Machine learning is the study of algorithms that infer patterns from data, typically with the express goal of making good predictions about future data. Classic examples of such techniques are neural networks and support vector machines, which have seen a wide impact far beyond their humble origins; indeed, the last decade has seen a flurry of applications of machine learning techniques to areas ranging from e-commerce (Netflix's video recommendations) to environmental science (MAXENT for species distribution modelling) to game playing (Google’s champion-beating Go program). Despite these successes, it is not always obvious how to apply machine learning to a new problem. In part, this is because to the uninitiated, it can appear to be a black box comprising an overwhelming number of techniques and terms. In this talk, we open this box and de-mystify some of these techniques. We introduce some simple ideas -- matrix factorisation, logistic regression, and neural networks -- and show how they underpin many recent advances in machine learning. We illustrate some of these advances via a series of vignettes that illustrate their applications to real-world problems, including: - determining what makes YouTube videos popular (via self-exciting stochastic processes) - recommending trips to travellers in a city (via structured prediction) - generating realistic images of cats, dogs and more! (via adversarial neural networks) Our aim is to stimulate discussion and generate ideas as to how machine learning could be applied in the astronomical sciences. In presenting these applications, care will thus be taken to emphasise the abstract problems underpinning each of them.