With his art-making software dating back to the 1960s, the programmer and artist paved the way for today’s confluence of art and technology

Brian Droitcour

Understanding the legacy of Harold Cohen, the world’s first AI artist

Is artificial intelligence a tool to be used by humans to creative ends or is it a creative agent in its own right? The late artist and programmer Harold Cohen considered these questions in his 1973 essay ‘Parallel to Perception’ and articulated a position shared by many artists working with AI today: ‘Tools generally serve to extend or to delimit various human functions,’ he wrote, ‘but of all the many tools invented by man only the computer has the power to perform functions which parallel those of the mind itself, and its autonomy is thus not entirely illusory.’ In other words, the computer can play a more independent role in the creative process than, say, a paintbrush or a camera.
Cohen proved this point artistically with his best-known project ‘AARON’, an art-making software he conceived in the late 1960s and developed until his death in 2016. As explored in ‘Harold Cohen: AARON’, a new exhibition at the Whitney Museum of American Art in New York, Cohen designed AARON specifically so that the software, and thus the machine running it, could make decisions with as little human input as possible. Today, AARON is known as the earliest example of an art-making AI. Reflecting on AARON’s relative simplicity, in comparison to the AI models used by contemporary artists, can show us how far we have come, but also what has remained constant.

When it comes to generative or AI art, the question of the machine’s autonomy has always been central. Cohen wrote AARON to mimic the decision-making process artists follow intuitively, linking lines to shapes to figures, to produce results that look more organic and spontaneous than the geometric abstractions that dominated computer art in the 1960s and 1970s (and are still common in contemporary generative art). Over time, AARON became increasingly complex, yet the code was still the work of just one person. In contrast, image-generating AI tools today rely on visual data, made by thousands of anonymized creators, scraped from the internet; this data is then connected to users’ prompts through databases of tags compiled through vast human effort. Perhaps in response to these conditions, many contemporary artists working with AI seek out collaboration and collective decision-making with technology, rather than framing their work as a solitary effort the way Cohen did.

Mario Klingemann, a pioneer of AI art, used to maintain ‘Sketchmaker’ (2006-2014), an autonomous drawing program that produced brushy abstractions, often with mirrored patterns. When AI became more advanced, he created ‘Botto’ (2021-), an artificial artist based on the open-source neural networks VQGAN and CLIP. Designed to learn and make decisions on its own, Botto’s style has varied widely, from the grainy, wavy figuration characteristic of early AI art to pastiches of Modernist paintings to HD renderings of quasi-Futurist sculptures of liquid metal, as seen in last year’s ‘Artificial Idol’ series. 
Botto’s range of styles is the direct result of a decentralized autonomous organization (DAO), a blockchain-based voting structure, established by Klingemann. Here, collectors of Botto’s works can vote on the outputs they like the best, influencing the direction Botto takes in subsequent works and determining which pieces will be released for sale as NFTs. Klingemann acknowledges that this market-oriented process means Botto’s weirder outputs are often discarded and the overall quality of the work veers toward mediocrity. But this is a conceptual interest for him, to think about how AI both shapes and imitates human tastes.

Artist Mat Dryhurst and musician Holly Herndon also developed a DAO for ‘Holly+’, an AI sound model based on Herndon’s voice. Before releasing it for public use in 2021, the voice model was used to produce Herndon’s 2019 album PROTO, one of the first bodies of music written with neural networks. Now, other artists who use the model can submit the works they create to the Holly+ DAO. If the members of the DAO vote to authorize a work, it is minted as an NFT and sold, with half the proceeds going to the artist and the rest to the DAO and Herndon.
Where these artists follow in Cohen’s footsteps is through the personification of software. By naming it AARON, Cohen lent some biblical heft to the project, as if the program was the conduit for his own prophetic art. When PROTO was released, Herndon referred to the voice model as an ‘AI baby’ and called it ‘Spawn’ before adjusting the nomenclature to Holly+, suggesting it is an augmentation of her creative self. Botto, meanwhile, is a portmanteau of ‘robot’ and ‘Giotto,’ as if this artificial artist is the harbinger of an AI renaissance. Personification positions the software as an agent; naming endows it with characteristics that shape the audience’s perception of its outputs as well as the creative relationship with its human counterparts.

‘DOUG’, which stands for Drawing Operations Unit Generation X, is the simple, friendly name of the program that artist Sougwen Chung has been developing since 2015. It is a neural network that mimics the sinuous shapes and lines that Chung’s own hand draws, sometimes identifying novel patterns in the artist's style for them to explore further. In live performances, Chung moves across large sheets of paper in tandem with robots, drawing together in an improvised dance. Their interaction with and presentation of DOUG as a robot drawing live before an audience dramatizes the collaborative aspect of her practice while also echoing Cohen’s work: AARON was often exhibited in live drawing performances, where part of the spectacle was the suspense of watching the machine pause, as if the program was ‘thinking’ about what to draw next.
A reassessment of Cohen’s legacy today, particularly through the lens of artists extending the path he paved, involves revisiting what is – or can be – meant by ‘art and technology.’ In midcentury manifestations, this commonly referred to a team of scientists and engineers working to realize an artist’s vision with their special expertise. But this process involved a substantial investment of resources by labs and corporations, so it gradually became more common for ‘art and tech’ to refer to artists who mastered specialized equipment and tools to pursue projects independently. This was the case with AARON, and that independence was a source of pride for Cohen.

Today, with AI and computing tools widely available, ‘art and tech’ can mean using consumer-grade hardware and software for conceptual ends. However, the most interesting work being made with AI is often a hybrid of these understandings. Klingemann, for example, often works with specialists to customize AI models that were originally developed by research labs and corporations, but he also challenges the expectations of how these tools enter popular use through the activation of open-source networks. If Cohen sought out the limits of what computers could do, today’s AI artists do the same while simultaneously thinking about the ends of these actions, imagining what tech and its applications can look like in alternative economic and social structures.
<hr>
‘Harold Cohen: AARON’ is on view at the Whitney Museum of American Art from February 3 through May 2024.
Brian Droitcour is editor-in-chief of Outland, an online magazine about digital art.
Published on January 25, 2024. 
Caption for full-bleed image: Harold Cohen, AARON Gijon, 2007. Screenshot. Artificial intelligence software. Whitney Museum of American Art. © Harold Cohen Trust.&nbsp;A dark filter has been applied over the image for readability.

title_style_2

‘The Road to Art Basel’, Part 3: In our series spotlighting the behind-the-scenes world of the four fairs, two artists tell us about their upcoming participation in Art Basel Hong Kong, and the work it takes to get there

Stories

From the studio to the fair floor: what artists do to show at Art Basel

For co-founder Sandra Terdjman and Julia Morandeira Arrizabalaga, the director of KADIST’s Paris space, the relationship between patronage and market is ripe for reinvention

Discover a truly unique art foundation: KADIST

The Singapore-born, UK–based artist pieces together ceramics, installations, and animations with intuition and emotion

The many whims of Kara Chin

With his art-making software dating back to the 1960s, the programmer and artist paved the way for today’s confluence of art and technology

Welcome,

DISCOVER MORE RELATED CONTENT BELOW: