Four legs good, two legs bad? You can rent space on women’s legs in Japan now.

Human drawing on the left, cold dead robotic fingers drew the one on the right. 

Tresset’s robots use computer vision to identify their subjects—they can recognize faces—and then they spend about 30 minutes on each portrait. (One of his earlier-generation robots, Pete, will actually doodle when there are no faces in sight to draw.) The early versions were crude and involved not physical robots but simulated drawing created with computer-aided drafting programs. But over the past 10 years or so, Tresset and Frederic Fol Leymarie, his co-director at the Aikon project at Goldsmiths University of London, have made tremendous progress. Can you tell which image [above] was made by a computer and which was created by Tresset before he lost his inspiration?

Robots face some of the same problems in learning to draw as humans do, Tresset says. “When we draw, the difficulty is not in making the lines. The difficulty is in the perception of the subject and the perception of the drawing in progress.” But sometimes, it may help to make it seem that the robot has difficulty in making the lines—Tresset has found that people feel more empathy for the machines when they make human-esque mistakes like crooked or tilted lines.