Artificial Intelligence & Art
French Art Collective Sold AI-Generated Portrait for $432K, Raising Questions from Creative Coding Community
On Oct. 25, 2018, Cristie’s New York auction house sold what was promoted as the first piece of fine art to be created by artificial intelligence (AI).
Art critics estimated that the grainy portrait of a fictitious 18th century character donning a white blouse and black robes would sell for $10,000 at most for its novelty. What no one foresaw is that the “Edmond de Belamy” print would sell for $432,000, raising questions from the AI community, who identified the algorithm used to generate this painting as that belonging to a West Virginia teen called Robbie Barrat.
The 70x70 cm print is one of 11 portraits that three French students, who go by the name of Obvious, created using a type of algorithm called generative adversarial network (GAN). The painting’s title is a play on bel ami, French for “good friend,” and a reference to Ian Goodfellow, a machine learning researcher who first came up with GAN as a tool for creating art in 2014.
Obvious founders Pierre Fautrel, Gauthier Vernier, and Hugo Caselles-Dupré explained through a series of self-published articles that their goal is to introduce AI-generated art in a way that is accessible to the general public. Nevertheless, the young entrepreneurs received flak from the AI community for presenting the concept in a way that was so simple it made it seem as if a computer conjured the images by its own free will.
“They used [Barrat’s] scraper, that’s the code he used to pull images from WikiArt, and essentially replicated his project,” said Jason Bailey, digital art blogger and founder of Artnome, over the phone. “There is some debate over whether they trained their own networks using the same code but the work is still derivative.”
When Bailey first saw Barrat’s work six months ago, he was stunned. “He’s a creative genius,” said Bailey, commenting on Barrat’s Nude Portraits series. “I was able to position his work within a rich aesthetic of art history, but it was new and fresh like nothing I’d seen before in digital and traditional art,” he said.
Barrat’s work isn’t as simple as layering a stylistic effect on a picture. Creating images using GAN includes feeding a network a dataset, such as historical paintings, to develop replicas while a second network judges whether the generated visuals match the original data and rejects those it deems as fake. The process includes constructing algorithms and sifting through hundreds of generated images to select the best samples.
One of the AI creators who was quick to recognize the similarities between Obvious and Barrat’s work is Tom White, a computational design lecturer at the University of Wellington in New Zealand. White downloaded Barrat’s original code and generated a series of paintings that could very well be Belamy’s long lost brothers.
Barrat took to social media after the sale, posting pictures of the work he created over a year ago with a caption that said, “Does anyone else care about this? Am I crazy for thinking that they really just used my network and are selling the results?”
White does not put all the blame on Obvious for concealing the true nature of work. “Cristie’s twisted the narrative,” he said. “They didn’t lie about it but they worded the sale in a way that they knew would grab people’s attention.”
In an article published by Cristie’s after the sale, they state that the portrait “is not the product of a human mind. It was created by an artificial intelligence.” The article also gives all the credit to Obvious and implies that GAN is a process created by the three French students.
This version of the story was echoed across many media outlets, which led to a flawed understanding of machine learning prior to the auction; and undermined generations of digital creators, said Bailey.
“When a really large institution tries to do something cutting edge, they don’t always get it right,” he said. “But then again there is nothing in traditional art that would prepare you to understand this new breed of art appreciators and creatives.”
This story was written in November, 2018. For any corrections please contact me directly.
