Locating the artist in AI artworks

The generative art group exhibition, In Our Code, is on display at Unit London between 13th September – 16th October 2022. Within the exhibition are two AI-generated works by Sofia Crespo and Helena Sarin. This article explores the creative process and public perception around AI-generated art, as well as how to understand authorship in AI art.

This summer, scandal befell the Colorado State Fair. The cause? Jason M. Allen’s Théâtre d’Opéra Spatial, and its winning place in the digital fine art competition.

At first glance, there is little about this artwork that could be deemed controversial. Three figures bathed in light stand gazing through a circular viewpoint at a strange, fantastical world. It is only when you reach the submission title that the source of controversy is revealed – the piece is attributed not to Jason M. Allen but to Jason M. Allen ‘via Midjourney.’

Jason M. Allen, Théâtre d’Opéra Spatial, 2022. Courtesy to the artist.

What is AI Art and how does it work?

Midjourney is an AI text-to-image software. Creating art on it might, at the outset, feel like making a Google Search. You enter a text prompt describing an image you’d like to see, and the software generates four pieces that match this description. But, Midjourney is not like Google Search. The images are not coughed up or compiled from a pre-existing database. They do not originate from beyond the computer, with a painter or photographer. Instead, the images are generated from scratch by Midjourney itself and are powered by the software’s machine-learnt understanding of how to visually represent linguistic terms.

As the news spread that an AI-generated piece had won an art award, the internet exploded with outrage. Critics insisted that where there was AI, there could be no artist. ‘This is the literal definition of “pressed a few buttons to make a digital art piece”’ read one Tweet. ‘‘If creative jobs aren’t safe from machines, then even high-skilled jobs are in danger of becoming obsolete. What will we have then?’ read another. The sense that AI renders human creativity obsolete was acute.

Sofia Crespo, essential_protozoa_1862, 2022. 

What happens when new technology enters the art world? 

Such fears surface every time a new technology makes its way into the artist’s studio. As experimentation with the photographic camera spread in the 1850s, Baudelaire famously lamented that it would become ‘the refuge of every would-be painter, every painter too ill-endowed or too lazy to complete his studies,’ and much of the art world insisted that photography was merely pressing buttons.

But just as photography is far more than the click of the shutter, AI text-to-image generation is far more than a text prompt. During an interview with the Washington Post, Jason explained that Théâtre D’opéra Spatial took him over 80 hours to create. The piece began with a simple prompt and mental image: ‘a woman in a Victorian frilly dress, wearing a space helmet,’ but as Jason pondered the pieces Midjourney generated, an epic dreamlike vision began to emerge.

It would take over 900 Midjourney iterations, for Jason to produce Théâtre d’Opéra Spatial, and it would take the use of two additional programs: Adobe Photoshop, which Jason used to add and remove elements; and Gigapixel AI, which he used to increase the sharpness and quality of Théâtre d’Opéra Spatial for the image to be competition ready.

As with photography, when the creative processes behind AI-generated artworks are revealed, it becomes clear that, contrary to initial impressions, the artist is indeed an artist, but unlike with the camera, it also grows increasingly difficult to see AI as an inanimate tool, and questions of whether to credit AI as a co-author in its own right often arise.  

Anne Spalter, Unconstructed Dream Space #2 (detail), 2022. Courtesy to the artist.

Is the AI a co-author? 

The works of AI artist Anne Spalter directly address this conundrum. Her piece Unconstructed Dream Space #2 which was recently auctioned as part of Phillips’ exhibition Ex Machina: A History of Generative Art is a 25-second kaleidoscopic video that depicts a small figure staring out at a vast and twinkling galaxy. This galaxy’s horizon grows dominated by a face that while initially dreaming later adorns a space helmet and embarks on intergalactic travel. Anne constructed the piece using a custom text-to-image AI pipeline that incorporated the phrase ‘unconstructed dream space’ and during a Twitter Spaces with Phillips, she described the software as doing ‘such amazing things, that it really is like working with another artist in your studio. It’s so hard to predict and so unknown in its processes.’

Perhaps the line between machine and person blurs when using AI text-to-image software because by drawing from a database of words and images that is continually learning from the updates and inputs of thousands of other humans, the machine is, as Anne says, generating images from the collective consciousness. The artistic process then can become a form of ‘collective culture dreaming’ that makes Anne’s use of space exploration in Unconstructed Dream Space #2 as a metaphor for the creative journey feel particularly apt.

So, how to credit these softwares and the many thousands of people whose inputs have helped shape the artworks they produce? It’s not a question with a straightforward answer, but increased familiarity has already helped us make sense of the role of the AI artist, and, before that, of the photographer. There may be a day when it helps us make sense of the role of the AI itself. 

Helena Sarin, Blue Stratagem: Improvising on the Keyboard of Manifolds, 2022. 

By Kitty Horlick