by Martin Veitch

Microsoft Research Cambridge MD on R&D, open source, cloud and more

May 06, 2010
Mobile Apps

At a Microsoft Research open day for press and selected others on Wednesday the only hard news of the day came early and from that great source of information for journalists, the cab driver.

‘Get many fares out to Microsoft Research?’ I ask.

‘Reasonable,’ he answers. ‘Pity they’re moving out from that site.’

It turns out he’s right. Microsoft Research Cambridge (the UK version rather than the US New England site] is indeed moving its big primary R&D shop in a couple of years, shifting to a new site right by the railway station and thereby costing cabbies 10 quid a pop, tip included.

“We’re absolutely staying in Cambridge but we need more space,” confirms Andrew Herbert, managing director of Microsoft Research Cambridge, when I interview him later in the day, adding that the total of 150 staff is supplemented by additional visiting researchers and others to over 200 at peak periods. Originally employing about 50 staff when set up in 1997 and the Web was a novelty, Microsoft has outgrown the building, he feels.

I ask Herbert, a 10-year veteran of the facility and former PhD student to founder Roger Needham, whether Microsoft — which even before the challenges of the current economy and market was always a value-conscious organisation despite its vast wealth — applies much in the way of KPIs to Microsoft Research.

“Not in any explicit sense,” he says. “As soon as you develop KPIs people optimise for them. There’s a huge amount of trust. We have a conversation with [Microsoft chief research and strategy officer] Craig Mundie every year about the budget.”

And is that budget going up, down or sideways?

“It’s consistent. The key thing is we’re still recruiting,” he says, adding that there is a requirement to be “more thoughtful … We’ll only hire a researcher if we’re confident we can give them the resources. [Recruiting is] not tougher but we’re competing more. Microsoft was less well known as a research organisation [when Cambridge was set up 13 years ago]. We had a job to convince people we were a research lab and that we would do exciting basic research. We find ourselves competing with Google, IBM and other companies. We do quite well [in recruiting top talent]. The quality of the research and the people is a tremendous magnet. The decision [to join] is often more what you want to do and the resources rather than the company badge.”

He draws a quite severe distinction between the attitude of Microsoft and rivals.

“It’s much more directed at product groups at IBM and HP. That can be frustrating as if companies change priorities you’re stranded. Google is interesting. Their people tend not to publish their research quite so freely. They encourage people to work with the engineering process. [Microsoft offers people] freedom to go with their own ideas.”

Like the youth academy at a football club it helps, of course, if every now and then one of the trainees turns into a superstar and Cambridge has pulled out a couple of plums recently in the shape of the F# (F stands for Fun and Functional it turns out) language that is now part of Visual Studio and Ad Predict, a technology that helps deliver more relevant adverts to place alongside search results in what is now Bing.

Given that we were on the eve of the British election, I considered it timely to ask Herbert what he thought of the state of computing in the UK. He answers, in true techie fashion, that there may be “an impedance mismatch“.

“We have a pretty good British software industry. It tracks back to getting kids at the youngest age passionate about developing software. That’s something we’ve kind of lost our way on. Computer teaching is about using email [and basic productivity tools]. It’s ironic [that many years ago] it was all about making the machine do something.”

Cambridge is working with the BCS to improve matters and get children interested in developing software “and a big part of that is educating teachers. A PE teacher can teach spreadsheets but not necessarily programming”.

In a rapidly globalising R&D sector, partnerships are important, he concedes, but he insists that shipping out R&D is A Bad Idea.

“In the crudest sense it’s buying in ideas. If you haven’t got your own people doing research at some level it’s hard to feel the pulse. In the UK a lot of software is buried in finance or engineering and they need to be connected to the research community. In hard times you need to be focusing your research investments in people with a track record who are going to be able to make the big advances. I need to be able to tap into global research [but] research deliverables [also turn out to equal] trained people.”

CIOs need to also “feel the pulse” of the coming waves of technology, he says, rather than just trusting in bought-in platforms.

“That’s not an IT strategy,” he argues. “It’s coming to things like today and saying ‘do I believe it?’ Partnering is great but you’ve got to make sure supporting new technologies and to place your bets you have some skin in the game.”

OK, so what’s the next breakthrough that (to steal a phrase from Apple rather than Microsoft) will be insanely great? What does he talk about at dinner parties? He suggests ‘inference’: the way clever computers can understand what dumb users are trying to do and hold their hands and help them.

“The one I find myself talking to friends about is this inference stuff and how that’s changing the relationship with the computer. There are all sorts of examples where we’ve let the machine take charge [such as relying on search engines to understand what we’re looking for]. It’s one of the most important tools for data mining rather than just counting and binning. If it’s a dinner party of CIOs then [I will talk] about improving productivity of developers.”

He talks about F# and other languages with inherent safety checking allowing companies to build and deploy very quickly, thus saving chunks of time to market and cost.”

Herbert is an academic but he has also worked at the sharp end of the commercial sector having built a startup called Digitivity in the 1990s which was sold to Citrix. So with both those hats on what does he think about the hype over cloud computing?

“It’s new so people are getting excited,” he says. “It makes a great deal of sense to me because of the economies of the datacentre; if you can scale up it drives costs out. It really is back to the future for mainframe sharing [but] it moves computing from the capital to the revenue budget. There are interesting discussions about a middle way…”

Ah, the famous private cloud, I interrupt. Wrongly, as it turns out.

Herbert flatly states that the private cloud makes no sense but what is interesting is ways to bridge between clouds and internal estates. With its experience of IT complexity, directories and so on, Microsoft is in a good position here, he contends, compared to some rivals.

“For Google and Amazon that’s a bit more of a hill to climb,” he argues. “The other story is you can just go to thin clients but if you want to have a rich user experience that actually requires quite a bit of computing in the user’s hands.”

Finally, I ask about open source and Cambridge’s attitude. He argues that what passes for “contributions” from many IT behemoths turn out to be only a last resting place for old editors and GUIs.

“It’s something we talk about a lot,” he says. “We release quite a lot of stuff through open source agreements so people can access them and lift the covers off. You can download F# and run it as a standalone without Visual Studio on other operating systems and stare inside. As a company we’ll do more to address interoperability and open-source solutions is a big part of that. The big question is where is the investment coming from for new ideas. Who’s paying for it?”

And with that my time with Andrew Herbert is up. I’m off to see the latest programs, gizmos and gadgets – which I’ll come to in my next blog.