How Microsoft Research banished the BSOD

Cambridge lab's Ken Wood on Microsoft Kinect, drivers and more

Ken Wood

"Somebody used to say 'you can't schedule breakthroughs'," laughs Ken Wood. "Whereas you can schedule the next version of Office."

Wood is Deputy Managing Director at Microsoft's Cambridge Research Lab. He's the bridge between Microsoft's busy product divisions and the unhurried – but focused - light and airy lab on an unassuming academic estate on the outskirts of Cambridge, the ultimate university town.

"I look and see what research is happening and how we can take advantage of it," he explains. "I see how we can get things out to the product groups or, if there's no natural Microsoft outlet, if there are any deals we can do or partnering with people outside."

Cambridge's role in Kinect

Does work come to the labs from the product divisions? "It can do. It's quite rare, but it does happen that way. My job is more of a sales job… we just work under very different schedules [to the product divisions].

"They're under more pressure, we have to be sensitive to that. We need to know when they're in their research phase and give them ideas in advance."

"The Xbox guys came to us about Project Natal (now Microsoft Kinect) because they knew we had expertise in that area – it turned out that our object recognition was most important to them," explains Wood. "So that was an example of them coming to us."

"Look at TrueSkill which was done here. If you think 'Xbox Live is great because I always get matched with people I can have a good game with', you may not realise that's down to some Bayesian Inference that was done here a few years ago!"

Wood is affable and is more than happy to talk openly about the Research carried out at the Cambridge labs. We were interested to know how results-driven the labs actually were. Does Wood consider any of the research carried out by the lab as superfluous?

Reducing BSODs through Four Colour Theorem

"Research can be airy fairy stuff, but even if it is, it can have an impact. We have a good example of this. You've probably sat in front of a Windows machine that's bluescreened on you."

"There are quite a lot fewer blue screens than there used to be, right? We have a guy in the lab here, Georges Gonthier, who's a mathematician. And he's famous now for having proved the four colour theorem," explains Wood.

The Four Colour Theorem is the mathematical theory proving what the minimum number of colours you need to colour a map or diagram so no two bordering countries are the same colour.

"People thought it was probably four, but there was no mathematical model. So this had been an open question, people had never proved it – a couple of people had computer-based proofs, but the mathematicians wouldn't accept them as they couldn't be sure the software was doing the right thing.

"So George came up with a computer-based proof where he could prove mathematically how the computer stuff he was using was sound. And the mathematicians accepted it.

"Now, the four colour theorem itself is of no use to anybody! But, the techniques that George used to prove that his computer-based proof was correct happens to have direct application to a set of tools that we give to people that build PC peripherals. When your computer blue screens, almost every time it's not Microsoft software, it's driver software.

"It's a driver for a mouse [or other peripheral]. So we've got a suite of tools that lets them throw their software in and test… you've got a memory leak here... We don't let people release drivers [without this].

"And that's down to the work that Georges did – it directly had an impact on the number of blue screens you see. I like that story, it's one of the best research to practical applications stories there is."

Where do ideas come from?

We asked Wood where new research topics come from. "Ideas come only from the researchers themselves. One of the tenets of this place is that we hire smart people and let them do what they want.

"They tend to work like they would in a university. They think of a hard problem that they want to study and they work on it. What happens over the years is that you tend to hire people [in specific areas]. In that way we tend to build areas of expertise.

"We have a very strong Human Computer Interaction (HCI) group…but they tend to grow organically. It's not like [each lab] will do [different things]. Sometimes there might even be some friendly competition. It just kind of works out. It's very much bottom-up.

"The people we hire are clever, so if they've done something they think will be valuable to Microsoft, they'll patent it, but there's no permission needed to publish something. It's a very open atmosphere, that's how we have to work."

Article continues below