Machine learning methods are inspired by neuroscientific discoveries—but current methods are fragile, do not generalize well, and require massive amounts of data. We use visual computing methods to better understand the similarities and differences between human and machine intelligence.