Glitch in Cuffs -Detroit Police 2020 Case-
In January 2020, the Detroit Police arrested a man called Robert Williams for stealing luxury watches. The whole thing turned out, however, to be a false accusation, and the only “evidence” against him was a doubtful facial‑recognition match.
The Police had taken a blurry security footage from the crime and used an AI system to compare it against 49 million driver’s license photos in the database. The system judged Williams’ old license photo to be the guilty one, and instead of verifying the lead, officers placed his photo with five other Black men and presented it to a store employee who didn't even witness the crime. This identification alone led to his arrest.
Even though charges were dropped after Williams spent 30 hours in a crowded jail, the police refused to apologize or reconsider their use of the AI system. Williams sued the city, and so Detroit Police can no longer arrest anyone based only on facial‑recognition results or use photo lineups from AI systems without independent evidence. Officers now need to be trained on the system's racial bias and high error rates, and courts are to be informed whenever facial recognition is used in an investigation.
I think these are legitimate countermeasures, and am relieved to see that at least for some cases there are actual changes made. However, arresting some random citizen by relying completely on unsure systems that are apt to make errors should never be allowed, and not just us normal people but also authorities need to be careful not to get carried away by AI tools.
-References-
Why Police Must Stop Using Face Recognition Technologies | TIME
Comments
Post a Comment