AI art, drifting along

Luba Elliott

It has been five years since DeepDream was released by Google, its distinct hallucinatory aesthetic captivating audiences, journalists and creators into this so-called new field of AI art. Of course, under various guises and in different communities, AI-related technologies have been employed by artists such as Harold Cohen and Ernest Edmonds decades prior. This time, the mainstream media met with excitement the colourful “puppyslugs” and the rhetoric of algorithms “dreaming” up such creatures out of ordinary landscapes, oblivious to the history of computer art and fascinated instead by the deep learning revolution. It ushered in a new wave of artists eager to try their hand at whatever new technologies researchers from Google, OpenAI, NVIDIA and other labs may unleash.

Like with many new tools, the spectrum of communities experimenting with deep learning has been broad, each looking to coin, command and conquer a novel application. Research scientists, computational creativity academics, advertising creatives and artists from new media, fine and contemporary art began to engage with deep learning tools to generate images, texts and music in many a niche field. As AI research progressed, it became possible to generate highly realistic human portraits with StyleGAN, produce poems with GPT-3 and make music with Jukebox. Meanwhile, tools like Runway and Wekinator made it increasingly easy for artists to incorporate machine learning into their work without being proficient in Python.

The question that now plagues many AI art aficionados is, what next? Now that the quality of generated images is photorealistic, the attraction of the glitch effect as a creative goal fades. The first GAN models, heralded as a major breakthrough in image generation, produced much better images than alternative methods, yet exhibited difficulties with the accurate placement of facial features and human limbs. What were considered problems in the benchmark-driven AI community became features to artists, who attributed the lopsided eyes and legs sticking out at odd angles as examples of algorithms being creative. With the rapidly improving models no longer able to offer these features, we realise the need to provide something more meaningful to capture the attention of an increasingly AI literate art audience, particularly as much of the low-hanging fruit has been picked. It is no longer enough to badly train a machine learning model on a found dataset online, even if it is a niche one - this has exhausted the minds of curators and journalists alike given the prolific experimentation back in 2016-2017. It is no longer enough for artists to plug in their artwork as data and comment on the algorithm’s uncanny insights regarding their practice. It is no longer enough to give AI full authorship, driving eyeballs and clicks to any AI-related works initiated, managed and curated by a human. 

As the field matures, new actors and new ideas appear. Fine artists engage with non-generative systems in their work, digital artists apply machine learning tools to underexplored mediums and computer scientists strive for both technical excellence and artistic novelty in their work. The AI art gold rush is waning off and that is a good thing: liberated from its frequent function as a PR gimmick for advertising clients and technology companies, AI art can now aspire to be considered on its own merit, competing alongside the full spectrum of contemporary art. 

Luba Elliott is a curator and researcher specialising in artificial intelligence in the creative industries. She is currently working to educate and engage the broader public about the latest developments in creative AI through talks, workshops and exhibitions at venues across the art and technology spectrum including The Photographers’ Gallery, Victoria & Albert Museum, Seoul MediaCity Biennale, Impakt Festival, Leverhulme Centre for the Future of Intelligence, NeurIPS and ICCV. Her online curatorial projects include  aiartonline.com and computervisionart.com. She has advised organisations including The World Economic Forum, Google and City University on the topic and was featured on the BBC, Forbes and The Guardian. Recently, she was part of the Lumen Prize selection committee. Previously, she worked in startups and venture capital and has a degree in Modern Languages from Cambridge University. http://elluba.com/