Abstract: Arthur Samuel (IBM) in 1959 defined Machine Learning as the “…field of study that gives computers the ability to learn without being explicitly programmed.“ . As an example, statistical language translation using convolutional neural networks (CNN) to translate from one language to another by examining many examples of translated texts, where no rules of grammar or word definitions are employed. This same type of learning is now revolutionizing many technologies in our society such as self-driving cars, robotic control, speech recognition, image and pattern identification, video generation, and unsurpassed mastery of games such as chess and GO. Machine Learning is now becoming a highly active research field in geosciences. In this talk I present the mathematics of CNN and overview some of its practical applications in geosciences. These examples include discovery of exoplanets from data recorded by the Kepler space telescope, rock crack and object identification in photos, fault and lithology interpretation in seismic images, and vegetation identification from hyperspectral data.