By Peter Griffin 22/12/2016

Microsoft, the tech company best know for the Windows computer operating system and Office productivity suite, spent over NZ$17.3 billion on research & development in the last year.

To put that in perspective, Microsoft spends more than Google, Amazon and Apple on R&D, companies that are generally more often associated with cutting-edge innovation and futuristic technologies.

To give you an idea of the scale of Microsoft’s spend, consider this – in 2014, New Zealand businesses and the government spent a total of $2.7 billion on R&D. As they strive to come up with the ideas that will power the next wave of technology, the biggest tech companies are out-spending many countries – Microsoft’s spend on R&D was 14 per cent of revenue ($123 billion) last year.

It doesn’t disclose the exact make-up of its spend within R&D, so its hard to know exactly how much of Microsoft’s spend is going into pure research and what is spent on development of new software and hardware products.

Last month I visited the headquarters of Microsoft Research, which this year celebrates 25 years in business, and has over 1,000 researchers working in research labs in Redmond, Santa Barbara, Cambridge (Massachusetts), Cambridge (UK) and Bangalore.

I spoke to Jim Pinkelman, senior director, Microsoft Research Outreach, about where Microsoft Research has come over 25 years and the impetus for creating a research division back in 1991.

Click below to listen to Jim Pinkelman explain Microsoft’s approach to research.

Microsoft Research's Jim Pinkelman
Microsoft Research’s Jim Pinkelman

‘Don’t focus on products’

“Really it was the brainchild of Bill Gates and Nathan Myhrvold,” says Pinkelman of the Microsoft co-founder and the former chief technology office.

“They saw that computing was not going to be highly predictable. They felt that they needed people that were thinking about the future of computer science and computer-related disciplines in ways that were removed from a particular product. We needed researchers in machine learning and in human computer interfaces.”

Microsoft was already dominant on the desktops of millions of computers all over the world, but Gates knew the company had to invest more in basic scientific research to influence the next wave of computing.

“The notion was not to create an advanced technology group, something with a certain product focus. They literally wrote down ‘don’t focus on products’,” says Pinkelman.

“Product groups were great at thinking about how their current product was going to be competitive in the market and how the next one might be. But they wouldn’t think about the advances in technology that weren’t related to a product or that fell in between products.”

What you didn’t see then, back in 1991 as Microsoft Research got off the ground, staffed by computer scientists from Carnegie Mellon and other top-flight universities, was “a research team literally within Microsoft Word really trying to push the boundaries of spellchecking”.

Two key values

It was more fundamental than that and the Microsoft Research team was formed with two key values in mind:

– The researchers would remain connected to the university world and would be committed to open research – 95 per cent of Microsoft’s  research is published in open academic journals, subject to peer-review scrutiny.

– Microsoft would hire the best researchers in their fields, empowering them to push the envelope in their area of research.

As a result, the peer-reviewed papers started to mount up as Microsoft researchers worked on some of the big computer science-related developments of the last 25 years.

But it wasn’t just about blue skies research.

“It was important that they weren’t isolated and interacted with engineers in the product groups on an ongoing basis,” says Pinkelman.

“The engineers in product groups needed to bring their biggest challenges to the research organisation. Researchers needed to tout their advances on an ongoing basis to product groups.”

Some have argued that this hasn’t been done effectively enough in the history of Microsoft Research, leading the company to miss opportunities to commercialise some cutting edge technologies that researchers were working on at a fundamental level.

Indeed, as this Bloomberg piece explains, Microsoft in 2014 reorganised Microsoft Research to make sure research advances could be better harnessed to produce new products:

To break down the walls between its research group and the rest of the company, Microsoft reassigned about half of its more than 1,000 research staff in September 2014 to a new group called MSR NExT. Its focus is on projects with greater impact to the company rather than pure research. Meanwhile, the other half of Microsoft Research is getting pushed to find more significant ways it can contribute to the company’s products.

The approach seems to be working with research from the group going into some of Microsoft’s flagship products, such as Skype, Office and its new augmented reality headset the HoloLens.

“There was fundamental research that was done in computer vision to enable HoloLens, in gesture recognition, and in sound as well,” says Pinkelman as he ran a demo of Microsoft Research-developed voice recognition for noisy situations that ended up being used with the Xbox Kinect and aspects of which are built into HoloLens.

Microsoft Research has as a result ended up with three ambitions, says Pinkelman – advancing fundamental computer science, transferring those advances to improve products and create new ones, and using research to incubate new businesses.

When I ask Pinkelman for the major breakthroughs at Microsoft Research, his answer surprises me – I thought he’d point to Xbox or even the Surface computers.

“The high points are the research that went into our major server products, Windows Server, SQL Server, over the last five years Azure, and going forward, Azure to a huge degree,” he says.

“Running compute workloads and storing data in the cloud. What we’ll see over the next five or ten years is that its the services in the cloud that are of huge value to companies, to developers, to all of our customers. the notion of having world class machine learning algorithms available through azure to every data scientist who wants it, that’s mind boggling.”

“Whereas ten years ago, that wouldn’t have been available in a download package. One of the key outputs for Microsoft Research moving forward is figuring out what the service is and that’s a shorter product cycle than it is to build hardware or software.

Quantum computing and artificial intelligence are being researched at a fundamental level at Microsoft.

“We do research in quantum computing because we believe it is going to be an important computing mechanism years out from now,” says Pinkelman.

The next game-changers

Looking forward, he sees the 1,000 strong team pushing forward in three main areas – the fundamental computer science domains relating to capturing data from the real world, harnessing machine learning and artificial intelligence so computers can process what they see, and transferring those insights and recommendations to empower people.

So everything from figuring out how sensors work through to managing server loads to human computer interfaces are areas of interest for researchers.

Ultimately, says Pinkelman new products will emerge for Microsoft.

“Its not an extension of Excel or a new visualisation product for Excel or a new security mechanism for Azure, you are going to see new products come out of our research efforts. Some will be modest, they won’t necessarily lead to a billion dollar business.”

But some of them just might…

Peter Griffin visited Redmond, Seattle as a guest of Microsoft

Audio courtesy of Christian Sachsinger from Munich public service radio station Bayerischer Rundfunk

NEXT: Underwater data centres

Site Meter