3 ms·

I wouldn't want facial recognition software to identify them because I can't understand its failure cases. If it is allowed in court as evidence, prosecutors wi

by verall 5y ago

I wouldn't want facial recognition software to identify them because I can't understand its failure cases. If it is allowed in court as evidence, prosecutors will talk up the recognition as "state-of-the-art" and juries will be influenced by its opinions.

ggreer 5y ago
Do you also apply this reasoning to fingerprinting, DNA analysis, tire prints, and ballistics comparisons? Like most people, I don't understand all of the failure modes involved with those technologies, but they do seem to be helpful tools for bringing violent criminals to justice. I also think eyewitness testimony has many failure modes. If anything, it's probably less accurate than current facial recognition tech and biased in ways that are harder to determine. Yet I wouldn't want to ban all use of eyewitness testimony. Banning facial recognition seems like overkill. Instead, it makes more sense to restrict it so that we can get the good parts (catching violent criminals) without the bad parts (oppressive surveillance state). Instead of banning all fingerprinting and DNA analysis, we have rules for how police can use them. Why not have similar rules for facial recognition?
- verall 5y ago
  Yes, I do. Ballistics comparisons in particular been found to be nearly complete bunk, yet experts continue to use them in court.[0] Tire prints are also pretty bad.[1] We live in an age of unprecedented HD video surveillance. We don't need automated facial recognition to bring violent criminals to justice. [0] https://www.sciencemag.org/news/2016/03/reversing-legacy-junk-science-courtroom https://www.sciencemag.org/news/2016/03/reversing-legacy-jun... [1] https://www.apmreports.org/story/2016/09/27/questionable-science-tire-track-shoe-print https://www.apmreports.org/story/2016/09/27/questionable-sci...

3 ms·

I wouldn't want facial recognition software to identify them because I can't understand its failure cases. If it is allowed in court as evidence, prosecutors wi

by verall 5y ago

ggreer 5y ago
Do you also apply this reasoning to fingerprinting, DNA analysis, tire prints, and ballistics comparisons? Like most people, I don't understand all of the failure modes involved with those technologies, but they do seem to be helpful tools for bringing violent criminals to justice. I also think eyewitness testimony has many failure modes. If anything, it's probably less accurate than current facial recognition tech and biased in ways that are harder to determine. Yet I wouldn't want to ban all use of eyewitness testimony. Banning facial recognition seems like overkill. Instead, it makes more sense to restrict it so that we can get the good parts (catching violent criminals) without the bad parts (oppressive surveillance state). Instead of banning all fingerprinting and DNA analysis, we have rules for how police can use them. Why not have similar rules for facial recognition?
- verall 5y ago
  Yes, I do. Ballistics comparisons in particular been found to be nearly complete bunk, yet experts continue to use them in court.[0] Tire prints are also pretty bad.[1] We live in an age of unprecedented HD video surveillance. We don't need automated facial recognition to bring violent criminals to justice. [0] https://www.sciencemag.org/news/2016/03/reversing-legacy-junk-science-courtroom https://www.sciencemag.org/news/2016/03/reversing-legacy-jun... [1] https://www.apmreports.org/story/2016/09/27/questionable-science-tire-track-shoe-print https://www.apmreports.org/story/2016/09/27/questionable-sci...