• redballooon@lemmy.world
    link
    fedilink
    arrow-up
    12
    ·
    24 hours ago

    They have the best models that are currently available. They’re genuinely usable in limited areas. That’s something. It’s not the vastly overpromise they like to brag with, but it’s certainly not worthless.

    How much this is worth exactly the market is still determining.

    • EightBitBlood@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      ·
      22 hours ago

      The key word is “currently available”

      The market will probably react the same way it did with Blackberry, Diamond Rio, Yahoo, and every other company that invented tech first and then spent their dollars promoting it as the best out there rather than R&Ding it into the better versions that were eventually widely adopted. (iPhone, iPod, Google).

      Hostorically speaking, all of these companies are going to fail once a single competitor with an easier to use product who spends less can outperform them. Which is already a very visible problem with AI and local models.

      IMO. These companies all spent Trillions on tech that will cost most people only hundreds to eventually get the same performance running custom models locally.

      • mad_lentil@lemmy.ca
        link
        fedilink
        English
        arrow-up
        2
        arrow-down
        1
        ·
        21 hours ago

        They’ll never ever, ever. Ever. Ever… Ever. be able to make up their insane capex.

        As soon as the bubble pops some competitor without all their overleveraged baggage will come in and buy up all their assets for pennies on the dollar and eat their lunch.

        Although even then it’s hard to compete with basically $0/token that local models offer.

    • UnderpantsWeevil@lemmy.world
      link
      fedilink
      English
      arrow-up
      5
      arrow-down
      2
      ·
      edit-2
      22 hours ago

      They have the best models that are currently available.

      That doesn’t justify a $2T market cap. That’s two Berkshire Hathaways.

      How much this is worth exactly the market is still determining.

      Clueless Retail Investor: “How much weight can this rug hold?”

      Investment Bank: “Step on it and find out”

      JP Morgan had heavily divested from their Anthropic position by end of week. This isn’t market price exploration, it’s just a share dump.

    • bstix@feddit.dk
      link
      fedilink
      arrow-up
      2
      arrow-down
      1
      ·
      23 hours ago

      No doubt AI is worth something, but not this. Not this amount and not like this.

      In a good scenario they could make a service that is worth paying for when you need that special service of hundreds of computers crushing through data to get the most probable statistic results for whatever purpose that might be.

      The one company that succeeds in making a profit from its specialist service is going to bury the rest.

      It’s just not what’s happening currently. All the companies are betting on it becoming a mainstream search engine replacement for a broad user base, but it sucks at it and nobody wants to pay them for doing that. The value is entirely thought up because of the idea of a massive user base that doesn’t exist and likely never will.

      • redballooon@lemmy.world
        link
        fedilink
        arrow-up
        1
        ·
        20 hours ago

        Reminds me of “there is a world market of maybe two or three super computers” – IBM 1950something

        The demand for compute power is far from saturated.

        • bstix@feddit.dk
          link
          fedilink
          arrow-up
          1
          ·
          edit-2
          20 hours ago

          I guess it is, since they have to make up uses for it that nobody knows about or even requested.

          Personally I was fine with what we had in the 80s …

          and apparently so was the flight computers…

          They spent 40 years updating that shit and then it breaks. You know… it worked fine before that.

    • boonhet@sopuli.xyz
      link
      fedilink
      arrow-up
      1
      arrow-down
      1
      ·
      22 hours ago

      best models that are currently available

      That depends on your use case among other things. I find that Gemini 4 hydrogen and especially neon are better than Opus 5.x like 2 months ago or whenever I last used Claude. And Opus has a tendency to make things very expensive.

      Run gemini 4 hydrogen on the cheapest Google AI plan (which for the first 3 months is a couple of euros and then it’s the same as everyone else’s lowest tier) and you can have two parallel sessions without ever having to run out of the 5 hour limit (admittedly my workflow isn’t exactly “fire off 2 hour one-shot prompt jobs” so it’s not generating 100% of the time, sometimes it’s waiting for my input).