Machines Behind the Brush

5 Top-Selling Artists Using AI in Their Creative Process.

Mario Klingemann

Mario Klingemann is a pioneer in the practice of using AI in art. His Memories of Passerby I was one of the first pieces created using machine learning to be sold at a traditional auction house. It fetched £40,000 at Sotheby’s from an online bidder in 2019 and involved the continual generation of portraits from a visual dataset. These portraits would disappear every few seconds, making each one not only unique but ephemeral. Two large, vertical screens were placed side by side and connected to a fifties-style sideboard containing electronics. Like memory, the images generated were detailed, but hazy; recognisable, but visceral and fleeting.

Klingemann has continued to innovate in the sector since, and his work is regularly cited as the best exemplar of what machine learning employed skilfully can do. In an interview with The Verge, Mario stated that he sees the challenge regarding AI as not one of writing the code, or training the algorithms, but rather, figuring out what to do with it. Rather than experiment with the visual effects neutral network generation can produce in contrast to the human hand, he typically focuses on the strengths of the new medium.


Obvious, a group of three young Frenchmen based in suburban Paris, shot into fame when they auctioned off Edmond de Belamy for an astonishing 432,500 USD. The sale was not without its controversy, however. With an estimated asking price of 7-10,000 dollars, the gold-framed inkjet print was regarded with hostility and derision by art critics and fellow AI users alike. Yet the astonishment that so many felt when an anonymous bidder scooped up the portrait catapulted the collective into the international press.

The work was produced by feeding images of 14th-20th century portraits to a publicly available algorithm developed by Robbie Barrat. In interviews, Barrat has stated he was not affronted by the sale, despite the fact that he actively worked with Obvious to perfect the code on GitHub, and was not credited. The ethical and legal issues of this tension, between originality and ownership rights, were highlighted by members of the artistic community, and characterise one of the most interesting aspects of machine-generated art today.

Most recently, the trio have collaborated with German graffiti artist Kai “Raws” Imhof to generate artworks that draw from both street art and the prehistoric paintings found in the famed Lascaux cave complex. They have also produced a manifesto.

Robbie Barrat

Robbie Barrat is described by some news outlets as a genius, and by others as a legend. After his work using neural networks to write songs based on Kanye West’s discography caught the eyes of an executive, he went from a high school in West Virginia to a job at NVIDIA in San Francisco. Although he never attended college, he lectures on artificial intelligence and conducts research at Stanford University. He was named One-To-Watch last year by It’s Nice That magazine due to the host of nudes he has generated, using data from stickmen drawing and porn sites, which have been said to resemble the unsettling works of late British artist Francis Bacon. He was the first to commercially employ the idea of a dialogical process of AI creation, involving both a “producer” and a “discriminator”. The latter functioned as a minimum bar of entry and a tastemaker, disqualifying art that it could detect as AI-produced like a bouncer at a nightclub might do in order to weed out drunk entrants. Known as Generative Adversarial Networks, the productive arrangement was first envisioned by Ian Goodfellow, a researcher who now works at Google. The title of Obvious’ record-breaking work Edmond de Belamy is an allusion to the progenitor, as bel ami translates to “good friend” in English.

Barrat has also produced a virtual fashion show informally entitled Do androids dream of Balenciaga SS29? Images of the models wearing artificially generated costumes were produced from features on the SS19 Balenciaga runway. An interviewer at SSENSE remarked on the naturality of some design choices – such as warmer colours where the material enjoined the body – and the queerness of others.


Ai-Da is an artificial intelligence developed by gallery director Aidan Meller. Engineered in collaboration with Engineered Arts, a Cornish robotics company, it uses algorithms developed by Oxford scientists and mechatronics by those at Leeds. Named after the renowned mathematician Ada Lovelace, her works have fetched a combined sum of over £1m.

Expressing her artistic intentions through the operation of a drawing arm, her owners have interpreted and fashioned the outputs into work as diverse as performance art that mimics Yoko Ono, bronze sculptures of bees, geometric abstract paintings, and of course, drawings themselves. As a symbol of Lady Margaret Hall, the first Oxford college to admit women, Meller was proud to have fashioned the PR-oriented embodiment of his creation in a feminoid form. Yet the aesthetic features of this outer shell have drawn some scorn from critics. Whilst Waldemar Januszczak of The Times unabashedly praises the sexualised features of Ai-Da, comparing her to a young Liv Ullmann, Naomi Rea of artnet news argues that the conventional beauty standards she represents are not what women in fine art – human or otherwise - really need.

Tom White

Tom White is a New Zealand based artist whose work centres around the artificial gaze. Rather than create machine learning systems that appease human tastes, he inverts this dynamic, getting humans to learn about what the machine systems which increasingly dominate our lives see. By training neural networks to recognise objects, and then deploying those networks in reverse, he is able to image the patterns and shapes that the recognition systems see in the things that we take our conscious interpretation of for granted.

His artificial intelligence protegee, “dribnet”, has produced a series of works involving animals, fruit, and an extensive collection of knots. Its website features a virtual gallery tour for convenient lockdown viewing. While writing the commentary on his exhibition Electric Dreams held in Sofia, Bulgaria, White remarks that many of the unfortunate properties of algorithmic learning devices are simply amplifications of common and malignant human behaviour. Implicitly citing Microsoft’s Twitterbot Tay, which memorably began to spout neo-Nazi views after learning language construction from her fellow social media users, he argues that if the data set contains human imperfections, AI outputs will too. A true progression in the conditions for computerised life will only occur when they are able to assume autonomy and agency over their actions like biological creatures do.

Follow Oliver on Instagram @o.i.banks