Algorithms usually need thousands of examples to learn something. Researchers at Google DeepMind found a way around that.

"Making AI Smarter, Faster" is from the January/February 2017 issue of Technology Review.