News

Christie’s Sells “AI-Generated” Art for $432,500 As Controversy Swirls Over Creators’ Use of Copied Code

Before the auction, Christie’s refused journalists’ requests for comment on the questionable attribution of the AI-generated artwork. Now, the code’s original creator, a 19-year-old working at Stanford, has publicly voiced concerns that the tech group misrepresented their moneymaking intentions.

“Portrait of Edmond Belamy” (2018) created by GAN (Generative Adversarial Network) and sold for $432,500 on October 25 at Christie’s in New York (image courtesy Christie’s/ © Obvious)

Artificial intelligence (AI) has finally arrived on the auction block, smashing sales expectations by over 40-fold. The computer-generated portrait, which was sold by Christie’s on behalf of a trio of French techies who call themselves Obvious, surpassed original estimates ranging between $7,000–$10,000 to earn an astonishing $432,500 price tag.

Days before the hammer fell, however, reports slowly trickled in that Obvious may have cribbed its code from 19-year-old artist and high school graduate Robbie Barrat.

Within the programming world, Barrat has built a reputation for his exceptional generative adversarial network (GAN) algorithms that train two discriminant networks to effectively act as “artist” and “judge.” The artist builds a system of patterns while the judge checks the result against an original sample; from there, the artist revises his work and the system repeats itself until the GAN has created a passable work of art.

Pictures created using AI GANs have a distinct aesthetic that’s somewhere in-between Surrealism and melted candle wax: borders between distinct objects often get blurred or ultra-shiny. Many of the preexisting AI GAN artworks also derive from Romantic landscape paintings or portraiture, which contributes to the spookiness of the style.

Members of Obvious readily admit that they borrowed elements of Barrat’s code, but says they tweaked the algorithm to produce different results. “If you’re just talking about the code, then there is not a big percentage that has been modified,” says Hugo Caselles-Dupré, the tech lead for Obvious and a machine learning PhD in Paris. “But if you talk about working on the computer, making it work, there is a lot of effort there.”

This exchange is preserved on one of GitHub’s discussion boards where Caselles-Dupré simply goes by the name, “Caselles.” There, he asks Barrat for a version of the code with enhanced resolution capabilities and questions the original programmer about his pre-trained models on the website.

One day before the Christie’s auction of Obvious’s AI-generated artwork, Barrat publicly questioned precisely how much the French group changed his original code on Twitter. The teenage programmer even posted side-by-side comparisons of Obvious’s “Portrait of Edmond Belamy” (2018) and nine of his own works, which show an uncanny resemblance.

Obvious responded on the Twitter thread showing a video demonstrating their AI’s training process as separate from Barrat’s GAN algorithm. The group also shared an April correspondence with Barret, in which they asked the teenage programmer for his blessing to use his code for an undescribed project.

“We do not want you to think that we stole your work or anything like that. We were inspired by your work to create our collection,” Obvious writes.

“Yeah — I’m 100% okay with it; sounds like interesting work and I’m glad that you found some of my code useful. You don’t owe me any credit or anything,” responded Barrat, back in April.

However, Barrat has now voiced concerns that Obvious misrepresented their moneymaking intentions as an open source project of democratization. Because of that, he takes issue with the lack of recognition he’s received as an instrumental component to the “Portrait of Edmond Belamy’s” existence.

Responding to comments on Twitter, Barrat also clarified that he actually did retract his permission to Obvious once he realized that they weren’t doing an open source project, but were instead selling the outputs of its algorithm.

In a statement by Obvious provided to Hyperallergic through Christie’s, the collective said:

We would like to thank the A.I. community, especially to those who have been pioneering the use of this new technology, including Ian Goodfellow, the creator of the GAN algorithm, who inspired the name of the Famille de Belamy series, and artist Robbie Barrat, who has been a great influence for us. It is an exciting moment and our hope is that the spotlight on this sale will bring forward the amazing work that our predecessors and colleagues have been producing. We are grateful to Christie’s for opening up this dialogue in the art community and honored to have been a part of this global conversation about the impact of this new technology in the creation of art.

A Christie’s spokesperson also noted that the collective has given Barrat credit in their Medium post about the project, during a panel with Bloomberg, and in the lot’s cataloging.

Code-sharing is a very common practice in the programming community because it’s seen as a democratic way to expedite the development of technology. Unlike appropriation in the art world, such practices rarely attract controversy. The “Belamy” case is an exception because it raises the difficult question of how much code Obvious should have altered for their program to become sufficiently different from Barrat’s original. And if AI-generated art is going to accrue hundreds of thousands of dollars, then what compensation — if any — is owed to the primary coder?

comments (0)