Introduction to the Special Section

Machine Learning (ML) is exploding in popularity. What little more than a decade ago was the realm of specialists is now becoming a commonplace tool in the development of technology products. For example, as of January 2019, Amazon has sold more than 100 million devices running its Alexa software—whose speech synthesis, recognition, and dialogue systems are all built with ML [Bohn 2019]. Alongside this growth in the application of ML, universities are investing massive resources in hiring faculty who are experts in artificial intelligence, ML, or data science, with approximately 35% of recent faculty searches in top 100 Ph.D.-granting computing departments targeting such faculty [Wills 2019]. These faculty, in turn, are teaching an ever-larger number of ML courses, thereby preparing a wave of software engineers to use ML in their work. These in-person university courses are complemented by online ML education offerings from Coursera, Udacity, EdX, and others that tout enrollments in the millions. ML-focused industry job advertisements are also are rapidly growing in number. The number of industry job postings mentioning ML or artificial intelligence on Indeed rose about 30% between 2018 and 2019, after doubleand triple-digit growth in the preceding 2 years [Indeed Editorial Team 2019]. This rapid growth in ML-based products, ML faculty positions, and ML engineering jobs is propelled by one basic phenomenon: ML is capable of leveraging ever-increasing amounts of data to inform computational tasks such as decision making, reasoning, and even design and creation of new systems. This means that systems created with ML can be far better than those created with traditional programming approaches at capturing the nuances of many phenomena that matter to people: human bodies, human communication, human societies, and the natural and human-built physical world. Some examples illustrate this point: