We know D.C. Get our free newsletter to stay in the know.

Over the past year, Pindar VanArman has spent nearly $12,000 and hundreds of hours in his McLean basement trying to teach Zanelle how to paint. Zanelle’s OK: Her acrylic portraits are accurate, if standard; her more interpretive works might someday find a home in a coffee shop or a hotel lobby. Consider that Zanelle is a robot, however, and her works become considerably more impressive.

“Zanelle” is a common term for art created by a robot. VanArman’s version is “fairly primitive,” he admits. “Nobody’s figured this out yet,” he says. “People have been working on artificial intelligence since the 1930s 1960s, and if we did figure it out, we’d all be replaced, like in the movies.”

Spending more than 10 grand in the pursuit of science fiction was “enough to make my wife pretty upset,” says VanArman. But he has made considerable progress toward his goal: creating a machine that can produce art to rival a human’s. At first, Zanelle functioned “pretty much like a printer,” he says, creating literal reproductions of photographs using paint and brushes. Then, VanArman was able to program Zanelle to “really interpret an image, like any human would do,” he says. Soon, Zanelle was producing increasingly abstract renderings of source images. At her most complex, Zanelle “tries to create completely original work with emotional impact,” says VanArman. “She really sucks at that.”

Initially, VanArman was able to offset the cost of his project by using Zanelle to paint commissioned portraits based on photographs, which he’d sell for about $400 a pop. But soon, “I reached an artistic dilemma,” he says. “People began to send me pictures of their dogs for Zanelle to reproduce. I’m not an art snob, but that was just ungodly.” Zanelle, whose work has been shown in galleries across the country, is “better than that,” says VanArman.

But bills do need to be paid and tweaking ’bots is not cheap. Portraits are still commissioned, but VanArman has one condition: Zanelle gets to interpret the source images however she wants.