How the N.Y.P.D.'S Facial Recognition Tool Landed the Wrong Man in Jail
nytimes.com
nytimes.com
As a layman, it looks to me like the legal system is unprepared for the unintuitive issue that if you run a facial recognition search with a 0.000001 false positive rate on a database 10 million people, you get someone who:
1. Likely looks extremely close to the target
2. Has a 99% chance of not actually being the target
Eyewitness confirmation does little to shift that 99% because it'll be based largely on the same factors, rather than being independent evidence.
Even scarier with the black box that is AI…
How can I refute what a computer says if nobody knows how it came to a conclusion?