Research and Development of Future AI Technology

American companies are spending more and more time and money on the study of artificial intelligence for future applications. One such company that is paving the way in AI research and development is Microsoft. Eric Horvitz is a principal researcher at Microsoft Research's Adaptive Systems & Interaction Group. Horvitz is also the president of the AI trade group the American Association for Artificial Intelligence (AAAI).

“I'm working at a modern-day Bell Labs, in probably the largest software research company in the world. IBM may be larger, but they're hardware and software; we're just software. About one-quarter of Microsoft's research is AI research,” Horvitz said, referring to AT&T's legendary Bell Labs research division. He explained that AI has grown so diverse that there are subsets of pure research that use AI in some way. Horvitz estimated that among the 800 researchers employed by Microsoft around the world—in places like Cambridge, Massachusetts; California's Silicon Valley region; Beijing, China; Bangalore, India; and the company's Redmond, Washington, headquarters—about one-quarter of the work done involves AI. Some of those groups include the Natural Language Processing Research Team, the Adaptive Systems Research Team, the Speech Research Team, and the Vision Research Team.

Horvitz is part of Microsoft's Adaptive Systems and Interaction Team. Every day he handles a broad range of research on topics that deal with integrating technology with software to make services more adaptive. One such system uses satellites to help people reach their destinations. Horvitz explained, “For the last two years, we've provided volunteers with fifty global positioning systems, for two weeks each, to use as they drove around Seattle. We then amassed data on hundreds of trips in Seattle, so we've created something called Pre-Destination. The idea is to predict where they are driving just as soon as they start moving their car. With that knowledge, we can layer in a number of services for their dashboard screen such as directions, advertising, and other opportunistic placements. We also have something called Shortstop where we can predict how long a stop will be based on stored criteria that can be applied in a smart way. We can determine, based on the length of the stop, whether or not there's time to deliver messages.”