Algorithms are the exciting new artists-to-collect in the world of fine art. But what makes works by a bunch of bots so appealing? Surprisingly, it’s all about the glitches and the infinite, artistic anomalies that they bring. That’s more or less what AI art is about — at least for those that sell. It’s a real stroke of creative genius.
Mario Klingemann, one of the pioneers who leveraged on AI in art, explains: “A machine enables you to forcefully provoke creativity because it’s much easier to glitch, or bring off course, than a human brain. In the process of doing that, often, some interesting things happen, which
Earlier this year, the German artist sold his video installation, an unending and unexpected stream of AI-generated portraits, at a Sotheby’s auction. It raked in its highest estimate of £40,000 ($66,700). This was the second time a major auction house successfully sold a work of art created by an algorithm.
Memories of Passersby I (2018) was created using a class of algorithms called generative adversarial networks (GANs). A GAN is a package of two networks pitted against each other. The first, called a generator, tries to make images that the second, called a discriminator, will accept. The discriminator’s role is to differentiate between generated images and a data set of images that it has already been fed with.
When it “fails” an image that is passed to it by the generator, it provides feedback from which the generator can learn and make another attempt. But eventually, the generator learns how to fool the discriminator — not by accurately creating images but by creating strange and surreal images that are convincing enough to slip under the discriminator’s radar.
In the case of Memories of Passersby I, Klingemann trained his algorithms using thousands of portraits from the 17th to 19th centuries. He created a Tinder-like application to accelerate the learning process and taught the machine his own aesthetic preferences, influenced by surrealist figures such as Max Ernst. The result? A mash-up of discombobulated faces and features, fascinating enough to pass off as art. At times, the images melded into abstract arrangements of pixels as the machine struggled to create a new portrait.
Paradoxically, there were earlier experiments with GANs by other users who fed it a set of landscape paintings that didn’t go as wrong as the human portraits. However, the relatively convincing landscapes it generated were arguably far less “creative”.
Thus, the idea of AI art done this way, with portraits specifically, is one that underscores spectacle and kitsch. And that’s possibly why Christie’s was able to sell last year’s Portrait of Edmond Belamy — created by Paris-based art collective Obvious using GANs principles — for US$432,500 ($599,300), more than 40 times its original estimate. It also marked the first time an auction house had put an AI-generated artwork under the hammer.
Some call this watershed moment for AI art a frothy piece of marketing but AI-generated art may very well have become the artistic medium to collect, if recent sales are anything to go by. Yet, it’s not all fun and GANs. Some artists are utilising the AI programme to expand their creative limits beyond binary glitches, bringing more humanity into the mix.
Such is the case with Chinese-Canadian artist Sougwen Chung, who trains AI on her own drawings by having the machines transfer what they have learnt about her style to a robotic arm that paints alongside her. This results in a clear mark made by hand and another by machine — a co-creation of sorts, an interplay between her and her robo-version.
“The creative process is about capturing a uniquely human nature. Drawing, for example, is like playing the violin, which is gestural, abstract and improvisational but even more raw. There is an explicitness to it. It captures a sense of time and an emphasis on the hand. Its immediacy, humility and incompleteness are its defining characteristics. Drawing traces the contour of the moment; it strips away excess.
“So in the space between drawing and mark making, both artist and machine can be at their barest as shared movements get stripped to their essential atomic units. It is an expansive reduction; it can reveal unseen qualities in both agents and trace the process of co-evolution,” Chung explains.
Similarly, there’s Austria-based American artist Addie Wagenknecht who programmed a robot with an algorithm to paint a canvas. The only difference is, instead of a “paintbrush duet” like Chung and her machine, Wagenknecht — for her series of mechanically-assisted paintings titled Alone Together (2017) — reclines naked in her programmed robot’s pathway to obstruct it as it navigates and paints around her nude form. Still, what unites these artists is a critical examination of the tech age and a hyperconscious exploration of how to fuse human creativity with new forms of innovation.
Naturally, AI artists attract collectors from a variety of new industries such as science and research, gaming and blockchain, who share the same innovative lens. But pundits aren’t as impressed.
Pulitzer Prize-winning art critic Jerry Saltz has publicly expressed that he finds the work produced by AI artists boring and dull. When reviewing works of art created by AI on HBO, he explained that he was looking for “humanity, dignity, horror, originality”. His conclusion was that even the best efforts of today’s AI “doesn’t come close to art”.
It’s been said that for something to be considered art, all it takes is for someone who identifies as an artist to declare it as such. But to most of us, art is created, not just declared, and it should be delivered with a healthy dose of creative human intervention. While there are plenty of artists who fear that relying on machines will hinder the essence of art and could even lead to humans being replaced as creators, AI artists insist that machine-generated art is a new opportunity. Much like the famous artists of the past whose work was informed by concurrent social and political issues, technology working in conjunction with humans can result in art through comparable experiences.
To Klingemann, the process is becoming an art form in itself. He says: “The tasks of curating the data, selecting the architecture, changing the architecture, training the model — it’s all art. Even choosing neural networks to produce art that looks appealing is part of my art.”
And let’s face it, the AI artworks so far have had a significant amount of human input. So even if AI art is still in its infancy, the potential of the technology appears huge — this raises questions about the future role of humans in an increasingly automated world. Certainly, it is blurring the definitions of creator and creativity.
Even so, Oxford mathematician Marcus du Sautoy argues that we should see the relationship between AI and humans not as adversarial but as collaborative. He believes AI can get creative humans out of their rut and incite them to think in new directions.
He sums it up thus: “All technological achievements so far have pushed us beyond our biological limitations, and it is possible that these developments with AI creativity will only expand our human creativity and push the boundaries of art as a whole.”
Additional AI Art In Action
IBM’s Chef Watson uses AI to help develop recipes and advise humans on food pairings and unique flavours.
Creative technologist Ross Goodwin uses neural networks (a large statistical model that predicts linear sequences) to write poetry, screenplays and literary travel fiction.
I AM AI (2017) was the first album released by Amper, an AI music composer, producer and performer.
British choreographer Wayne McGregor uses an AI-driven machine that can generate its own independent choreography based on hundreds of hours of video footage it has been fed.
This story first appeared in the September 2019 issue of A.