• xapr [he/him]
    link
    fedilink
    English
    21
    edit-2
    5 days ago

    I only read the text, didn’t watch the video, but from the text it didn’t sound like that’s what they were describing. It sounded like they will present images to people and ask them to confirm whether or not there’s an object there?

    We introduce “MapTCHA”, a CAPTCHA that leverages the uncertainty of interpreting imagery with computer vision, and provides human verification for AI predictions: users are asked to identify images containing correctly interpreted objects, e.g. building outlines.

    Edit: also, here’s the github they seem to be working from - https://github.com/ciupava/maptcha_dev

    Edit 2: I hope they succeed, because it would be great to have an open source captcha that benefits everyone.

    • Em Adespoton
      link
      fedilink
      English
      25 days ago

      The big question though is: why?

      CAPTCHA was implemented because machine learning was worse than people at solving the problems, so they could filter out the bots AND train their algorithms.

      Today, the bots are often better than the humans, so both reasons for having CAPTCHAS no longer make sense. Even user verification can be easily faked by AI now.

      • xapr [he/him]
        link
        fedilink
        English
        45 days ago

        I’m not sure.

        Perhaps getting computer vision that’s reliably better than humans costs a lot of money that a project like OpenStreetMap doesn’t have?

        • Em Adespoton
          link
          fedilink
          English
          35 days ago

          Could be… it makes sense to reinvent the wheel if the previous inventors won’t share.

      • ubergeek77
        link
        fedilink
        English
        35 days ago

        They explain it in the video. They already use algorithms to detect if things are buildings or not.

        But if their algorithm can’t make a determination or is uncertain below a certain threshold, they send it to Maptcha to get a bulk human opinion.