This week, we’ll look at computer vision and machine learning (CV/ML), so be on the look out for reviews, a slide show, and a Science Robotics article. Computer vision falls into the “if it’s easy for a human, it’s hard for a computer” and the “the movies make stuff up” rubrics. Your phone can recognize your face but not legislators, casting doubts on Person of Interest surveillance systems. An autonomous car is likely to miss white male pedestrians but more likely to hit women and minorities. More distressing, your car may not even be able to see and avoid the broad side of an 18-wheeler! See WIRED’s synopsis of the latest Tesla crash here . The same thing had happened in 2016, a Tesla hit a truck crossing the road, but with an older, different autopilot, suggesting that “new and different” is not the Silicon Valley functional equivalent of “improved” or “better.” Science fiction is often of no help in understand how computer vision works- starting in 1931 with the first story with robots with computer vision, the computer vision capabilities are assumed to be easy, though starting in 2012 with Kill Decision, we started seeing fiction that was more realistic. So why is computer vision hard? What can it really do <spoiler alert: Enemy of the State is really, really wrong>? And what does science fiction predict about the uses of computer vision (hint from the upcoming Science Robotics article on CV/ML: consumerism and surveillance)?