<Mounts hobbyhorse>
To me the scariest thing about the march of technology is 1) facial recognition, 2) voice transcription, and 3) natural language processing.
At the moment there are no real privacy concerns. Although we're videotaping lots of public spaces, in general there isn't anyone watching the camera. Nonetheless, the data is out there (more and more so every day) just waiting for someone to figure out a way to start snooping with it.
There are two practical problems with large scale electro-snooping:
1. It's computationally intensive. (That's a comp.sci euphemism for slow.)
2. Even given unlimited access to computing power, the algorithms for addressing the three problems are still in their infancy.
In regards to the first hurdle, the EE types pointed out that machines are getting faster all the time. A problem that is prohibitively difficult today is not likely to be so in 10 years.
For instance, in 1976 the U.S. adopted an algorithm called DES (Data Encryption Standard) for secure communications. It was handed out to civilian industries that had a legitimate need for secure communications--mostly banks, if memory serves. About 10 years later the DES was useless. The problem was that in the intervening decade, computers had gotten so much faster it was now possible to just do an exhaustive search of all the possible keys for a given message.
Anyway, given
Moore's law hardware will sooner or later be fast enough and cheap enough to devote lots of processing power to every bit of video, voice, and text data everywhere.
Which brings me to the second hurdle.
Regardless of how fast and cheap your hardware is, there is currently no good way to, say, have a machine monitor conversations. Natural language processing (machine 'understanding' of human speech) is a tough nut to crack.
However, there are a lot of really bright Comp. Sci. graduates tilting at this particular windmill.* I'm reasonably confident that in the next 50 years or so we'll have something approaching true understanding of natural language by machines.
So, here's the kicker:
1. The data is out there waiting to be sifted.
2. It's more or less inevitable that the computing power to do large-scale sifting of the data will be available shortly.
3. It's less inevitable that effective sifting techniques will be developed, but I personally am convinced that they too are on the way.
So why should you care?
First, let's assume for the sake of argument that you're one of those from the "if you're not doing anything wrong you don't have anything to worry about" camp. Even so, it's scary.
Let's say there's an election coming up, and polls show that it's going to be a tight one. Let's further say that political strategists have identified two or three districts that are likely to be pivotal. If the party in power has access to an adequately sophisticated surveillance system, is it not possible that they might sift through the data looking for felonies commited by opposition voters? (Felony conviction == disenfranchisement.)
If that's too far-fetched for your taste, what's wrong with the following: In 2020, Bob is running against the incumbent candidate for senate, and he's winning. The incumbent candidate, by virtue of being on some senate committee or other, has access to our surveillance system. He asks his buddy in the FBI to sift through all the ATM, grocery store, and whatever video footage from Bob's college days and see if they can't find pics of Bob smoking something illegal, kissing a member of the same sex, or whatever. Is that really so unlikely?
(At this point the reader is encouraged to construct their own paranoid fantasies. )
