Generative adversarial networks (GANs) and other recent neural network architectures are bringing machine creativity closer to reality. Artificially generated images, videos, sounds and texts will change the way people experience the world, make decisions and create novel work.

We’re among the first to bring these new kinds of models to the industry, and are pioneering their deployment into real-world applications.

After enabling breakthroughs in supervised machine translation and information extraction, machine learning algorithms are now tackling higher-level reasoning and manipulation of abstract concepts using human language, just like people do. In the coming years, machines will assist our thought process and make our conversations more productive and creative.

We’re already testing out some of these advances in live situations, where businesses must provide personalized support to their customers at scale.

Reinforcement learning is one of the next frontiers of computer science. Beating the Go world champion, although unthinkable a few years ago, is only the tip of the iceberg. To make those algorithms practical, we’ll need to discover how to make them work in uncertain, fast-changing environments and improve their efficiency.

Although much of the current attention is focused on games, we’re already investigating how highly complex automated decision-making will help us save energy and manage resources more efficiently.