After training, the system can look at new pictures of cereals it's never seen before and correctly identify them with over 90% accuracy. While this example uses cereals, the same process could be used to teach the computer to recognize almost anything visual, making it useful for many different businesses and applications.