• nottheengineer@feddit.de
    link
    fedilink
    arrow-up
    54
    arrow-down
    1
    ·
    8 months ago

    The study claimed to be able to predict cannabis impairment using only the smartphone data with about 67% accuracy

    As someone who studied data science and did a good bit of machine learning in that time: That’s terrible and it means that the whole story is a big nothing burger. It’s barely better than random guesses that match the distribution.

  • webghost0101@sopuli.xyz
    link
    fedilink
    English
    arrow-up
    13
    ·
    8 months ago

    Thats alot of words for saying your smartphone is listening in and possibly watching.

    The article does nothing to detail what kind of smartphone sensors are used. I bet the gyroscope isn’t gonna be it when people are cough locked.

    But microphones are sound sensors and cameras can also fit the bill so…

    • nottheengineer@feddit.de
      link
      fedilink
      arrow-up
      9
      arrow-down
      2
      ·
      8 months ago

      You don’t even need sensors, you can tell from the websites/apps people use and especially from their typing behaviour. I can do 50WPM on my phone when sober, but less than half that when I’m not.

        • nottheengineer@feddit.de
          link
          fedilink
          arrow-up
          1
          ·
          8 months ago

          We don’t know exactly. That’s why it’s called machine learning.

          But we do know that it doesn’t do it well, 67% accuracy can be achieved by guessing if you try often enough.

      • webghost0101@sopuli.xyz
        link
        fedilink
        English
        arrow-up
        5
        arrow-down
        1
        ·
        8 months ago

        That i could see but not everyone who is stoned is typing on their phone all the time. Article makes it sound like it can do this automatically unobstructivly.

        Seems far easier to detect the sound of a lighter, crunchers hitting eachother and detecting sentence like “i am going outside to smoke my joint” combined with daily/weekly patterns in time.

        Thanks to this i did find a great nerdy app to play around with my phones electromagnetic field sensor though.

  • owenfromcanada@lemmy.world
    link
    fedilink
    English
    arrow-up
    11
    ·
    8 months ago

    To be fair, the senseless philosophical question you texted your buddy combined with the $65 mobile order for taco bell are pretty easy to interpret.