 Jean Arp made collages in a series called According to the Laws of Chance. MoMA, quoting the artist Hans Richter, describes paper scraps pasted down where they fell, and adds that the ordered look of the results suggests Arp never fully gave up control (MoMA). That balance between accident and authorship is the same one a generative artist sets in code.

Vera Molnár bridged the hand-made and the computed. The Centre Pompidou lists her Quatre éléments distribués au hasard (Four Randomly Distributed Elements) from 1959. It says that in the 1960s she worked out a method she called the "machine imaginaire," and in 1968 became the first artist in France to make digital drawings with a computer connected to a plotter (Centre Pompidou).

In 1967 Sol LeWitt gave the idea its slogan in Artforum: "The idea becomes a machine that makes the art" (BCS). His works were often made by assistants following his instructions, and Tate notes that he accepted the different outcomes as part of the process (Tate).

What happened at the 1965 exhibitions in Stuttgart and New York?

Stuttgart came first. Georg Nees was a mathematician and software engineer at Siemens, where, according to the V&A, he helped the company acquire a Zuse Graphomat, a drawing machine driven by punched tape, and used the scientific language ALGOL to make aesthetic images (V&A). The philosopher Max Bense invited him to show at the Technische Hochschule.

Frieder Nake, then a student, later recalled in a conversation published by DAM that the show held roughly eight to twelve graphics. The next day he told Bense he could show him similar work from the university computing center. That led to the November show at the Stuttgart gallery and bookshop of Wendelin Niedlich, where Bense insisted that Nees be shown alongside Nake (DAM).

Nake's Hommage à Paul Klee, 13/9/65 Nr. 2 shows the method. The V&A describes an algorithm built on the proportions of Klee's 1929 painting High Roads and Byroads, drawn on a pen plotter, with random variables written in so the computer could choose among a set number of options (V&A).

In New York, Noll and Julesz were researchers at Bell Labs in Murray Hill, New Jersey. Howard Wise saw a color pattern by Julesz on the February 1965 cover of Scientific American and asked to exhibit his work; Julesz invited Noll to join. The gallery's press release says the pictures were executed by an IBM 7094 computer with an SC-4020 microfilm plotter (Noll).

Noll's memoir records that not a single work sold and that AT&T tried to stop the show. Stuart Preston reviewed it in The New York Times: "The wave of the future crashes significantly at the Howard Wise Gallery." Noll's Computer Composition With Lines was based on Piet Mondrian's Composition With Lines of 1917 (Noll).

Authorship was contested from the start. Noll's first attempt to register Gaussian-Quadratic for copyright was refused because a machine had generated the work, and the V&A quotes his account that it was eventually accepted (V&A).

Who shaped generative art after 1965?

Manfred Mohr came to the computer from abstract expressionism after encountering Bense's information aesthetics, and programmed his first computer drawings in 1969. Since 1973 he has built his work on multidimensional hypercubes (Manfred Mohr).

In 1971 the ARC at the Musée d'Art Moderne de la Ville de Paris showed Mohr's Computer Graphics: Une Esthétique Programmée from May 11 to June 6. Mohr's archive describes it as the first one-person show of digital art in a museum. A Benson 1284 plotter and its tape drive were installed in the gallery beside more than 20 framed drawings (Manfred Mohr).

In 1968 the field reached London. Cybernetic Serendipity, curated by Jasia Reichardt at the Institute of Contemporary Arts in London from August 2 to October 20, included Noll and Bell Labs among many others (DAM). Nees published the instructions for his 1965 drawing Axis-parallel Maze in the accompanying book: 4,000 sections of random length, each under 15 millimeters, alternating horizontal and vertical (V&A).

Where does the term "generative art" come from?

The word reached computer art through Stuttgart. Nees wrote a doctorate on generative computer graphics under Max Bense, which DAM calls the first doctoral thesis on the subject anywhere (DAM). The sources differ on the year: DAM gives 1968, the V&A 1969 (V&A). Nake has said that the important word in that title was Bense's invention, and that many people now call this kind of work generative art without referring to the computer at all (DAM).

What was Harold Cohen's AARON?

AARON is an artmaking program that Harold Cohen conceived in the late 1960s at the University of California, San Diego, named in the early 1970s and kept developing until his death in 2016. The Whitney Museum of American Art calls it the earliest artificial intelligence program for artmaking (Whitney).

Cohen, previously an established painter in London, built his own plotters and painting machines to execute AARON's output, and over the decades the program produced drawings, paintings, and imagery for monitors and projections. The Whitney says he understood the work as a collaboration and spent his life translating an artist's knowledge and process into code (Whitney). In 2024 the Whitney, which says it is the only museum to collect versions of AARON from different periods, showed its drawing process live in the galleries for the first time since the 1990s (Whitney).

How did software art change generative art in the 1990s and 2000s?

Artists began building the programming tools themselves. Processing grew directly out of John Maeda's Design By Numbers project, released at the MIT Media Lab in 1999. Ben Fry and Casey Reas started Processing in spring 2001 as graduate students in Maeda's Aesthetics and Computation Group, and in August 2021 they marked twenty years since its first release (Processing).

Processing has always been free and open source (Processing). Casey Reas used software as his own primary medium. His studio's site describes his MicroImage (2001) and Process (2004) series as showing how simple rules and simulated behaviors can generate complex drawings (Reas).

openFrameworks came from the same lineage. In his foreword to the openFrameworks book, Zach Lieberman writes that it began around 2004-05 while he was teaching at Parsons School of Design, as an alternative to ACU, a C++ library from Maeda's group that was tied to late-1990s machines and was not open source (ofBook).

