“There’s no way to get there without a breakthrough,” OpenAI CEO Sam Altman said, arguing that AI will soon need even more energy.

  • Telodzrum@lemmy.world
    link
    fedilink
    English
    arrow-up
    9
    arrow-down
    9
    ·
    11 months ago

    General artificial intelligence has the potential to be actually useful. Generative AI, like Chat GPT (Generative Pre-Trained Transformer) absolutely does not. It’s a glorified autocomplete.

    • Blue_Morpho@lemmy.world
      link
      fedilink
      English
      arrow-up
      15
      arrow-down
      3
      ·
      11 months ago

      Chat GPT (Generative Pre-Trained Transformer) absolutely does not.

      I had to help my wife with an Excel VBA script. I know a few programming languages but I don’t know vba. Hours of googling turned up no useable scripts that did what I needed. ChatGPT wrote a working VBA script in seconds.

      Anyone who thinks it is a fancy auto complete hasn’t used it.

      • Womble@lemmy.world
        link
        fedilink
        English
        arrow-up
        5
        arrow-down
        1
        ·
        11 months ago

        It is a fancy auto complete, just an incredibly fancy auto complete. In the same way that a computer that can run VR games, simulate the evolution of solar systems let you access something close to the sum total of human knowledge over the internet is a very fancy pocket calculator.

        • fidodo@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          11 months ago

          I think it’s better explained as a search engine that works at the word level of granularity. It lets you do a word level search of all written human knowledge which allows it to adapt to your specific prompt. It’s the next step in searching knowledge based. First we had libraries, then document search, now it’s word search. I think it’ll be impactful on the same level as the creation of libraries and search engines.

      • Holzkohlen@feddit.de
        link
        fedilink
        English
        arrow-up
        2
        arrow-down
        2
        ·
        11 months ago

        I’d tried using it for python and it made up some fake library that does not exist to handle the complex task I was asking it do it. Utterly useless. The way it works is so complicated is simply has you fooled thinking it’s more than a fancy auto complete, but it is literally exactly that. Just the “smartest” auto complete there is. Don’t worry about it, at least you did not fall it love with it and then left your wife to be with ChatGPT, right?

        • Womble@lemmy.world
          link
          fedilink
          English
          arrow-up
          2
          arrow-down
          1
          ·
          11 months ago

          it’s a tool to be used like any other. If you are relying on it to produce working code with no effort on your part then yes it is likely useless. If you are using it to get you most of the way there its very useful. For example, I’m most comfortable writing python but have been doing a lot of c# recently, asking a code focused LLM questions like “how do you rename a column of a c# datatable” is so much quicker and more useful than trying to search through blogspam its not even funny.

    • R0cket_M00se@lemmy.world
      link
      fedilink
      English
      arrow-up
      13
      arrow-down
      2
      ·
      11 months ago

      As someone that works in IT and has seen the actual real world merit of LLMs, you guys clearly don’t work in a white collar field or you’d realize that you sound like someone in the 90’s claiming the internet is a fad.

      Microsoft continues to integrate it, and on prem data and cloud engineers are hooking it up to company resources for everything from helping with data creep to handling low level repetitive tasks.

      In the next ten years all bottom level data entry will be performed entirely by on premise AI models and the position of secretary will be a thing of the past, it’s not a gimmick like NFTs or crypto, it’s an actual tool that we are finding more uses for everyday.

      • fidodo@lemmy.world
        link
        fedilink
        English
        arrow-up
        3
        arrow-down
        1
        ·
        11 months ago

        VR and Crypto were bullshit, but AI is the real deal. On a side note I find it hilarious that out of the three options, Zuckerberg bet billions on the two wrong options. Unfortunately the rich have so much power that they can make catastrophic mistakes and still have plenty of money to finally bet on the right one.

        • R0cket_M00se@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          11 months ago

          How is VR bullshit? No Mans Sky, Elite Dangerous, Half Life Alex, more games everyday are using VR.

          Unless you mean the “meta verse” decentraland kinda deal then yeah for sure.

          • fidodo@lemmy.world
            link
            fedilink
            English
            arrow-up
            1
            ·
            11 months ago

            It being the future of computer interfaces is bullshit. I enjoy it as a novelty but Zuckerberg bet on it as the “next big thing”, on par with the Internet.

            • R0cket_M00se@lemmy.world
              link
              fedilink
              English
              arrow-up
              1
              ·
              11 months ago

              They never stopped to ask if it was actually an improvement and just assumed it’s what everyone wanted.

              Too bad some things are just more convenient on a screen in front of you. I don’t want to walk around a virtual grocery store, the website is fine.

    • HubertManne@kbin.social
      link
      fedilink
      arrow-up
      6
      ·
      11 months ago

      Its a little useful. Its another level of meta. initially research was libraries and books/docs but you could get synopsis from atlases, encyclopedias, dictionaries. Internet allowed for gopher searching and then web browsers allowed search engines but the results did not initially have synapses so you had to check each link. Then synapsis allowed looking through links and just checking the most promising ones. generative ai allows for the most promising links to be identified without perusal. One of the most useful things the ai assitants do is provide you the main links they got their info from.

    • Holzkohlen@feddit.de
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      3
      ·
      11 months ago

      Don’t know why you are being downvoted. This is exactly it. If ChatGPT can find you a solution, then that solution is already out there and it just stole it from some human. That is fundamentally how ChatGPT works, the damn thing can’t think.

      • dezmd@lemmy.world
        link
        fedilink
        English
        arrow-up
        2
        arrow-down
        1
        ·
        11 months ago

        How, exactly, do you think learning in humans works? We don’t just spontaneously know things, we build on others works and iterate improvements towards our learning and discovery. You didn’t just know English, you learned it from someone else who learned it ad infinitum until you get back to the earliest stages of complex chemicals we evolved from.

        When considered objectively and rationally, intellectual property is among the most insane imaginary bullshit humanity every came up with and actually adopted as a practice, but here we are. Thoughts and prayers.