Artsy, the online art database, has an ambitious vision: A world in which art is as popular as music. It has pursued that goal since its founding in 2009—I heard the mantra repeated at least twice while visiting the company’s lower Manhattan office—and recently took a big step toward achieving it when it bought the data science startup ArtAdvisor.
Art and data science make for an unlikely pairing, but data is especially well suited to toppling a key barrier to making art as ubiquitous as music: taste.* *“More sophisticated taste just comes from spending more time with something, just like music or wine or film,” says Artsy CEO Carter Cleveland. “You start with what catches your interest, and then you spend more time with it, and then you get bored with the things you’ve heard or tasted ten times, so you look for those deeper levels.” In other words: Taste is simply a function of information. The more you have, the more finely tuned your sensibility. Artsy intends to expand your palate by turning the art world into a series of data points, to reveal what you might like.
Exactly how this will work remains to be seen, but it expands on Artsy’s. That initiative, run by six art historians, involves assigning artists and their work “genes,” or traits. A Picasso painting from his Cubist period, for example, gets the Cubism gene applied at 100 percent. But a Mondrian that merely draws on Cubism might be assigned a value of 30 percent. The Art Genome Project creates a constellation of artists—27,000 to date—that fuels Artsy’s recommendations to its users. If you like Picasso, the genome might suggest that you’d also appreciate Tatiana Berg, a young Brooklyn artist who paints in big, gestural brushstrokes.
With ArtAdvisor in house, Artsy aspires to go beyond connecting styles, periods, and movements. It wants to create what it calls an art brain—a system that understands the whims and nuances of the art market much like an insider does. Cleveland uses a scenario to explain: A gallery shows a promising young painter. A MoMA curator suggests he might want to buy two pieces. The gallery owner mentions this to a collector, who now has a better sense of what the work might be worth. “That’s such a common thing you’d hear,” Cleveland says. “A lot of what we’re doing is figuring out how we capture that information in a structured way.”
Hugo Liu, ArtAdvisor’s founder and Artsy’s new chief scientist, plays a central role in this. The MIT-trained AI researcher built ArtAdvisor to see if machine learning could paint a more holistic picture of art valuation. “It’s not easy to find out why certain things are worth certain amounts,” he says. “It’s more than just auction results and prices, there’s a ton of rich cultural signals.”__ __Some of that has to do with debunking assumed truths. If one renowned gallery has a reputation for vaulting artists to fame, but a smaller one tends to represent artists who go on to enjoy longer careers, an algorithm could more accurately determine which gallery produces more culturally significant artists.
To understand those signs, Artsy already has a few information inputs in place. The Art Genome Project, for one. Many of the 4,000 galleries that use Artsy’s platform also supply the company with data. And Artsy can watch users’ surfing and buying patterns. Next up: Liu will design algorithms that can triangulate existing data points, and suss out other so-called cultural signals.
How useful you find this depends upon whether you aspire to become an art admirer or an art collector. If you simply want to develop an appreciation for fine art, you’re in luck. “Computers are infinitely patient,” says data engineer Glenn McDonald, who works on Spotify’s playlists. Algorithms can sift through millions of things to choose the best one to present to you at any given moment. If you intend to buy, leaning on algorithms becomes trickier. Every year, people ask McDonald’s team to predict the summer’s hit song. “Can you apply machine learning to find the next hits? I think no,” he says. Summer hits, he says, tend to sound more upbeat. “You could write an academic paper on it. But you can’t make your song faster and have it be a hit.” McDonald also points out that a platform like Spotify has hundreds of millions of users making decisions each day; the art world has less data to consider.
Artsy knows this. For some works—Cleveland mentions the hundreds of David Hockney watercolor prints that have sold for consistent prices in auctions—Artsy can make clear predictions. “Those are like your table stakes,” he says. “Then there’s completely new works by new artists where you have zero data.” But as more people engage with the art market, whether through auction or Instagram, those data points will emerge.
Artsy also insists that taste in art does not require purchasing power. It comes with having knowledge, opinions, and some anticipation for what will arrive on the scene next. To have that, art neophytes need access to more layered information beyond major museum retrospectives. If the works from blue-chip artists are your Top 40 radio hits, Artsy wants to acquaint you with obscure B-sides. Knowing them, following them, and maybe even loving them, the way you would a musician—that’s taste. To do that Cleveland and Liu will need to create a web of data that’s a lot like art itself: nuanced and open to interpretation.