21 Inc is the company that has created a small computer preloaded with Bitcoin software and primitive APIs to build apps that do bitcoin transactions. Their idea is that every computer, indeed, every device, on the internet should have a Bitcoin processor, contributing a small increment to validating the global ledger, and also capable of microtransactions. They’ve spoken of “lightbulbs” that are plugged in to the bitcoin network, grinding away processing transactions in the background, and so on.
The use cases for this computer are still iffy, the most believable ones are some form of ‘add a coin slot to X’ app, enabling pay as you go micropayments for digital services.
On another front, I have commented on one naïve and misguided form of public sensing of environmental data. Specifically, the naïve assertion that simply by deploying lots of cheap sensors and aggregating the readings via the internet, you can create useful or even meaningful datasets. For example, I criticized pigeon backpacks as producing data that has little use for actually understanding environmental conditions, as it was proclaimed to do.
Hey, let’s combine these two bogus ideas into one comprehensively bogus product!
I give you ‘Sensor21’, a sensor that is part of the 21 Inc. network, so you can buy and sell numbers. Well, the numbers are from sensors that measure temperature, barometric pressure, and so on. Or any way, that is what the numbers are supposed to be.
This is bogus in so many ways, it’s hard to even cover them all.
First of all, there is the obvious question of why we need a market in these readings, or to “incentivize” the collection of data in this way. Why do I want to read barometers in other cities? Why would I want to read one in real time, except my own
I think the main thing one would do with such data would be to use it as input to some kind of analytics or models, right? Perhaps some people want to run their own weather or air quality monitoring or prediction model. (Especially if you are sure that “the man” is lying to you about pollution or climate data.) This is cool, and I’m happy to have people learn about computational modeling and how to use real sensor data to understand the world.
Does Sensor21 enable this? Hardly.
First of all, what are the sensors measuring?
Hint: there is no way to know.
I note that the installation instructions seem to indicate that the sensor connects to your computer via cables a few centimeters long. So, one possibility is that the sensor is in your house or machine room where the internet connected computer lives. Not exactly the environmental reading we might have hoped for—indoor weather is not the same as outdoor weather.
Of course, the sensor might be outdoors somewhere, cabled to a computer in a housing. Where would this be? We have no idea. Is it is the sun or shade? On the ground or on a roof? Surrounded by plants and soil, or by buildings and pavement? Near a pool or river, or on a hilltop? For that matter, is the sensor permanently mounted or mobile? And so on.
The point being that environmental sensor readings themselves have little use without adequate understanding of the context in which they are taken. “Garbage In, Garbage Out” is a motto pretty much invented by computational modelers.
To really use this kind of sensor data, we would want to have quite a bit of information about the sensors, so we could identify the ones that are likely to be giving us data that is relevant, reliable, and representative.
If I want to estimate the average temperature at ground level in a city, I can’t really use readings from inside air conditioned buildings, and I have to carefully adjust for readings that come from the top of tall buildings, from the middle of wooded parks, and from the middle of paved parking lots. I probably need to know what kind of sensor is making the reading, too. (Instruments vary.)
Basically, the metadata is at least as important and valuable as the raw data, and the raw data is basically useless without it. Sensor21 could provide this kind of information, as real sensor nets do. But it doesn’t.
What would it take to make this data useful? First, you would need someone to do some serious and careful work setting up the sensors and documenting that they are working as they should. Then we would need some way to acquire information about the different sensors, so we could estimate their quality and usefulness for our own purposes.
It is important to realize the trust that is implicit here. Setting up a sensor net, especially a public environmental sensor net, is an exercise is trust. We need to know that the data is collected correctly and in contexts that we understand.
Does Sensor21 help this problem? After all, it uses blockchain’s “trust” protocol!
But just because somebody wants to sell be sensor data doesn’t mean that it is good data. In fact, Sensor21 incentivizes people to produce as much data as they can, with no quality assurance at all—every reading is equally good, according to Sensor21. So Sensor21 is actually making things worse, commoditizing the data in ways that render it pretty much useless.
I think the Sensor21 people would wave their hands and say that the data sources will have a “reputation”, which will let you figure out what to trust for what use. But that is just wishful thinking, in the absence of the real work that making such ratings require.
What I’m talking about is not really new. For example, Unidata has been distributing environmental data from various sources since before the World Wide Web, and works hard to provide trusted data .
These days, pushing data around the internet is the least of the challenges. Figuring out what it means and how it should be weighted is the hard part. Efforts such as the Semantic Sensor Web  have been underway for a decade or more, and has led to interesting work on provenance models .
In short, Sensor21 is basically trivial and technically backwards, and not only solves the wrong problem, it incentivizes the wrong behavior.
- Lazarus, Steven M., Jennifer M. Collins, Martin A. Baxter, Anne Case Hanks, Thomas M. Whittaker, Kevin R. Tyle, Stefan F. Cecelski, Bart Geerts, and Mohan K. Ramamurthy, 2012 Unidata Users Workshop Navigating Earth System Science Data. Bulletin of the American Meteorological Society, 94 (10):ES136-ES143, 2013. http://search.ebscohost.com/login.aspx?direct=true&db=a9h&AN=91672724&site=ehost-live
- Sheth, A., C. Henson, and S. S. Sahoo, Semantic Sensor Web. IEEE Internet Computing, 12 (4):78-83, 2008.
- Umuhoza, D. and R. Braun. Trustworthiness Assessment of Knowledge on the Semantic Sensor Web by Provenance Integration. In Advanced Information Networking and Applications Workshops (WAINA), 2012 26th International Conference on, 2012, 387-392. http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6185293&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6185293