In 2018, a portrait painting named Edmond de Belamy made headlines by selling for an astonishing $432,500 at a Christie's auction. This price dramatically exceeded its pre-auction estimates, which had ranged from $7,000 to $10,000. What made this sale particularly noteworthy was not just the final hammer price, but the artist behind the work: a generative adversarial network (GAN), an artificial intelligence system. Created by the Paris-based arts
collective Obvious, Edmond de Belamy became the first artwork generated by artificial intelligence to be featured in a Christie's auction, marking a significant moment in the intersection of technology and the art market.
The painting is part of a larger series called "La Famille de Belamy," or "Belamy's family" in French, which consists of eleven generative images. The name "Belamy" itself is a clever tribute to Ian Goodfellow, the inventor of GANs; "bel ami" translates to "good friend" in French, forming a pun on Goodfellow's name. This particular piece, Edmond de Belamy, is considered a fictional descendant within this AI-generated family tree, with Madame de Belamy being another AI artwork by Obvious. The work is an ink print on canvas, measuring 70 cm by 70 cm (27.5 inches by 27.5 inches), and depicts a somewhat blurry man, primarily focused in the top-left corner, surrounded by whiter color. Its dominant colors are brown and beige, and it is presented in a gold-colored gilded wood frame.
The Auction and Its Surprising Outcome
The journey of Edmond de Belamy to the auction block began when Christie's announced its intention to sell the piece as the first AI-created artwork in its "Prints & Multiples" sale in New York. The painting was initially displayed next to a bronze work by Roy Lichtenstein, highlighting its placement within a traditional art context. Bidding for the piece commenced on October 23, 2018, with five parties participating. Within just six minutes, the price soared to $350,000, already far surpassing the initial estimates. Seven minutes into the bidding, an anonymous phone bidder secured the artwork for $432,500. The final sale occurred on October 25, 2018.
This sale made Edmond de Belamy the second most expensive artwork in that particular auction, only outpriced by an Andy Warhol piece titled "Myths," a 254 cm by 254 cm artwork from 1981 that sold for $780,500. The collective Obvious stated that the proceeds from the sale would be used to refine their algorithm and create works that increasingly appear to have been made by a human being. This outcome not only brought significant financial success to the collective but also ignited widespread discussion about the value and authorship of art created by artificial intelligence.
The Method Behind the Machine
The creation of Edmond de Belamy involved a sophisticated algorithm, specifically a generative adversarial network. The Obvious collective, composed of Hugo Caselles-Dupré, Pierre Fautrel, and Gauthier Vernier, utilized an open-source algorithm originally developed by Robbie Barrat, which was forked from Soumith Chintala's work on GitHub. This algorithm was trained on a vast dataset of 15,000 portraits from the online art encyclopedia WikiArt, encompassing works from the 14th to the 19th centuries.
The GAN system operates with two main components: a generator and a discriminator. The generator creates new images based on the training data, while the discriminator attempts to distinguish between human-made paintings and those produced by the generator. The goal is for the generator to create images that are so convincing that the discriminator cannot tell them apart from human art. The final print of Edmond de Belamy is manually signed at the bottom-right with a mathematical formula, which is part of the loss function metaheuristic algorithm code that produced it, further emphasizing its technological origins. This method highlights a new frontier in art creation, where algorithms learn from historical masterpieces to generate novel works.
