Why didn't the robot cross the road? Because he couldn't see.

In the economist's Babbage podcast from October 23 (their first), there was a discussion on computer vision. They mention that the reason we don't see more robots in our everyday lives is that they really aren't able to see where they are going. While they can be trained to interact with specific types of objects, developing an algorithm for computer vision has its own challenges. They mention that the most successful computer vision algorithms out there begin with edge detection and try to build up to larger objects.

There is an enormous community working on the problem of computer vision, so it seems like the Economist's coverage cannot give a very complete overview of the state of research in this area. What I think they can do is talk about some of the most complicated tasks that computer vision can and cannot accomplish.