Thread

1. nomilk · 1790396076 · focus · HN ↗

   Had the 'evidence' been a random dashcam rather than Flock cam, this wouldn't even be a story beyond severe police incompetence.

   Flock didn't fail to inspect the car for damage and not find any, flock didn't take 13 days to conduct a cursory review of evidence, flock didn't fail to request cell tower triangulation data and rule out her being in the location of the accident...

   I get that people feel their privacy's invaded by being filmed, but this story doesn't appear to have any more to do with Flock than what the officers had for breakfast or the brand of air freshener she had in her car at the time.

   1. selfhoster1312 · 1790403619 · focus · HN ↗

      A random dashcam does not serve the police someone's identity on a platter. It had to be a stupid "smart" object for such an inexcusable offense to be committed.

      With dumb camera footage, the cops looks at the plates, visual indicators, etc. That does not make them more competent or less prone to abuse. But here, Flock's software literally told them this woman was the one they were looking for, despite no evidence supporting this.

      This is 100% both on the police and on Flock. Some technologies don't deserve to exist. Facial recognition and ALPR is one of those, because the error rates are too high and the stakes are too high.

      1. nomilk · 1790405301 · focus · HN ↗

         > A random dashcam does not serve the police someone's identity on a platter.

         Nor does Flock - Flock merely provides a photo and timestamp, which should have been scrutised like any other evidence.

         The Flock data showed:

         - Car of same make/model nearby at similar time to the accident

         - Possible colour match (eyewitnesses gave varying reports on the colour of the perpetrating vehicle). But a match could have been ruled out in minutes had police inspected paint transfer onto the struck vehicle

         - A number plate non-match (first three characters were reported to 911 and did not match)

         - That's all!

         So police didn't do their job (scutinise the evidence), and possibly worse, they may have fabricated evidence (since police claimed Isaacs' vehicle was damaged, but it was in factory-new condition) - which may explain why they're shifting blame to Flock!

         1. selfhoster1312 · 1790408516 · focus · HN ↗

            That's exactly the same. Suggesting an output is the same as serving a suspect on a platter. I'm sure there's many studies about how humans (except statisticians) are not comfortable with confidence intervals.

            A computer software is supposed to deliver reliable, predictable, reproducible results. When it does not every now and then, it's a bug. When your entire system is built in "recognition" and "confidence" and leads to wrongful arrests, it's not a bug. You are just not making computer science. You are making a power abuse machine.

Unofficial Hacker News client; not affiliated with Y Combinator.