A unique work of art is up for sale at the famed Christie’s auction house this week. Titled the "Portrait of Edmond de Belamy", the piece’s color palette and its subject’s attire suggest a creation from the 18th century, whereas its unrefined nature, muddled facial features, and areas of blank ‘canvas’ could be attributed to a troubled artist in the throes of a tragic emotional breakdown. However, the signature in the bottom-right corner – a neatly printed algorithm – alludes that the portrait is actually an image produced by an artificial intelligence (AI) program that was trained to mimic classic human art.
Sounds like the perfect start to a discerning 21st-century art lover’s private collection.
The humans behind the portrait – because there are no AI entities freely producing art without being programmed to do so, at least, not yet – are a French collective called Obvious. Based in Paris, Obvious seeks to explore questions about the increasing prevalence of AI and machine learning in modern life, using art as a lens.
“One of our goals is to explain and democratize these advances through our artworks,” the three members state on their website.
The Portrait of Edmond de Belamy, one of a series of “painted portraits” of the nonexistent de Belamy family, came into being after Obvious began tinkering with a particular type of AI algorithm called a "generative adversarial network", or GAN.
“The algorithm is composed of two parts," Obvious member Hugo Caselles-Dupré explained to Christie’s. “On one side is the Generator, on the other the Discriminator. We fed the system with a data set of 15,000 portraits painted between the 14th century to the 20th. The Generator makes a new image based on the set, then the Discriminator tries to spot the difference between a human-made image and one created by the Generator. The aim is to fool the Discriminator into thinking that the new images are real-life portraits. Then we have a result.”
Each time the Obvious GAN produced one of these passing images, Caselles-Dupré and his collaborators, Pierre Fautrel and Gauthier Vernier, used an inkjet printer to transfer them onto canvas and framed them. A few in the series are currently on exhibit in France, and several have already been purchased, according to the Obvious website. But Christie’s notes that their offering of the fictional likeness of Edmond, on the block (online) from October 23 to 25, represents the first AI-created art to be sold on the world auction stage. It is estimated to fetch a minimum of $7,000 to $10,000.
Though the humans behind Obvious will receive the profit and own the copyright, issues of ownership are one of the very puzzling concepts that Caselles-Dupré, Fautrel, and Vernier want us to consider.
“If the artist is the one that creates the image, then that would be the machine,” Caselles-Dupré said to Christie’s. “If the artist is the one that holds the vision and wants to share the message, then that would be us.”
As it continues to learn how to trick itself, the collective’s GAN will produce more and more realistic-looking portraits. The road to creating examples that stump people may take some time, however, as our brains have evolved over millennia to quickly recognize and analyze human facial features. This creative challenge is precisely why Obvious chose portraiture over something easier to dupe, like landscapes or nudes.