"One of the things we're trying to do is get to the next level in distributed systems," said Richard Rashid, vice president of Microsoft Research, which is the official name of Microsoft's laboratory operations. Other thrusts are in data mining, language recognition and computer vision.

As with most of Microsoft Research's personnel, Rashid was a recognized computer authority long before he headed to the Pacific Northwest. He holds a doctorate in computer vision and a professorship at Carnegie Mellon University (Pittsburgh).

More to the point, Rashid was the impetus behind the Mach operating system, a significant driver of mid-1980s systems software. "Mach was influential in a lot of different OSes," Rashid said. "You see it in variants of Unix, and Apple's latest OS [Rhapsody] is based on Mach."

The knowledge he gained in forging Mach puts Rashid in a good position to lead the way into the distributed-computing future. "One of the things we're pushing on is, how do we make distributed computing in some sense go away," he said. "The idea is that developers shouldn't have to be so acutely aware of their environment. Today, we still put a huge burden on the developer, requiring them to make huge choices."

One method of accomplishing this would be to move the OS abstraction layer above each individual computer in the network. "That way, you'd make managing distributed computing a task of the OS, not the applications," Rashid said.

With microprocessor clock speeds pushing toward 500 MHz, true distributed networks may become a reality sooner rather than later. Yet, if such networks are to accomplish useful work, the capabilities of systems software must catch up with the demonstrated prowess of computer hardware. "There is an imperative here," Rashid said. "If we want to expand the type of applications we can do on these networks, we have to make it easier [for developers]."

Beyond distributed computing, Microsoft Research is pursuing several far-flung areas of investigation. Most ambitious, according to Rashid, is an effort to make use of simultaneous inputs from multiple media types. "We're trying to move toward integrating different types of knowledge," he explained. "For example, taking information from cameras and integrating it with information from acoustics, such as speech or typing. The objective is to do better task modeling and provide better kinds of interfaces."

Microsoft Research has recently put together a team devoted to the topic. Indeed, Rashid noted that this attempt to find ways to integrate different types of knowledge has long been a holy grail in artificial-intelligence research.

Several other areas are on the lab's radar screen. Data mining, which entails poring through large information bases in a bid to discover new types of correlations, is one promising field. Long term, Rashid believes data mining is going to change dramatically the way science is conducted.

Microsoft Research also has a natural-language group, which investigates ways for computers to parse and comprehend normal English sentences. An early result of their work is a program called Mind Net, which is posted on the company's Web site.

Away from Microsoft Research's home base, a group of researchers led by Jim Gray is working at a satellite facility in San Francisco. Called the Bay Area Research Center (BARC), the site is studying scalable systems. The most visible fruit of that effort was the TerraServer, a huge cluster of Alpha-based servers displayed in public by Microsoft last year.

BARC is also home to Gordon Bell, the famed computer architect who developed Digital Equipment Corp.'s minicomputers in the 1960s. At Microsoft, Bell is investigating Telepresence — essentially, the technologies required to perform complex computing and communications tasks by remote means. To that end, Bell is currently looking into different forms of reliable multicast technologies.

Image problem?
Despite such formidable scientific prowess, Microsoft's researchers must contend with an overall corporate image that has taken something of a beating in the press lately. Most of the negative news revolves around an ongoing government antitrust case against Microsoft.

Still, Rashid doesn't think the research group has been affected. "Nobody likes to see your company vilified in the press," he said. "But we know the reality. If you come here and see what we're doing, morale has not suffered at all."

In its earliest days Microsoft Research had to work to counter perceptions that the company couldn't possibly be serious about building up a world-class, academic-quality operation. "In the early days, people would ask if Microsoft was really capable of sustaining this level of research," Rashid said. "By sheer will power, between [Microsoft chief technology officer] Nathan Myhrvold and me, we were able to bring in some good people."

"The basic approach I've taken with the research group is to organize it like I organized my department at Carnegie Mellon in the mid-'80s," Rashid said, adding that the organization has now reached "critical mass."

Indeed, the research operation today is home to such leading lights as Alvy Ray Smith, a founder of Pixar, and graphics guru Jim Blinn.

"The good thing is, it's big names, not big egos," Rashid added.

Currently the lab has a staff of roughly 300. But Rashid is ramping it up quickly and hopes to have some 450 researchers on hand by next June. One of the key projects that's being staffed is research into computer vision. Although the field has long been thought of as applicable mainly to factory automation, Microsoft believes it has major potential in the consumer arena.

Looking off the beaten path is necessary if Microsoft is to remain a software force into the next century. "You have to continually reinvent yourself," Rashid said. "It's already the case that software systems have evolved to be much more componentized. Clearly, we're trying to create technology that will let the company adapt to whatever comes along."

Perhaps most important, Rashid doesn't see his researchers as living in an ivory tower, but as responsive to the realities of the marketplace. "I look at this as a full-contact sport in terms of getting our ideas into the development groups at Microsoft," he said. "We're working really hard to get technology into the products."