Researchers at Carnegie Mellon University are working on a system called NEIL, or, Never Ending Image Learning, to recreate common sense in a computer. They are attempting to teach NEIL to think for himself by analyzing and identifying shapes and colors in pictures. It is also slowly discovering connections between objects on its own. For example, NEIL figured out that zebras tend to be found in savannahs. But it did get a couple wrong. Like “rhinos can be a kind of antelope” and “actors can be found in a jail cell.” Well, I wouldn’t say that one is necessarily wrong. What’s next? NEIL will analyze vast numbers of YouTube videos to look for connection between objects.
[ Read the Article: Computer Teaches Itself Common Sense By Analyzing Web Images ]