Llama 2 is about as factually accurate as GPT-4 for summaries and is 30X cheaper | Anyscale
www.anyscale.com
From the creators of Ray, Anyscale is a framework for building machine learning applications at any scale originating from the UC Berkeley RISELab.

Its interesting to see so much investor money chase AI unicorns at the moment. Something tells me many of them might be making the wrong bets.

  • niemcycle@lemmy.worldEnglish
    211·
    1 year ago

    I feel like with a lot of new technology, you have the frontrunner who proves such a thing is possible, then you have those who build on that work to make things cheaper and more efficient, etc.

    • FaceDeer@kbin.social
      10·
      1 year ago

      Indeed. The first mover often doesn’t reap the benefits of their move. They have to spend a lot of resources trying out dozens of wrong ideas before they manage to find the right one, and then everyone else can just look over at what they’re doing and get started straight away on that. Often with a nice fresh clean start that can be more useful as a foundation going forward from there.

      On the one hand that feels sadly unfair, but on the other hand OpenAI abused the concept of being an open non-profit to get to where they are so I don’t really have any sympathy.

      • Potatos_are_not_friends@lemmy.worldEnglish
        3·
        1 year ago

        I’m trying to remember the book I was reading. First mover advantage… Only something like 35% of first movers stay in business after two years.

        In two years… OpenAI is definitely going to be absorbed by a big company, probably Microsoft. But by that time, Facebook and Google would have had strong contenders ready to go.

        • fresh@sh.itjust.worksEnglish
          3·
          1 year ago

          What’s the baseline though? If only 10% of non-first movers in a new industry stay in business, being a first mover is still a comparative advantage.

      • TheDarkKnight@lemmy.worldEnglish
        2·
        1 year ago

        Eh whether legit or not they tried to seek funding through alternative means like educational institutions but were getting nowhere.

        As much as it sucks, the Microsoft deal made a ton of sense for both parties, they got the massive amount of compute needed to train their models and MS got AI.

        Truth is they are an AI company and it is EXPENSIVE to push that tech. Idk if they could’ve done it a better way, maybe but no way they’d be where they are at now without MS.