Our consumer electronic devices are talking back. Sure, there are privacy problems but they are creating an incredibly rich source of useful data for all of us. Some of the best data comes when no human intervention is necessary so the data flow isn't gated by human attention to the process.
Here's a partial list of applications I've heard about recently, some without links because I don't have time to do the research. Great if commenters could add to the list or add the links.
Earthquake detection and prediction: Many laptop computers have accelerometers in them: very cheap devices which can sense motion. Their primary use is to sense if the laptop is falling off a desk and park the disk to avoid damage. But they can also detect minute and not so minute movements of the earth. There is an application which aggregates their information and uses it for gathering extremely sensitive and distributed seismic data.
Weather: The weather stations that we weather nerds and sailors like to deploy are now networkable. WeatherUnderground is a huge network of stations with near realtime data available and some weathercams. Historical data can be retrieved with files, feeds, and APIs. The laws of chaos seem to say that this won't help much with long term predictions but it sure helps with short term predictions of what conditions you'll meet down the road, over the mountain, or across the lake. Also good historical data for siting wind turbines or solar panels and predicting their output.
Traffic: I've already posted in rhyme and prose about Dash Express, a GPS which uploads the speeds it has traveled at for aggregation into realtime and historical road condition databases which it then downloads and overlays on its roadmaps.
Disease spread: Reportedly Google is experimenting with tracking the spread of flu by the geographical origin of searches with words like sneezes and sniffles.
Electric outages: One of the advantages of smart meters is that, when they stop communicating, utilities can quickly spot an outage – often even before you call them about it. Software can deduce the location of the map by the response and non-response of the meters.
TV watching: Your DVR is watching you. It knows whether you fast-forwarded through the ads or not as well as which shows you selected. Not sure if this is good or bad but it's a lot more accurate than the old survey method of detecting the popularity of shows.
One thing that all these applications have in common is that the devices were NOT distributed for the purpose of aggregating data; they already had a use which justified their deployment and the data collection is just a fringe benefit. That's very important for getting a network application going because there is no value in a vary sparse network. nevertheless, some data the value of the aggregate data collected may be as large as the value of the original application.
Another interesting thing (to me) is that very large networks are required to make this data useful BUT the individual stations DON'T need to be professionally certified. Even fraud isn't a problem let alone malfunction. With enough datapoints, it's simple enough to simply ignore the outliers.
One predicted application: nautical charting. Ever since mariners started making charts, they have been a collaborative effort. But now many of us are sailing or powering around with depthfinders and GPSes. We can upload realtime data on the location of even the shiftiest sandbars; we can download the aggregated uploaded information. Even if our uploads aren't uploaded until we get close to shore and WiFi and we can only download when we moor in a marina, the data would be much better than what we have now on charts which are sometimes literally 100 years old even if they are sold in GPS format.