Playing Around: Games and scale

I saw a great presentation last week when I spent the day down at Harvard doing my thing there. The presenter was Shwetak Patel from the University of Washington, and his subject was some work he has been doing in ubiquitous computing.

I did some work in this area myself some time ago, both as part of an NRC committee and within Sun Labs. But Shwetak's approach to the problems of this area are rather different than most. Most people doing ubiquitous computing take the term literally, and try to spread computers (usually in the form of wireless sensor nodes) all over the place. The result is that there is always a computer observing, and there are lots of interesting problems like making sure that the battery life of the sensor node is maximized and how to deal with very lossy networks. There are times that I've gotten a bit frustrated with this sort of research, since it seems less about sensing the environment than about building networks out of small, low-powered, not very competent computers.

But Shwetak's approach is less to spread computers all over the place than it is to use a small number of sensors to gather information from the existing infrastructure in the environment. Rather than trying to track items by tagging them and then putting tag readers throughout a space, for example, his research tracks tags by plugging one or two sensors into the existing electrical system of an area, and then using the existing wiring as an antenna that can be used to track the tag. He has also found that, by attaching a sensor to a single water faucet, he can track the use of water throughout the house. And a central air conditioning system will let him track movement through a house by adding a couple of sensors in the air exchange system.

Since I also work on privacy, my first response was one of concern. But then I started thinking about the approach to ubiquitous computing. Rather than putting computing everywhere, this approach is to compute on information that is available from what is already everywhere. Or, as my colleague Harry Lewis put it, one man's noise is another man's signal.

The great thing about this approach is that it is minimally invasive, requires very little investment, and uses information that is already around but being discarded. It can also be very useful-- the sensors using the electrical system of a house as an antenna can also detect arcing from bad wiring or faulty appliances, which is a common cause of fires. I wonder how many other channels of information are around that we are ignoring, and what we could do with it once we start to measure and monitor that information. It is certainly simpler than putting sensors everywhere, and could give us a lot of environmental information that would be of great value as we try to conserve energy, understand the environment, and make our world safer. And it is all based on thinking about the information that is already available to us, rather than on spending to add to the sources of information.