Intel Casting for More Sensor Data

April 20, 2011
261 Views

My health-care chapter in The Numerati focused on Intel’s efforts to mine the minutia of elderly people’s lives. In pilot projects, they wired people’s homes with all kinds of sensors, and then crunched petabytes of behavioral data: Their walking patterns around the house, the strength of their voices, even the tilt of their bodies in the kitchen. The idea was to puzzle out the patterns of diseases like Parkinson’s and Alzheimer’s–and hopefully to intervene more quickly.

My health-care chapter in The Numerati focused on Intel’s efforts to mine the minutia of elderly people’s lives. In pilot projects, they wired people’s homes with all kinds of sensors, and then crunched petabytes of behavioral data: Their walking patterns around the house, the strength of their voices, even the tilt of their bodies in the kitchen. The idea was to puzzle out the patterns of diseases like Parkinson’s and Alzheimer’s–and hopefully to intervene more quickly. (Here’s a New Statesman article with some of the details.)

Now, writes my friend and former colleague, Olga Kharif, Intel is planning to expand research to 10,000 seniors in a $200 million study. However, in the example’s Intel’s Eric Dishman provides, the intelligence derived from the data is more elementary. Instead of looking at sensors producing rivers of behavioral data, the emphasis now appears to be using sensors for alerts. This could mean less fun for dataminers. But the simple alerts could pay off more quickly.

For example: You don’t need much statistical analysis to gain an insight from one single piece of data: Mom didn’t get out of bed this morning. Or Grandad hasn’t gone into the kitchen in a week. Or he’s been in the bathroom for 4 hours. Valuable data. No machine learning required.

This doesn’t mean the end of the original vision, in which computers would spot oncoming diseases, perhaps even before humans knew they had them. But according to Eric Dishman, Intel’s lead researcher on the project, Intel can’t develop the knowledge and technology alone: …ldquo;It…rsquo;s too expensive even for Intel to single-handedly produce the clinical and financial evidence that these technologies detect diseases and lower costs,…rdquo; Dishman said. …ldquo;Even competitors need to come together and co-invest….rdquo;

In the meantime, I’m betting that middle-aged children will urge their distant parents to install ever more sensor devices in their homes. In the short term, these gadgets will send alarms. But over time, they’ll produce the rich rivers of behavioral data that the Numerati feed on.

We drove out to Madison last weekend to visit family, and made stops on the way home in Detroit and State College, Pa. The weather, for the most part, was miserable. But the visits were great. Above is my view from the restaurant of the Edgewater Hotel, in Madison, looking out on Lake Mendota.