Machine Learning: the Power and Promise of Computers that Learn by Example

In this April 2017 guide, The Royal Society provides a current-day assessment of the discipline of machine learning. 6 chapters, 128 pages.

Machine learning is a branch of artificial intelligence that allows computer systems to learn directly from examples, data, and experience. Through  enabling computers to perform specific tasks intelligently, machine learning systems can carry out complex processes by learning from data, rather than following pre-programmed rules.

Recent years have seen exciting advances in machine learning, which have raised its capabilities across a suite of applications. Increasing data  availability has allowed machine learning systems to be trained on a large pool of examples, while increasing computer processing power has  supported the analytical capabilities of these systems. Within the field itself there have also been algorithmic advances, which have given machine  learning greater power. As a result of these advances, systems which only a few years ago performed at noticeably below-human levels can now outperform humans at some specific tasks.

Many people now interact with systems based on machine learning every day, for example in image recognition systems, such as those used on social  media; voice recognition systems, used by virtual personal assistants; and recommender systems, such as those used by online retailers. As the field  develops further, machine learning shows promise of supporting potentially transformative advances in a range of areas, and the social and economic  opportunities which follow are significant. In healthcare, machine learning is creating systems that can help doctors give more accurate or effective  diagnoses for certain conditions. In transport, it is supporting the development of autonomous vehicles, and helping to make existing transport  networks more efficient. For public services it has the potential to target support more effectively to those in need, or to tailor services to users. And in science, machine learning is helping to make sense of the vast amount of data available to researchers today, offering new insights into biology,  physics, medicine, the social sciences, and more.

Machine learning is a vibrant field of research, with a range of exciting areas for further development across different methods and applications. In  addition to those areas of research that address purely technical questions, there is a collection of specific research questions where progress would  directly address areas of public concern around machine learning, or constraints on its wider use. Support for research in these areas can therefore help ensure continued public confidence in the deployment of machine learning systems. These areas include algorithmic interpretability, robustness, privacy, fairness, inference of causality, human-machine interaction, and security.