Remember Google Arts & Culture’s Face Match? Back in 2018, Google’s Arts & Culture app—which primarily exists to give users access to a ton of famous works of art at their fingertips—briefly went viral thanks to a particular feature that allowed users to upload selfies, which the app would then match with existing artwork. For about a week, you couldn’t escape people posting side-by-side photos of themselves next to Google’s chosen artwork match. Sometimes the matches were eerily accurate, and sometimes they were hilariously off-base.
Recently, the next logical step in this kind of AI-based art appreciation appeared online in the form of AI Gahaku. Selfies are also involved with this particular app, but this one’s not just about matching your face with existing works of art. AI Gahaku is, in fact, an AI artist who will create Renaissance-style paintings based on the selfies you submit.
Below, for instance, is an artwork generated by AI Gahaku based on a selfie taken by the author of this article a few months ago:
AI Gahaku is the creation of a Japanese full-stack developer named Sato. He has come up with ten different styles to choose from when you upload your selfie and the software that analyzes the photo. All of the filters look like variations of portraits by Leonardo da Vinci, Michelangelo, Albrecht Dürer, and other such famous Renaissance-era artists.
AI Gahaku might seem like little more than a classier version of, say, Instagram, which has a variety of filters to give your run-of-the-mill smartphone pictures a more “artistic” look. But AI art of the kind AI Gahaku traffics in isn’t exactly a new phenomenon. In fact, AI art has been part of the art world for the past few years now, albeit not to quite the degree of mainstream acceptance that Sato’s creation has achieved in the past week or so.
Perhaps the most notorious example of AI art that stirred the art world came in 2018 with “Portrait of Edmond de Belamy.” The “Portrait” was an artwork created not by a single artist, but by a French collective named Obvious, whose members didn’t even create the algorithm that generated the work but instead altered an existing developer’s code to come up with the Renaissance-like portrait. The piece not only ended up at a Christie’s auction but also sold for $432,500, far surpassing any appraiser’s pre-auction estimates.
An even more ambitious—not to mention historic—example of AI art came the following year with Faceless Portraits Transcending Time, an exhibit at the HG Contemporary gallery in New York’s Chelsea neighborhood. If “Portrait of Edmond de Belamy” was interesting more for the way it approximated the Renaissance style in its machine-generated canvas, the portraits created by the artificial intelligence AICAN marked an attempt by its creator, Rutgers University computer science professor Dr. Ahmed Elgammal, to try to use AI to create stylistically new art. The results, with their blurred faces and general tilt toward abstraction, look both retro and modern in equal measure: a mix of Renaissance, impressionist, and abstract expressionist styles all presented with a palpable digital sheen.
Even if one doesn’t find the end products of these pieces of artificial intelligence-generated art all that interesting, Dr. Elgammal suggests that it’s not so much about the result as it is about how the result came about. “It’s about the creative process,” Dr. Elgammal wrote in the January-February 2019 issue of the magazine American Scientist, “one that involves an artist and a machine collaborating to explore new visual forms in revolutionary ways. For this reason, I have no doubt that AI-produced pieces are conceptual art, a form that dates back to the 1960s, in which the idea behind the work and the process is more important than the outcome.”
There’s certainly precedence for this kind of thinking in other forms of arts & culture. The pioneering 20th-century avant-garde composer John Cage, for instance, spent much of his career experimenting with the idea of “aleatory music,” or “chance music,” in which the music was dependent not on a fully notated score, but on whatever sonic elements were available at a given moment. (That’s the whole idea behind Cage’s most famous experiment in aleatory music, 4’33”—essentially 4 minutes and 33 seconds of silence where whatever sounds popped up during that time were the “music.”) AI art of the kind pioneered by Obvious and AICAN appear to be similarly aleatory in nature, according to Dr. Elgammal.
Still, as AI Gahaku suggests, the results do matter to some degree, especially in a culture as goal-oriented as America’s generally is. Why else would it offer, towards the bottom of its website, an opportunity for users to create their own custom title and frame, allowing all of us to “frame” this algorithm-generated artwork as if it was ready to be exhibited in a museum? The process might be fun, but that process isn’t necessarily going to appear on someone’s Instagram page, ready for all the likes and comments the end product will encourage.
In any case, even aforementioned Google’s Arts & Culture app seems to get in on digital artistic creation with its recently unveiled Art Transfer tool, which, like AI Gahaku, can turn your photos into old-style paintings. Perhaps art created by artificial intelligence is here to stay after all.