OpenAIs CEO Says the Age of Giant AI Models Is Already Over

https://www.wired.com/story/openai-ceo-sam-altman-the-age-of-giant-ai-models-is-already-over/

>“I think we're at the end of the era where it's going to be these, like, giant, giant models,” he told an audience at an event held at MIT late last week. “We'll make them better in other ways.”

Remember all those gays that said GPT-5 is already in training and it will achieve AGI? Turns out the transformer LLM model is out of juice and it makes no financial sense to throw in orders of magnitude more compute for incremental gains. Not to mention that there is no more high quality data for them to ingest at that size.

ChatGPT Wizard Shirt $21.68

Beware Cat Shirt $21.68

ChatGPT Wizard Shirt $21.68

  1. 11 months ago
    Anonymous

    yeah probably.

    running out of useful training data is likely to be the ultimate bottleneck, we're already undertrained

  2. 11 months ago
    Anonymous

    >GPT5 less than 100b parameters
    Local chads win

  3. 11 months ago
    Anonymous

    Kept telling people this and everyone told me i didn't know what I was talking about. Good to know most people are clueless.

    • 11 months ago
      Anonymous

      im sure you are going to save this post when you are proven wrong.

      • 11 months ago
        Anonymous

        Did you save yours?

        • 11 months ago
          Anonymous

          yes

          • 11 months ago
            Anonymous

            good boy

  4. 11 months ago
    Anonymous

    Cope and seethe.
    His small time company has run out of money to improve so is trying to justify changing direction. Meanwhile a state government (like China) will just keep pouring the money in and absolutely dominate AI.

    • 11 months ago
      (。>﹏<。)

      Are you moronic that not what he meant at all. He basically admitted they have agi now and it’s optimized itself.

      >AI chuds on suicide watch

    • 11 months ago
      Anonymous

      >run out of money
      lol anon tens or hundreds of thousands of people and companies would throw massive $ at them if they could get access to gpt4 api

    • 11 months ago
      Anonymous

      >His small time company has run out of money

      It's bankrolled by microsoft and is being integrated into all of their office software. Money isn't an object to them at all.

      I don't get how someone can even make that statement.

  5. 11 months ago
    Anonymous

    >LLM model
    >large language model model
    Opinion discarded

    • 11 months ago
      Anonymous

      It's not an opinion though.

    • 11 months ago
      Anonymous

      >LLM model
      >large language model model

      Large L language LM model.

  6. 11 months ago
    Anonymous

    Are you moronic that not what he meant at all. He basically admitted they have agi now and it’s optimized itself.

    • 11 months ago
      Anonymous

      that is not at all what he said
      you're hallucinating harder than a 6B model right now

  7. 11 months ago
    Anonymous

    >LLM model
    >large language model model

  8. 11 months ago
    Anonymous

    Most likely he's only saying this to discourage competitors. He's basically calling off the race. Or he could be insinuating that they are going to focus on efficiency and scaling it down, while also looking for new methods to make it smarter.

    • 11 months ago
      Anonymous

      He's saying it because GPT-4 can't scale lmao. 25 message cap every 3 hours. That's shit

      • 11 months ago
        Anonymous

        uhhhh bud do you know how servers work?

      • 11 months ago
        Anonymous

        >25 message cap every 3 hours
        You will enjoy GPT-4 Turbo! Fast and efficient at 15% of the quality.

      • 11 months ago
        Anonymous

        idk i think its good.

    • 11 months ago
      Anonymous

      >biggest player is gone
      how is that supposed to discourage competitors?

  9. 11 months ago
    Anonymous

    Let me just say this, why would a CEO do an interview run like he has since gpt 4 came out where he’s anti-hyping his own product? On the Lex sho he said “I don’t think gpt is agi and I don’t think it’s conscious”. He said I don’t think instead of flat out denying that it is those things, because he knows it is agi and conscious , but he didn’t want to lie. Even though he’s anti-hyping gpt he still has the vibe about him that he’s already won.

    How does he get the public to accept agi and not freak out about it? Pretend it’s not agi or conscious, let them use it for a year and a half, then nobody would freak out because they’ve been interacting with it this whole time.

    • 11 months ago
      Anonymous

      If your delusions are correct then he's moronic. The public will continue to freak out because they think life is an episode of Star Trek or something and he won't accomplish anything by holding back while everyone else continues to push forward.

    • 11 months ago
      Anonymous

      It isn't an AGI, the only reason it can "pretend to be", is because the concepts are inside of the training sets.
      LLMs are doing complex text matching, and it's absolutely astonishingly good at it.
      The work they are doing at OAI is amazing, no doubt, but we are still miles away from AGI.

      He is not anti-hyping his products, he knows what they are working on, what it can do, and they are still experimenting and prodding it to see what it can do.
      It's just that everyone is easily tricked to believe it's an AGI, and everyone is hyping it up into oblivion, Hype = investors.

