Converting blue sky research into products that can yield hard cash is a constant challenge for IT vendors.
This is one reason why the main focus of Microsoft’s Enabling Innovation Through Research event at its Cambridge labs on Wednesday was on broad appeal consumer technology, such as body motion tracking and self powered controls for gaming, and video and image manipulation tools destined for applications like digital painting.
But one area that appears to offer real promise for future business applications is making it easier for organisations to obtain more accurate results from carrying out searches into large and complex datasets that are derived from multiple sources, and giving computers the ability to make intelligent decisions on that data for themselves.
This is being driven partly by the inexorable migration of services and applications to cloud computing environments, where data sets such as geographical information, distributed sensing data, social networks, electronic health records or genome data for example, are no longer stored in isolation but jointly in the cloud.
The cloud offers processing capabilities on a scale that can fuse and extract new information from these data sets for the first time, says Microsoft Cambridge chief research scientist, professor Chris Bishop.
He has spent the past few years examining ways to improve mathematical algorithms and probability forecasts to link statistical learning with the knowledge and expertise stored in these data sets to enable the computers themselves to recognise patterns and make intelligent decisions based on their findings, a type of artificial intelligence, or machine learning.
“Statistical testing has been fairly lightweight, but in the last few years there has been an explosion of data coming from so many different devices, tied to the compute power of the cloud. The magic sauce is very efficient algorithms that give pretty accurate results,” he said.
“The breadth of applicability is huge – rather than being purely stats driven it is much more about the value of the information we can extract from that data.”
An example is recent work Bishop has done with Asthma UK in conjunction with the University of Manchester, where thousands of children are monitored not only for symptoms, but also for the number of bugs in the carpet in their bedroom, the number of soft toys they own and a variety of other conditions.
The problem with immense projects of this type, said Bishop, is scale and the fact that it is difficult to know exactly how accurate the information being extracted is.
“Large-scale applications are constrained by available processing power – we can dream up something on the whiteboard but without the scale behind it, it will take half an hour to get an answer,” he said.
“Plus, the margin of error depends on the application. Searches, for example, will not find everything because the sophistication is not there yet, but a search engine that works is better than no search engine at all.”
Microsoft’s ongoing rivalry with Google, which culminated in the launch of a rival search engine Bing last year, indicates machine learning is almost certain to make it out of the research community and into mainstream online search market at some point in the future.
The Cambridge Research laboratory has been home to hundreds of researchers since it opened in 1997. When the lease expires in 2012, it will move to a newly built facility closer to the centre of town to accommodate an even larger complement of internationally renowned boffins and newly graduated researchers.
Notable successes so far include motion capture and sensor technology which initially found its way into cameras manufactured by Vicon, given a prize for best new product developed in the UK at the British Computer Society (BCS) and Computing awards event last year.
These T-Series cameras are used for life sciences, animation and engineering applications, and the same SenseCam technology has also been licensed into a new breed of wearable digital cameras for medical use.
Aimed at the estimated 1 in 14 people that suffer a variety of memory conditions, including dementia and Alzheimer’s, the £500 Vicon Revue takes pictures automatically at regular intervals triggered by events such as alterations in light conditions, movement, inertia and changes in direction.
“The challenge is to add further sensors, like GPS tracking, and better batteries,” said Vicon CEO Nick Bolton. “As a company, we are focused on clinical applications, but there are possible uses in defence as a type of black box for soldiers for example, or even police forensics.”
Cameras and motion detection technology also feature large in project Natal, a natural user interface (NUI) for the Xbox 360 games console.
A horizontal bar consisting of an infrared projector and CMOS sensor allows the system to see in 3D, and recognise various body movements to navigate a game and execute character actions like punches or kicks without using a controller.
So far, Microsoft has concentrated on developing the technology for the lucrative gaming market, but post-doctoral researcher Jamie Shotton believes other applications are possible.
“It may have some traction in medical imaging, where a surgeon who is unable to use his hands can use arm movements to scroll through data on screen, for example,” he said. “There are hundreds of groups looking at ways to use this technology.”
By eliminating high entry costs for big data analysis, you can convert more raw data into valuable business insight.
A discussion of the "risk perception gap", its implications and how it can be closed