,
5 tweets,
6 min read
Read on Twitter
(1) @NYPDONeill is in the @nytimes today, defending the #NYPD’s #FacialRecognition, but his defense left us with more questions than answers. S.T.O.P. and other advocates have challenged the #NYPD’s facial recognition dragnet. #NYC needs reforms, not PR. nytimes.com/2019/06/09/opi…
(2) The truth is that the #NYPD’s #FacialRecognition system is biased and broken. Their database will return more than 200 possible “matches” for every single photo. Meaning that AT BEST it is wrong 99.5% of the time. All too often, NONE of the “matches” will be the real suspect.
(3) #FacialRecognition can be a powerful tool, but only when used correctly. Per @GeorgetownCPT, #NYPD artists #Photoshop images before testing them, altering the faces and copy/pasting #celebrity photos. This isn’t #ForensicScience; it’s an art project. law.georgetown.edu/privacy-techno…
(4) Even if the #NYPD didn’t alter photos, even if they used scientific methods, it still wouldn’t overcome the fact that #FacialRecognition is #Biased. This broken #technology is far more likely to get it wrong for women and #NewYorkers of color. media.mit.edu/articles/facia…
(5) @NYPDONeill’s most alarming admission: 7,024 requests in 2018 alone and 1,851 so-called “matches”. Thousands of New Yorkers of color targeted by the latest form of digital #StopAndFrisk. Terrifying to think how many New Yorkers will be “facially profiled” in the years to come
