[fpgaminer]

That's been my feeling. Last I looked I recall public SOTA in facial recognition being something like an error every 1 in 1000 cases. Assuming that a private corporation's efforts are somehow 100x better that still gives you 3,000 suspects in America for any given photo. (And all those people are likely to look very, very similar.)

Hell, Apple's FaceID makes a mistake every million faces, and that system is both professional and has an order of magnitude more data to work with from the FaceID scanner. Clearview is just using blurry photographs.

[DangerousPie]

Why is that a problem? This is just the first step to generate a set of possible matches, there will always be further steps to narrow it down further.