• 
• 

 1 of 2

Gallery

The future according to Microsoft

Posted on 4 Jan 2013 at 14:07

We explore Microsoft's research labs, with projects to improve gesture recognition and eliminate blue screen errors

When it comes to carrying out research, Microsoft plays the long game. At the moment, its researchers are working on sociological improvements to search, improving gesture recognition systems, and solving hundred-year-old mathematical theorems. While that may not sound like a route to improving software and hardware, give Microsoft some credit – it’s already reduced the occurrences of the Blue Screen of Death.

Academic focus

Microsoft Research (MSR) has 1,000 employees working globally, including 150 at its Cambridge lab taking up three floors on the grounds of the famous university, says Ken Woodberry, its deputy managing director. The goal of the lab isn’t product development, but furthering academic knowledge in fields such as computing science and mathematics.

“The people who work here would otherwise be academics,” says Woodberry. “We measure ourselves using a couple of metrics, but one of the primary ones is simply as though we were academics.” This means using “strong, peer-reviewed publications [and] status in the academic world”.

Of course, Microsoft funds its research department with a purpose: bringing innovations and expertise to the larger company, to help as needed.

The kind of people we tend to hire want to have an impact in the wider world, so they tend to work on problems that will eventually have some relevance to the company

“It’s entirely bottom-up; we hire smart people and let them do what they want to do. The kind of people we tend to hire want to have an impact in the wider world, so they tend to work on problems that will eventually have some relevance to the company.”

“It gives you a deep pool of expertise in areas that the company might not realise is important now, but will turn out to be important later on,” he says. “Search was something that famously we were a little late on, as a company. However, as soon as we realised we had to do something in that area, we had such a tremendous amount of expertise in the relevant computer science areas in MSR that it turned out they could turn their hands to building Bing.”

Digits

One area of focus for Microsoft’s Cambridge lab is gesture recognition – this is the firm that invented Kinect, after all. While the games console system didn’t actually stem from lab research, it was refined by it. The Kinect team ran into one or two problems, and they knew that MSR had expertise in computer vision and related areas. “We were able to help them solve those problems... we’re quite proud of that,” says Woodberry.

Another gesture-related hardware product is Digits. “It’s a wrist-worn device that lets you track the movements of your fingers – if you turned it into a product, it might be a watch. As you move your fingers, imagine you’re typing in mid-air; it can know exactly what your fingers are doing.”

The interaction works as though the user is wearing data-collecting gloves, analysing the hand’s movement in 3D using lasers and an infrared camera, and comparing that to a 3D model of a hand – the researchers spent hours staring at their own hands to try to understand how the system should work. It allows users to interact with a device anywhere – accessing a smartphone that’s still in your pocket, for example. Aside from interacting with mobile devices, Digits could also be combined with Kinect for console gaming: demo videos show users pointing at the screen and pulling their “trigger” finger to shoot.

The first versions of Digits were made using off-the-shelf parts, so it’s bulkier than the researchers would like, but by making specialised components, it could be the size of a wristwatch.

Blue skies to stop blue screens

The work happening at Cambridge isn’t only about hardware. Researcher Georges Gonthier is “even more famous”, says Woodberry, after proving the Feit-Thompson theorem – which states that in “mathematical group theory, every finite group of odd order is solvable”.

That may make little sense to those of us who barely scraped an A-level in maths, but Gonthier’s proof of the idea took six years to solve and two books to publish. “The kind of techniques he had to use to prove that theorem may have applications in areas such as proving software is correct,” explains Woodberry. “It’s a good example of a blue sky thing that’s in Microsoft’s interest to fund, because the chances are it will have a positive impact on things that will affect the company.”

If that sounds far-fetched, consider the last famous mathematical quandary that Gonthier solved: the four-colour theorem. “If you have a map, how many colours do you need to colour every area on the map so that no two adjacent areas are the same colour?” asks Woodberry. “If you think about it, it looks like the answer is probably four – and it is four – but nobody had proven it since 1850 or whatever, and Georges did prove it.”

To achieve this, he used a combination of pencil-and-paper work and computers – and using computers to prove the theorem hadn’t been done before to the satisfaction of mathematicians. “However, the technique that he uses to do the computer-based side of the proof... has been used in something Microsoft built called the ‘static driver verifier’, which we give to third-party partners that are writing driver software for whatever they’re going to attach to the PC, to make those drivers less likely to cause blue screens,” Woodberry explains. “In the old days, when PCs blue-screened more often, it was almost always a driver that caused it. And that software we hadn’t written – somebody else had – and it was usually that software that caused the problem.”

The verifier, based on Gonthier’s work, lets third-party coders check their work – and, as a result, means users see fewer blue screens. “If you said to someone, would it be in Microsoft’s interest to pay somebody whose goal it is to prove the four-colour theorem, they’d probably say no, but in fact it was. In a nutshell, that’s why this model of a bottom-up, researcher-driven blue sky lab really does pay off.”

Searching for better search

The “ten blue links” design of search engines hasn’t changed much since browsers first arrived

Other projects have clear product applications. Woodberry says a lot of “exciting” work is happening around search – clearly to help give Bing a boost over rival Google. However, it isn’t only about technology and algorithms. “Social scientists – sociologists, psychologists, designers – are trying to unpick what people are doing when they go to a web browser,” he says, noting that the “ten blue links” design of search engines hasn’t changed much since they first arrived.

Researchers are looking at why that is, and whether different methods might improve certain types of search tasks. “People have been asking this question for a long time,” he notes.

While Google and Bing both try to deliver the fastest search results possible, that might not be the only way. “Maybe the idea of slow search, rather than instant, [where] you have a long-term interest in something,” he suggests.

Author: Nicole Kobie