What is machine learning?

Traditional software follows rules written by programmers, while machine learning systems discover the rules themselves by studying examples. In supervised learning, a model is shown labelled data, such as photos tagged cat or dog, and learns to label new ones. Unsupervised learning finds hidden structure in unlabelled data, and reinforcement learning trains systems through trial, error and rewards. The term was popularised by IBM researcher Arthur Samuel in 1959, who built a checkers program that improved by playing itself. Machine learning now powers spam filters, voice assistants, fraud detection and medical image analysis.