Google rolled out AI overviews across the United States this month, exposing its flagship product to the hallucinations of large language models.

Google rolled out AI overviews across the United States this month, exposing its flagship product to the hallucinations of large language models.

  • DominusOfMegadeusEnglish
    1191·
    6 months ago

    Testing in Prod. Stay classy, Google.

    “The vast majority of AI Overviews provide high quality information, with links to dig deeper on the web,” said a Google spokesperson in an emailed statement to Gizmodo, noting many of the examples the company has seen have been from uncommon queries.

    This is entirely fair. There is no way that anyone at Google could have anticipated that humans would search for strange things on the internet.

    • hydroptic@sopuli.xyzEnglish
      31·
      6 months ago

      The vast majority of AI Overviews provide high quality information

      According to some fuckwitted Google rep, and I wouldn’t trust them any further than I could throw them.

      • DominusOfMegadeusEnglish
        2·
        6 months ago

        I wouldn’t trust them not to kneecap me in the parking lot after they interview

    • Uriel238 [all pronouns]@lemmy.blahaj.zoneEnglish
      12·
      6 months ago

      Although any sociologist or veteran of the internet will tell you humans will engage in any exploit that yields a funny result. The Diet Coke + Mentos rule.

      And that means we’ll actively search for hilarious Google AI responses.

      Google is so f double-plus filthy rich, it is obligated to run its projects by experts or be relentlessly mocked. So it should have known this was the outcome.

      Unless this is 5D chess and Google is willfilly using itself as a cautionary tale to discourage future webservice sites from arbitrarily inserting AI into its features.

      • Agrivar@lemmy.worldEnglish
        8·
        6 months ago

        Unless this is 5D chess and Google is willfilly[sic] using itself as a cautionary tale to discourage future webservice sites from arbitrarily inserting AI into its features.

        Holy shit, can I live in that timeline, please?!? Pretty please?

        • Uriel238 [all pronouns]@lemmy.blahaj.zoneEnglish
          3·
          6 months ago

          What came to mind was the New Coke reformation of 1985, which seems like a brilliant idea (in retrospect) only if you imagine it a 5D chess move to get people to panic because the Coke Classic was about to be discontinued. In fact, the marketers of Coca-Cola since admitted they thought they were getting killed by Pepsi and were sincere in their new release, but the company was able to not only recover thanks to responding to public feedback, so they sold Coke Classic, and New Coke and to this day it’s a popular soft drink. (The company still does shit things like employing death squads to keep their offshore workers scared of unionizing, so it’s still a typical large publicly traded multi-national corporation)

          Coca-cola marketing didn’t have that kind of foresight, but there’s a tiny chance that some folks at Google have that kind of hindsight, knowing Google could absolutely afford a ploy like willfully goofing up and then recovering with aplomb by listening to the public. It’s also a way to sneak such a ploy past the shareholders by insisting they were sincere in their implementation of AI at the time.

          Unlikely, of course. Most of the time the upper management of big companies are glad to just half-ass everything. But it would make a cool movie at least.