Thinking up Microsoft's next advantage

The company's then-technology guru, Nathan Myhrvold--who tried and tried and tried unsuccessfully to convince Rashid to leave his Carnegie Mellon University professorship and become head of Microsoft Research--bestowed the nickname.

Chairman Bill Gates had better luck. These days, Rashid leads more than 600 researchers who work on giving Microsoft a weighty presence in academic circles while trying to come up with the ideas that will keep the software maker ahead of the curve in a fast-moving business.

Rashid joined Microsoft in September 1991. At Carnegie Mellon, he led development of an operating system core called the Mach microkernel, software that has become the heart of two Windows operating-system rivals: Hurd from the Free Software Foundation and the Darwin heart of Apple Computer's new Mac OS X.

Rashid spoke with CNET News.com during an event recognizing the 10th anniversary of Microsoft Research.

Q: Why were you initially reluctant to join Microsoft, and what changed your mind?
A: I was a professor at Carnegie Mellon University for 12 years, so I had a lot of emotional investment in the university. I basically grew up in that environment in terms of my career. I really believed in the way the department was run. It was a hard decision to leave. At the same time, I've always been intrigued in the idea of creating things from scratch. When I went to CMU as a student becoming a professor, I was working in an area where there wasn't anybody there.

When I met with Bill (Gates), my first reaction was, "I don't know if he really understands what he's getting into." I knew Nathan (Myhrvold) did. But does (Gates) really understand it's going to be a long-term thing, that you can't really predict what's going to come out of it? I kept asking questions, expecting to hear the wrong answer and getting the right answer. Ultimately, what got me to make the decision to come was the fact that it seemed like it's going to be so much fun. It was going to be a kick. I thought, "My reputation is well-established. I bet I could get a job if this doesn't work out."

I was impressed by how smart everybody was, especially at the executive ranks. It was to me really surprising coming from a university environment and meeting all the top executives in the company and finding out that these guys not only were incredibly bright, they also were very knowledgeable about research. Jim Allchin, a contemporary of mine, is a good example. He had been working on a distributed operating system called Clouds at Georgia Tech.

How is Microsoft Research culturally different from the rest of Microsoft?
It's a more academic group. One thing that distinguishes a researcher from a product-development person (can be seen when) they both have an idea. A researcher is more likely to say, "Well, let's try this, and if it doesn't work, OK." Whereas a product person might say, "I don't think this will work. I don't have time for this." A product person will say, "How can I narrow this down so I can assure success in a particular period of time?" There's more of a sense of expansion in the way a researcher is going to attack a problem.

How do you balance work that's abstract with work that applies more directly to actual products?
I don't even think about it. In my mind, it's a question of how do I have a balanced, healthy research organization that constantly produces great results? I know a successful research organization needs all those pieces...to solve new sets of problems that one area can't solve by itself. You want to have a mix of talents and attitudes. You need to have a range of different perspectives. A researcher in one group will collaborate with a researcher in another group. In the course of their interaction, they may come up with an idea (with) potential product activity.

Different parts of computer science work on different schedules. With work in my area--operating systems and signal processing--when something works, it can often be translated into product form very quickly. In other areas, there may not be an obvious vehicle. For example, computer vision or computer language theory has an impact, but how it may find its way into a product isn't really obvious. And other things are so darned hard it takes a long time to do them. We've invested 10 years in natural-language processing. We have a tremendous capability in a number of languages, and there's still a fair distance to go. There are problems that are just real hard.

Microsoft is mostly a software company, but Microsoft Research is working on hardware such as microelectronic mechanical systems (MEMS), the tiny machines produced by the same lithographic techniques used to make chips. Is MEMS and other hardware research directly relevant to Microsoft?
We need to understand where the limits are in a number of hardware areas so we can design software. Take the Tablet PC, for example. People in research were involved in building the reference design. We need to have people develop the new technology. The work we do in MEMS technology may be relevant for next-generation, high-resolution displays or looking at very small devices that could be part of the .Net infrastructure in the future. It's not a large commitment, but having people in those areas gives us an opportunity to understand the state of the art. The same thing happened with graphics chips and Xbox.

You've showcased many research successes, but what duds have you worked on?
In the mid-'90s, I directed a lot of Microsoft's interactive-TV work. I'm astounded that over an 18-month period, we developed the hardware for a set-top box, the software, the server environment, the fiber-optics to the curb in Japan. We developed it, and nobody used it. It was so far ahead of its time that there was no market. But we got a lot of good ideas. A lot evolved into what is the TV platform. But I think at the time we hoped to have more of an impact with that.

Does Microsoft fund the Research group for corporate prestige--the modern equivalent of the Medici family in Renaissance Italy funding artists? Or is it more pragmatic?
That was one of the concerns I had talking with Bill: having a research lab just so he could claim he had a research lab. But you don't have the kind of investment we made if that's all you want. IBM has made very long-term investment in basic research. It's been a fundamental tenet of their philosophy over the years. The old AT&T really had the philosophy, too, but as the company has split itself up, the child companies have had different agendas. Lucent has a strong group, but even they're under huge pressure right now.

There may be companies that invest in basic research for the show value. Those organizations don't wind up getting a lot of serious research done. It's too much money to do it just for show.