Why are you less impressed by small percentages than by numbers that are large only in the absence of a reference scale? It's generally not possible to design flawless systems, and failure rate a much more meaningful number.
Obviously a very small failure rate is impressive in the abstract. We're not talking about the abstract here, we're talking about foundational concerns about whether there can be an expectation of privacy in any online communication.
But! Taking these numbers at face value, the low error rate tells us that, yes, we can have privacy with pretty high probability. As with all things in life, it is not guaranteed, but these are much better chances than what we can expect in other aspects of life. I mean, tens of thousands erroneously picked up over many months out of the billions of emails exchanged daily? Why, you're more likely to die in a terrorist attack than be snooped on [1]!
[1] No, I haven't run the numbers, just saying so for dramatic effect :-P
You'll never convince me that any jihadist has the capacity to inflict harm on society more than a government with unchecked power to know all secrets. Knowledge is power, and complete knowledge corrupts completely.
You'll never be convinced because you are, presumptively, ensconced in the safe, protective shell of a (western?) country that does not experience terrorist attacks on a regular basis. The terrorist threat is real. People die everyday in terror attacks. The "power-mad" government scenario? Lets just say in all this noise I've seen very few instances of actual abuse of power, if any.
I don't care what the percentage is. The right to snoop arbitrarily is too much power, and if we accept it will be abused, it will get worse, and it will be very hard to undo. Safety is not worth the permanent eradication of human dignity and liberty. I'd rather die than live like a chickenshit.
BTW, the government and the courts agree that warrantless snooping is dangerous, which is why there are checks and balances in place. The recent leaks raised doubts that those were not effective. But if the numbers in the declassified documents are accurate, it's nowhere nearly as bad as people here are imagining.
Put simply, databases change the equation. Never in human history has it been possible to store and mine such a massive quantity of data in perpetuity. If the infrastructure exists to capture and process that much data, included via algorithm rather than by human, that represents such a radical shift in power that its abuse is inevitable, regardless of whatever restraints may or may not exist today. What happens if the next Snowden is a profiteer? What happens if China finds a backdoor, and can suddenly tap all data on every American citizen.
Capturing and storing all data is really really fucking dangerous, no matter who's doing it or why. (I'm not thrilled about Google or Facebook either, but at least it's voluntary, and they never throw anyone into a cage.)
> People die everyday in car accidents and safety is an illusion, so let's stop wearing seatbelts.
I think a better analogy would be a seatbelt that you can't take off, "for your own good". And if you're trapped by your seatbelt in a crashed and burning car, well, too bad, we meant well!