Jun 08, 2011, 01.54 PM | Source: Forbes India
Rick Rashid, head of Microsoft Research, tells Forbes India that pure research is not about managing ideas, but people.
How do you align long-term pure research to dovetail with MS’ product strategy?
We have a group of people doing full time technology transfers. Having worked in product teams in the past, they monitor research and keep track of products; match and connect them. They recognise opportunities and then bring people together, in many ways, like ‘personal glue’. It is one of the single most important things we do.
Technology transfer is not about technology but about building connections between people — specifically researchers and product teams.
Let’s take Kinect. The underlying research work behind that goes back to 1990s though the specific product work goes back two years. Todd Holmdahl [corporate vice president, Interactive Entertainment Business] came to me and said I want your team’s help. His people had begun to talk to our researchers, so there was already a link. They were looking for ways to handle huge variations in people’s homes, clothing and individuals; and being able to robustly connect actions of user to characters on screen.
For example, say a dog runs in front you while using Kinect. Now if all you had was skeleton information from a 3D camera, then you could lose track of the arms and legs, but Kinect can recover from those kinds of interruptions of information and create the illusion to the user that it always saw the user. It’s very similar to what you and I do — just because my arm or leg went behind a table you don’t think it disappeared.
All of this requires very sophisticated computer vision and machine learning, both of which we began investing on in the 1990s.
Windows Phone 7 has been getting rave reviews for re-imaging mobile User Interface. What MSR projects went into it?
We’ve been working with various versions of mobile, including Windows Mobile and Windows CE, since the early 1990s.
Over time, we’ve added more intelligence about what you’re doing. As you’re doing your email, as you begin to type, we try to guess who you’ve been talking to, or what words you’re typing, what keys you’re typing. The idea is to use more intelligence so [the] user has to do less.
The word recogniser in our Bing search engine effectively does machine translation and learning algorithms, translating what people type to what they probably meant.
Innovation will often lead to failure. How do you deal with it?
We don’t deal with it.
You want silent failure, don’t make a big deal of it and let people move on. If you punish failure, you will reduce risk taking. The biggest asset for research is risk taking; ultimately you’re not going to get value if that goes down.
The biggest mistake people make when they try to manage research is they ‘manage’ research. You can only manage researchers.
When it comes to pure research, how do you know when to pull the plug on something?
It’s not about managing ideas but people. A good researcher will figure out and self correct, or the community itself will self correct. I don’t need to do that. There are times if someone gets stuck and bangs their head against the wall, then there are management techniques to get them to move on.
For instance you say “Have you talked to so-and-so?”
Hiring and firing are the two most important things I do. Most people mismanage research by trying to pick and choose winners and losers.
We never try to predict what will show up in a product till we’ve announced a product.
For instance, we still don’t know where quantum computing is going even though people have been working on it for over 15 years. Yet we’ve got some good ideas on that too through a research group in Santa Barbara that is working with physicists. It’s one of those ideas that could never turn into anything or suddenly become the best — you won’t know till you get there.
It’s easy to dismiss something because it hasn’t happened yet, but technology to make it happen could be right around the corner because of investments made. Who knows, when it happens next, it could create a new billion dollar product.
Click here to read more
More Forbes India stories