In reality nobody gives a shit about the false negatives. If you want to make a statistical argument you have to base it on the false positives. It looks like there is no such thing as a "false positive" in Google. They've been hiring puzzle-solvers and phone-dictation-coders for 10 years, only to suddenly turn around and say "puzzles are not a good predictor for success, nevermind we used it for 10 years, still no false positives".

I doubt Google's false positives are better than anybody else's.