Google unveils TurboQuant, a new AI memory compression algorithm — and yes, the internet is calling it 'Pied Piper' | TechCrunch
techcrunch.com
Google’s TurboQuant has the internet joking about Pied Piper from HBO's "Silicon Valley." The compression algorithm promises to shrink AI’s “working memory” by up to 6x, but it’s still just a lab experiment for now.
  • Brewchin@lemmy.worldEnglish
    20 minutes

    This should come in handy for the recently projected need for 300 GB RAM* in upcoming self-driving cars.

    *Not a typo. 😳

  • hopesdead@startrek.websiteEnglish
    2 hours

    Okay, but did Google calculate how many dicks thet could jerk off for maximum efficiency?

    • x00z@lemmy.worldEnglish
      46 minutes

      Yes it was part of their quarterly circlejerk.

    • spy@lemmy.dbzer0.comEnglish
      2 hours

      Well that depends on a lot of factors. One of them being the distance of dick to floor of every one they would jerk. Call that D2F.

      Hopefully they thought of it.

  • Deconceptualist@leminal.spaceEnglish
    3 hours

    Should be called Middle-Out? That was the algorithm IIRC. Pied Piper was the name of the startup.

    Either way, funny shit.

  • AbouBenAdhem@lemmy.worldEnglish
    5 hours

    TurboQuant, meanwhile, could lead to efficiency gains and systems that require less memory during inference. But it wouldn’t necessarily solve the wider RAM shortages driven by AI, given that it only targets inference memory, not training — the latter of which continues to require massive amounts of RAM.

    I didn’t realize the RAM shortage was mostly due to training—I would have thought inference was at least a big a factor.

    • Dran@lemmy.worldEnglish
      3 hours

      Inference is dirt cheap in comparison. Hundreds to thousands of concurrent users can be served by hardware costing in the high-thousands to low-ten-thousands.

      Training those same foundational models is weeks to months of time on tens to hundreds of millions worth of hardware.

      • AbouBenAdhem@lemmy.worldEnglish
        2 hours

        Yeah—but in theory you only need to train once, while inference costs are ongoing and scale up with usage.

        I guess it’s ultimately a business decision by AI companies to weigh how often retraining is worth the cost.

        • JGrffn@lemmy.worldEnglish
          1 hour

          Yeah i don’t think they ever stop training is the thing. At this point I’d assume they have multiple training pipelines to try different shit out, just queued up to hit the big farms as soon as the last models are done training.

          Resting isn’t a thing in capitalism.

    • uuj8za@piefed.socialEnglish
      3 hours

      Yes, it should be Nucleus. Them calling it PiedPiper is a propaganda campaign to try to earn good will from people. Fuck Google locking down Android.

    • hopesdead@startrek.websiteEnglish
      2 hours

      Sure but didn’t the plot line with Nucleus come in a later season?

      Secondly, I am pretty certain the Google logo was always in the opening credits.

    • cecilkorik@lemmy.caEnglish
      5 hours

      No, that is what it would be if we were using traditional, deterministic compression and using a reversible and verifiable mapping of data. But this is the new era of memetic compression, “Pied Piper” is what everyone remembers from the show, so we compress it to “Pied Piper” to minimize the amount of memetic overhead and allow the smallest possible compression artifact. Like with “AI”, it doesn’t need to be correct, just close enough for people to think it is! /s

  • SalamenceFury@piefed.socialEnglish
    4 hours

    Funny thing that came with this: apparently Micron’s stock fell off a cliff, and apparently so did RAM prices? Can’t confirm that later one.

    • BetaDoggo_@lemmy.worldEnglish
      2 hours

      Neither are true, Micron has been plummeting since their earnings report on the 18th. This might have caused a small dip but it’s nothing compared to the cliff they just fell off of.