SenseCam wearable camera leads parade of Microsoft Research novelties

Judging by volume of footfall and word of mouth, the virtual prize for most interesting project at Microsoft Research Cambridge’s open day for media and others last week was probably the SenseCam, a device that is worn by the user and takes pictures automatically to create a record of activity. Although many see consumer applications of SenseCam — ‘life blogging’, for example — an early commercial product is from Vicon, an Oxford-based company which has licensed the design for its Revue product that is aimed at medical applications, including for people suffering from memory loss problems. Pricing is currently £500 a system but as volumes increase and prices fall, these devices could, as well as bringing serious gains, become hugely popular items for everyday use.

Another device that had visitors cooing was Peppermill, a prototype input device that is steered by twisting the top of the hardware just like twisting pepper pods in a mill. This twisting motion acts as a power source so no battery is required and the device will even provide feedback on direction and rate of input.

Other intriguing projects include (En)tangled Word Bank, a tool that identifies and maps changes to a text. In the case of the example shown in Cambridge the text was Darwin’s The Origin of Species and the resulting visual portrayals of change were themselves exquisite, looking more like exotic plants than data mining reports.

Another project, the purpose of which was described as ‘Making Datacentres Power-Proportional’ and codenamed Sierra, is intended to save power on storage in cloud computing by powering down storage servers at times of low activity. It has already been tested on Microsoft services including Hotmail and could end up as part of Windows Azure.

Not all the demonstrations were equally convincing. The Gathering Engine for example, described as a “look into the future of web searching tools”, is based on a dialogue between user and computer. Rather than using query analysis or clickstream data, the tool provides information that it anticipates will be interesting. The result is more akin to leafing through an encyclopaedia and happening upon snippets rather than a Google-type search for relevant information. Interesting, but would I really want to jump from ‘Bob Dylan’ to ‘Billy Bob Thornton’?

However, not all the demonstrations were far-fetched. Two Cambridge developments are already built into commercial products from Microsoft: the F# development language that is part of Visual Studio, and AdPredictor, an algorithm intended to serve relevant ads based on Bing searches.

If you missed it, I posted an interview with Microsoft Research Cambridge MD Andrew Herbert here last week.