The system works thus: you give it a picture of a suspect, and it tries to match the snap to one of three billion faces scraped from Facebook, LinkedIn, YouTube, and so on. If it finds a hit, it returns details about the profile or webpage it found the photo on, allowing investigators to discover the identity of the wanted person.
That sounds like an amazingly useful tool to me.
But "privacy advocates" or whatever they're called seem to think that this system will automatically launch a drone strike on the target which it will have incorrectly identified 90% of the time.
