For facial recognition experts and privacy advocates, the East Bay detective’s request, while dystopian, was also entirely predictable. It emphasizes the ways that, without oversight, law enforcement is able to mix and match technologies in unintended ways, using untested algorithms to single out suspects based on unknowable criteria.

  • Catsrules@lemmy.ml
    link
    fedilink
    arrow-up
    35
    ·
    1 year ago

    Didn’t facial recognition get some poor guy arrested and raped in prison and he was completely innocent of everything?

    • Ignotum@lemmy.world
      link
      fedilink
      arrow-up
      7
      ·
      1 year ago

      Considering how common rape is in American prisons and how often innocent people are getting locked up, that does not sound unlikely

    • agent_flounder@lemmy.world
      link
      fedilink
      English
      arrow-up
      9
      ·
      edit-2
      1 year ago

      Propublica did an article on that.

      https://www.propublica.org/article/understanding-junk-science-forensics-criminal-justice

      E.g.

      The reliability of bloodstain-pattern analysis has never been definitively proven or quantified, but largely due to the testimony of criminalist Herbert MacDonell, it was steadily admitted in court after court around the country in the 1970s and ’80s. MacDonell spent his career teaching weeklong “institutes” in bloodstain-pattern analysis at police departments around the country, training hundreds of officers who, in turn, trained hundreds more.

      In 2009, a watershed report commissioned by the National Academy of Sciences cast doubt on the discipline, finding that “the uncertainties associated with bloodstain-pattern analysis are enormous,” and that experts’ opinions were generally “more subjective than scientific.” More than a decade later, few peer-reviewed studies exist, and research that might determine the accuracy of analysts’ findings is close to nonexistent.

  • iAvicenna@lemmy.world
    link
    fedilink
    arrow-up
    27
    arrow-down
    1
    ·
    1 year ago

    wow nice to know that from DNA you can predict whether or not a person has a beard, or their style of hair

    • hansl@lemmy.world
      link
      fedilink
      arrow-up
      4
      ·
      1 year ago

      Even from a perfect witness (and witnesses are very imprefect) you wouldn’t be able to predict if they have a beard or not. That’s why you always multiple variations of the person when they actually distribute renditions.

    • Wage_slave@lemmy.ml
      link
      fedilink
      arrow-up
      2
      ·
      1 year ago

      I was was wondering what I’d look like with a sick tat on my face. And behold, the DNA and AI winning combination knew it, before I ever got it.

  • FluffyPotato@lemm.ee
    link
    fedilink
    arrow-up
    10
    ·
    1 year ago

    Getting a psychic to give them a suspect through the shadow realm or something would probably be more accurate.

  • dan1101@lemm.ee
    link
    fedilink
    arrow-up
    7
    ·
    1 year ago

    Boy that is just a garbage sandwich, garbage in garbage out with twice as much garbage.

  • The Pantser@lemmy.world
    link
    fedilink
    arrow-up
    6
    arrow-down
    5
    ·
    1 year ago

    I can see the abuse but what if this actually worked in a best case scenario? So dna is found say from a rape and that DNA is used to create a image of the person and then they find that person and then do DNA tests to match them. The image is not used as evidence but used to find the person. Honestly it seems like a good use, if it’s limited to that.

    • i_love_FFT@lemmy.ml
      link
      fedilink
      arrow-up
      18
      ·
      1 year ago

      I don’t know a lot about DNA, but i know about facial recognition.

      Facial recognition is highly inaccurate. It would be easy for people from the same country to “match” at facial recognition despite being totally unrelated.

      If “face generation from DNA” is only roughly accurate (ex: nose size or skin tone), then anybody from the same ethnic origin could be a match. Basically, the more you look like the “average person”, the more likely you would fit the generated face.

      Doesn’t it sound a lot like technology-enabled profiling?

      • fidodo@lemmy.world
        link
        fedilink
        English
        arrow-up
        3
        ·
        1 year ago

        I think this is a bad idea, especially the way it’s being developed, but let me play devil’s advocate for a second. What if it were only used to narrow a search radius, the same way cell pings are used to narrow a search radius? Cell pings are already used to direct resources. Being near a crime obviously doesn’t mean you committed the crime, but it does narrow down where to look, and once you start looking you can find real evidence more efficiently. You could pair this with other techniques to narrow down the search, and then find real hard corroborating evidence. Also, since they need DNA in the first place they’d need a DNA match from the suspect preventing random people from getting charged.

        Now to stop playing devil’s advocate, there are just so many ways this can be abused, and the police are the worst organization to lead it. They are not technology experts, they’re not even legal experts, and they’ve been shown over and over again to be easily biased, so even if they need corroborating evidence, that doesn’t mean they won’t be biased by the face match and then “find” evidence, or even plant it, plus, even just being accused can be hugely disruptive, and traumatizing when they target a non match. Imagine you’re innocently going about your day and you suddenly get snatched up for questioning and forced to give a DNA sample.

        If anything like this were to be used in any way you would need so many safe guards and it’s obvious the police don’t care about setting any of those up. You’d need a double blind approach to evidence gathering, extreme oversight, and a very restrictive legal framework and course close guarding and anonymization techniques on any personal data, and probably more things I’m not thinking about. The police are so irresponsible to treat this like a whatever thing that isn’t incredibly sensitive and invasive and needing tons of safe guards to not be a massive violation of privacy and a dangerous biasing vector that could easily catch up innocent people.

    • fidodo@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      ·
      1 year ago

      Don’t assume it wouldn’t be abused since the police have a shit track record. If anything like this we’re to be used then strict laws restricting how it can be used need to come first since the police are dumb and can’t be trusted to invent new applications of technology. They’re the last group that would be leading this.

    • OfficerBribe@lemm.ee
      link
      fedilink
      arrow-up
      2
      ·
      1 year ago

      I am on the same boat. Somewhat similar thing was already done by those forensic sketch people that drew how a person might look like after x years if they had earlier photo. It’s not like those sketches meant they are irrefutable proof, just a method to potentially find what you are looking for and then investigate further.

      That said, this feels like grasping at straws. I cannot fathom how with only DNA sample you could get an accurate portrait that face recognition could then match.