His student Theo Watson joined to build the Mac version, and Lieberman presented the toolkit in beta at a 2006 OFFF festival workshop. Arturo Castro, a participant, later joined to build the Linux version, and in 2008 the project won a prize at Ars Electronica (ofBook). Its authors aimed for parity with Processing so artists could move between the two (openFrameworks).

The browser came next. p5.js, a free JavaScript library in the Processing family built for audio-visual and interactive work on the web, lists Lauren Lee McCarthy as its creator and lead from 2013 to 2020 (p5.js).

When did Art Blocks bring generative art to the blockchain?

Art Blocks launched in November 2020 with Chromie Squiggle by Snowfro, and describes itself as the leading platform for on-chain generative art (Art Blocks). LACMA credits the Squiggle to Erick Calderon, the artist who founded Art Blocks (LACMA).

The artist's algorithm is stored on the Ethereum blockchain. When a collector mints, a unique hash is generated on-chain and fed into the script, and the same hash always produces the same output (Art Blocks).

Calderon told LACMA the Squiggle began as a demo of how much variation a single generative algorithm could produce. Chromie Squiggle #9999 was minted for LACMA's collection on August 21, 2024, as a gift of the artist (LACMA).

What are the key dates in the history of generative art?

What does generative art look like today?

Today generative art is shown running as well as printed: artists publish live code, toolmakers keep building instruments, and museums restage older systems in motion. Leander Herzog keeps his work on his own site as an index of pieces that run as code in the browser, each on its own page, from Tapete in 2007 to works dated 2026. Lieberman now leads the Future Sketches group at the MIT Media Lab, which studies software as a medium for art and design, builds new tools for creative coding, and lists generative form, machine learning and augmented reality among its research areas (MIT Media Lab).

The oldest systems are being shown running, too: the Whitney's AARON exhibition (February 3 to May 19, 2024) included plotters drawing images from different periods of the software in the gallery (Whitney). For a work on a blockchain, Calderon told LACMA, keeping the exact algorithm that recreates the outputs matters more than which chain it lives on (LACMA).

Why is generative art best seen running live?

Generative art is best seen running because the work is the process, and a recording shows only one pass through it. The 1965 press release for Noll and Julesz described reviewing a pattern, refining it and running it again until the result satisfied (Noll). Mohr had a plotter installed in the museum for his 1971 show.

Showing that at home takes a screen that computes. The Layer Canvas renders generative works live on its own processor rather than playing back video, on a square screen with a 59-inch diagonal and 2160 x 2160 resolution. Pre-orders are open in the US at $6,999 plus applicable taxes, with a $99 deposit that is fully refundable until the unit ships, and units are expected to ship in mid-2027. For why a running work differs from a recorded one, see living art.

Sources

1. 3 Standard Stoppages, Marcel Duchamp, The Museum of Modern Art, https://www.moma.org/collection/works/78990
2. Untitled (Collage with Squares Arranged according to the Law of Chance), Jean (Hans) Arp, The Museum of Modern Art, https://www.moma.org/collection/works/37013
3. Vera Molnár: Parler à l'œil, Centre Pompidou, https://www.centrepompidou.fr/fr/programme/agenda/evenement/PA7jRZ5
4. A machine that makes art (Computer Art Image of the Month, January 2012), BCS, The Chartered Institute for IT, https://www.bcs.org/articles-opinion-and-research/computer-art-image-of-the-month-january-2012/
5. Sol LeWitt, Display at Tate Modern, Tate, https://www.tate.org.uk/visit/tate-modern/display/materials-and-objects/sol-lewitt
6. Events Timeline, Digital Art Museum (DAM), https://dam.org/museum/exhibitions_ui/timelines/events/
7. About Georg Nees (talk with Frieder Nake), Digital Art Museum (DAM), https://dam.org/museum/nees-talk-nake/
8. Achsenparalleler irrweg (2), Georg Nees, Victoria and Albert Museum, https://collections.vam.ac.uk/item/O154793/
9. Hommage à Paul Klee, 13/9/65 Nr.2, Frieder Nake, Victoria and Albert Museum, https://collections.vam.ac.uk/item/O211685/
10. Gaussian-Quadratic, A. Michael Noll, Victoria and Albert Museum, https://collections.vam.ac.uk/item/O1193785/
11. The Howard Wise Gallery Show of Computer-Generated Pictures (1965): A 50th-Anniversary Memoir, A. Michael Noll (Leonardo, 2016), Engineering and Technology History Wiki, https://ethw.org/First-Hand:Howard_Wise_Gallery_Show_of_Digital_Art_and_Patterns_%281965%29:_A_50th_Anniversary_Memoir
12. Manfred Mohr, emohr.com, https://www.emohr.com/
13. Computer Graphics, Une Esthétique Programmée, 1971, Manfred Mohr, https://www.emohr.com/paris-1971/index.html
14. Harold Cohen: AARON, Whitney Museum of American Art, https://whitney.org/exhibitions/harold-cohen-aaron
15. Harold Cohen: AARON (press), Whitney Museum of American Art, https://whitney.org/press/harold-cohen-aaron
16. Overview, Processing, https://processing.org/overview/
17. Foreword, Zach Lieberman, ofBook, openFrameworks, https://openframeworks.cc/ofBook/chapters/foreword.html
18. About, openFrameworks, https://openframeworks.cc/about/
19. About, p5.js, https://p5js.org/about/
20. REAS.com, Casey Reas, https://reas.com/
21. Art Blocks Docs, Art Blocks, https://docs.artblocks.io/
22. Erick Calderon's Chromie Squiggle 9,999, LACMA Unframed, https://unframed.lacma.org/2024/08/21/erick-calderon%E2%80%99s-chromie-squiggle-9999
23. Leander Herzog, leanderherzog.ch, https://leanderherzog.ch/
24. Future Sketches, MIT Media Lab, https://www.media.mit.edu/groups/future-sketches/overview/