Researchers at the New York University have developed a new algorithm that enables machines to learn without having to be fed thousands of examples. In fact, after only one example, the algorithm was able to provide almost identical to those provided by humans.