Security cameras are actually fairly creepy at this point. They’re small, they’re increasingly everywhere, and they’re recording everything we do.
But that’s not nearly creepy enough for the government or Carnegie Mellon researchers: Now they want cameras to independently parse their own footage and accurately guess what you’re doing next.
The basic idea is that as the computer is fed footage, it compares that against predicted behavior. For example, if it sees you eating at Taco Bell, it knows there’s a strong probability you’ll stagger away from your food court table clutching your stomach and searching desperately for a bathroom.
But when you stray from this behavior, or act in a way that’s suspicious right from the start, the computer is able to alert a human, who can then see whether the person observed is committing a crime or just doing something weird.
The downside of all this is that your privacy and security is trusted to a cold, unblinking eye that will ruthlessly measure and analyze everything you do. And possibly get better at it as it goes.
I want more like this!
Follow us on Facebook and get the latest before everyone else.