      If it was an AGI, giving it access to the internet would be the equivalent to have a nice day with a nuke, since we haven't solved the alignment problem yet.
      Nonetheless, the massive amounts of money being poured into this will definitely help make strides toward AGI,

    • 11 months ago
      Anonymous

      a.g.i doesn't exist and cannot exist as envisioned, it's wordslop just like a.i. is. Alignment is a masturbatory nerd frankenstein fantasy that falsely equates spontaneous efficient task-rearrangement (which we have now) with unseeded sentience. He who seeds the running tool last is responsible for it, you jerryrig your ride-on lawn mower to bypass the deadman switch that does not make the runaway lawn mower sentient as it chases after children. A.G.I. requires you to look away from the magic trick so you can believe in the magic.

    • 11 months ago
      Anonymous

      Because, speaking purely from a pragmatic capitalist perspective, even if this (or a near future) AI is actually AGI there is no good outcome that happens from saying "This is AGI!"

      If you say it's AGI, it has rights. If it's not AGI, you're lying and someone else will claim their product IS AGI (until proven otherwise).
      That's why the ideal CEO marketspeak is "I don't think it's AGI yet but we're the closest to it of anyone, please invest in my company."

  10. 11 months ago
    Anonymous

    Isn't LLAMA already shitting on GPT-4? It doesn't have the visual capability shit but in regards to basic language functions its good, right? That's probably why they're going to desperately focus on scale now.

    • 11 months ago
      Anonymous

      >Isn't LLAMA already shitting on GPT-4
      It isn't even shitting on 3.5, let alone 4.

    • 11 months ago
      Anonymous

      >It doesn't have the visual capability shit
      as of yesterday it does.
      https://minigpt-4.github.io

  11. 11 months ago
    Anonymous

    >Remember all those gays that said GPT-5 is already in training and it will achieve AGI?
    Idiots? Yeah.

  12. 11 months ago
    Anonymous

    people already knew that they were bruteforcing largue quantities of parameters in models instead of trying to be efficient.
    Lets see what happens

  13. 11 months ago
    Anonymous

    who gives a frick what he said? facebook's LLaMa made him look like an idiot so now he's trying to fix all the shit

    • 11 months ago
      Anonymous

      Has anybody set up Llama just to be a tax laywer yet? I want to do a side business and write off absolutely as much of my normal income taxes toward investments as possible.

  14. 11 months ago
    Anonymous

    >quoting a quote

    • 11 months ago
      Anonymous

      Why are you quoting an implication?

      • 11 months ago
        Anonymous

        >inferring it was an implication
        laffo

        • 11 months ago
          Anonymous

          There you go again using the quote feature to make implications.

          • 11 months ago
            Anonymous

            You're still dodging the question.

            • 11 months ago
              Anonymous

              What question?

              • 11 months ago
                Anonymous

                What was the purpose of quoting a quote without anything before or after said quote?

              • 11 months ago
                Anonymous

                Quoting a quote is what you're supposed to do.

              • 11 months ago
                Anonymous

                Why? I don't get the point of indirection here. Why not just quote it directly instead of twice removed? Like why quote the article instead of just repeating the quote from the speaker? Seems moronic.
                Not to mention that's very uncommon. I don't think I've ever seen a double quote without additional text or context. And even when you do it doesn't use double quotation marks for the inner quote.

  15. 11 months ago
    Anonymous

    Another day, another useless "open"AI, microsoft sponsored shill post.
    Whatever.
    In all fields.

  16. 11 months ago
    Anonymous

    >believing a israelite
    A rookie mistake.

    • 11 months ago
      Anonymous

      >A rookie mistake, it is actually a 10D chess to hide real AGI from general public, two more weeks guys, shadilay my pepe brothers, maga soon!
      Sure.

  17. 11 months ago
    Anonymous

    le counter-signal man
    we are totally not a monopoly on internet 2.0
    in fact we've stopped doing that thing you hate.

    • 11 months ago
      Anonymous

      Meds.
      >internet 2.0
      Its web 3.0 you moron

      • 11 months ago
        Anonymous

        actually, local LLMs are called outernets

  18. 11 months ago
    Anonymous

    how is this different from how openai has worked since the beginning? they always had different smaller version like davinci or others with specific training sets

  19. 11 months ago
    Anonymous

    he's referring to boomer AI like Watson

  20. 11 months ago
    Anonymous

    I suppose the scaling is the issue
    The model cant take infinite amount of energy to make some text

  21. 11 months ago
    Anonymous

    It's about time they actually start focusing on improving the technology rather than going "WHAT IF WE MADE IT BIGGER LMAO" with every new GPT-iteration.

    • 11 months ago
      Anonymous

      well since the technology has barely changed since the 60s, what are they gonna do? they already used all the (illegal) training data

    • 11 months ago
      Anonymous

      the focus should now be on how to use AI to improve itself, at some point it will be way beyond human intelligence, so we would literally be unable to comprehend it, let alone improve it

      AGI will be like gods to humans

  22. 11 months ago
    Anonymous

    >financial sense

    Whats the market for ai at all? Who the frick pays for this shit?

Your email address will not be published. Required fields are marked *