• The headline is wrong. They don’t claim Gemini was trained on docs.

    " It also seemed to be able to regurgitate accurate information related to unreleased mechanics that the developer says they had only written into GDocs the day before the player’s AI interaction."

    If it was written only the day before there’s no way the model was trained on that data. The claim is Gemini accessed his docs while generating the answer. We know Gemini can access public docs so I think the question here is if it accesses private docs in a breach of security, the doc had wrong permissions or it was all just a coincidence.

  • elucubra ( elucubra@sopuli.xyz ) 
    link
    fedilink
    arrow-up
    15
    ·
    4 months ago

    Just as I have proof that Android phones are listening and sending our conversations constantly, I’m willing to bet that Google is scraping all our Google docs, files in drive, and Gmail emails.

    • infeeeee ( infeeeee@lemmy.zip ) 
      link
      fedilink
      arrow-up
      14
      ·
      4 months ago

      What proof lol.

      Google already had all the docs, they never said it was encrypted, as they also scan all the emails for “spam detection”. It would have been more surprising if they didn’t use it for training

        • infeeeee ( infeeeee@lemmy.zip ) 
          link
          fedilink
          arrow-up
          6
          ·
          4 months ago

          Could you elaborate on that? Because it was refuted countless times that current technology can’t do that, and I want to be up to date on the latest unhinged conspiracy theories.

          • “They are always listening” isn’t exactly unhinged. Infact it’s a bit unhinged to think they can’t listen through current technology when that shit literally listens to people talking in the vicinity.

            "The majority of the commands we issue to smart speakers are innocuous, perhaps asking to play a particular radio station or set an alarm. But whistleblowers have expressed concern that devices are waking up randomly and processing accidental recordings that should never have been made. 

            These are known as “false accepts” and data collected by one Google whistleblower shows that about 15 per cent of recordings occur by accident. This error rate indicates that smart speakers – essentially microphones in our homes – may not be as benign as we’ve been led to believe. “These technologies live in your home, they become part of your daily fabric,” says Schaub. “When you phone a call centre, you typically get a message explaining that the call may be recorded for training purposes, but that explicit acknowledgment is missing from smart speakers.”

            https://www.thenationalnews.com/arts/are-big-tech-companies-actually-listening-in-on-us-via-our-phones-or-smart-speakers-1.898574

            This is one source. 15% is too high even for accidental activations.

            Edit: Found the most relevant article. I heard about this one maybe 6 months ago:

            https://www.zdnet.com/article/new-report-claims-your-phone-tv-and-smart-speaker-are-spying-on-you-but-is-it-true/

            • infeeeee ( infeeeee@lemmy.zip ) 
              link
              fedilink
              arrow-up
              8
              ·
              4 months ago

              Yes, I know about this, but bugs in wake word detection is not “listening and sending our conversations constantly”.

              And on device STT capabilities will get better with each new generation of TPUs, so we are not far from ditching wake word detection as a technology, and phones will be able to do type everything they hear. But currently we are not there yet.

              • The tech is there for certain. That part is super easy. It would just be putting the software in place. Also, it would be far more covert and energy efficient to simply record the radio onto the phone itself and then send it as a file every so often. Only recording\saving when voice is being picked up could be done in as little as 30MB a day.