cross-posted from: https://lemmy.ml/post/47972724

i encountered this for the first time today while attempting to read something on archive.today.

i confirmed that decoding the qrcode using a computer and following the URL it contains is insufficient; the error it gave directed me here which is what the linked screenshot is of.

the old type of captcha remains available too, for now:

screenshot of text: Important: Mobile verification for Google Cloud Fraud Defense is an experimental challenge type in Preview. Visual and audio challenges are available as alternatives for users who can't complete mobile verification. To use them, click the Visual  or Audio  buttons.

screenshot of google support webpage with text: To complete the mobile verification, you must use a compatible mobile device. If verifying on iOS/iPadOS device with version 15.0-16.4, download the reCAPTCHA app in the App Store. After installation, return to the challenge to try again. If verifying on Android device with Google Play Services version older than 25.41.30, upgrade to the latest Google Play Services version. After upgrading, return to the challenge to try again. Scan to Verify You're Human. Use your phone's camera app to scan this QR code. This securely links your device. Mobile Verification. Verification will initialize the reCAPTCHA app on your device.
you are viewing a single comment's thread
view the rest of the comments
[–] 53 points 5 months ago (4 children)

Captcha has been one of the greatest google acquisitions ever.

They acquired it under the guise of improving OCR and have since morphed it into an AI data farm (how else is google lens gonna know what objects are what?) and now total insight into a users every single action from desktop to mobile, tying it all together into a surveillance nightmare.

I can guess the permissions that the recaptcha app needs now. Probably something akin to root access with all datapoints and considerations you could think of.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 20 points 5 months ago*

    I used to always add one incorrect tile and skip one correct tile.(It would still pass)

    I thiught I was such a rebel lol

    Then I figured, they'd be stupid if they didn't show the same image to multiple people...

  • source
  • parent
  • [–] 1 point 5 months ago (2 children)

    How would that teach Lens to recognise anything other than motorcycles and traffic lights really well?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 3 points 5 months ago (1 child)

    I've had many, many not traffic light and motorcycle/bicycle recaptchas. They're probably leaning a bit into self driving learning the past few years.

    Lens has a lot more data points nowadays after everyone's google photos was used for training for what, 10+ years at this point?

    Google harvested all human typed words 15 years ago with the google library project. They've been hoarding and processing data for models forever.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 5 months ago

    I was being at least partly facetious because I rarely get anything but motorcycles and traffic lights and even then it'll most likely ask me about buses or bridges. Not disagreeing that they're hoarding data :)

  • source
  • parent