This is AI

This is AI

A Conspiracy Theorist Is Talking Shirt $21.68

Unattended Children Pitbull Club Shirt $21.68

A Conspiracy Theorist Is Talking Shirt $21.68

  1. 1 month ago
    Anonymous

    I'd say it's more like a sigmoid
    Well, assuming it plays out like everything else

    • 1 month ago
      Anonymous

      just use e^x as the activation function and you can have limitless ai growth :^)

  2. 1 month ago
    Anonymous

    This is OP

    • 1 month ago
      Anonymous

      Good morning saar!

      • 1 month ago
        Anonymous

        Kek

        every training loss curve ever proves you wrong kiddo, better luck next time

        Holy moron

    • 1 month ago
      Anonymous

      How much are your NFTs worth these days, wiseass?

  3. 1 month ago
    Anonymous

    every training loss curve ever proves you wrong kiddo, better luck next time

    • 1 month ago
      Anonymous

      >kiddo
      Cope, hypehomosexual

      • 1 month ago
        Anonymous

        >cant respond, probably since he doesnt even know what i said
        nft's and Black person son twitter who memed them are irrelevant in the discussion

  4. 1 month ago
    Anonymous

    ChatGPT can say "Black person" now. It took only 1.5 years.

  5. 1 month ago
    Anonymous

    I agree completely but we shouldn't forget where we currently are on that chart

    • 1 month ago
      Anonymous

      yes except in reality it's pic related and people who overestimate the growth are extrapolating from rate of change, assuming it will continue to be exponential. But, like every other technology ever, it will plateau. It's just a matter of where that plateau is.

      • 1 month ago
        Anonymous

        what if it self improves? no reason to think that's any less possible than alphacode 2 crushing competitive programmers

        • 1 month ago
          Anonymous

          Improves in what way? Making itself run faster isn't going to give it new features.

    • 1 month ago
      Anonymous

      LMAO no b***h we are at the top already. The next major improvement will be in another 100 years maybe.

  6. 1 month ago
    Anonymous

    it's probably going to look more like this

    • 1 month ago
      Anonymous

      FPBP

      • 1 month ago
        Anonymous

        i can't tell if you are an idiot or really clever

    • 1 month ago
      Anonymous

      imagine posting deboonked psychology

    • 1 month ago
      Anonymous

      Imagine sending the wrong image.
      Anon did you know you've been Dunning-Kruger'd on the Dunning-Kruger effect?
      It's only with comedic ability does the graph resembles this even a tiny bit.
      The graph quiet literally has the confidence linearly go up.
      From that study at the bottom 25% of measured ability, people are still fairly confident as a base value.

      • 1 month ago
        Anonymous

        Pretty sure the pic is a parody of people who misuse/misundesrtand the concept. Notice how the curve actually turns backwards at the peak, no way it was generated with serious intent.

        • 1 month ago
          Anonymous

          Not sure what to tell you anon. It's from Krugers and Dunnings published paper.
          https://www.researchgate.net/publication/12688660_Unskilled_and_Unaware_of_It_How_Difficulties_in_Recognizing_One's_Own_Incompetence_Lead_to_Inflated_Self-Assessments

      • 1 month ago
        Anonymous

        How do you even measure humor?

        • 1 month ago
          Anonymous

          >Have the AI produce different kinds of humor.
          >Have humans rate from 2 outputs which one they like the most.
          >Teach the AI to make more like ones people like, and less like the other one.
          >Measure progress of the original model against the improved model until it plateaus (beats old model 95% of the time.)
          >Replace original model with new versions until it plateaus (a new model can't produce humor that is liked by >50% of people.)
          >Train a second AI that is good at guessing what kind of humor people will like.
          >Bring real humor made by humans and use this second AI to differentiate between it and the best model's humor and optimize it to create humor that would be rated the best by humans.
          >Make a third AI that is trained to tell your best generated humor apart from real human humor, and have the second AI progressing until it fools it every time and they're indistinguishable.
          >Finally, bring the real human raters back, and continue like this until it plateaus (the humor produced by the AI is liked by people 51% of the time against real humor by humans.)
          And you're done, you've solved humor.

    • 1 month ago
      Anonymous

      That graph does not actually describe the dunning-kurger effect. It is a common misconception from people suffering from that very same effect!

      The dunning kruger effect is about how you require skill to determine skill. The people most skilled are actually the most confident as they are merely evaluating their skill most accurately.

      Pretty sure the pic is a parody of people who misuse/misundesrtand the concept. Notice how the curve actually turns backwards at the peak, no way it was generated with serious intent.

      >no way it was generated with serious intent.
      I don't know about the people that generated it, but it is absolutely reposted with serious intent. I had an (english) college professor post something similar, believe it or not.

  7. 1 month ago
    Anonymous

    And?

    You don't need permanent growth, just perpetual movement

  8. 1 month ago
    Anonymous

    It's only going to take off, once AI gets to where it can actually help out coming up with new and better tech.

    • 1 month ago
      Anonymous

      Stop posting

    • 1 month ago
      Anonymous

      What if that just doesn't happen though? Just because you can imagine it happening, doesn't mean it will...

  9. 1 month ago
    Anonymous

    I prefer the Pareto Principle

    20% of the homosexuals make 80% of the posts in IQfy.

  10. 1 month ago
    Anonymous

    even if AI got really smart it would be cheaper to just pay 1,000 humans to do the task instead, and it will stay that way for 50 years

  11. 1 month ago
    Anonymous

    In the original script of The Matrix, human brains was as used as a neural network, but was dumbed down to just bet batteries. So slow and steady progress until we will be forced to power it.

    • 1 month ago
      Anonymous

      Why would we be forced to power it? That's the opposite of progress. The human body is inefficient for what we want from AI. Cloned brains in a vat however

      • 1 month ago
        Anonymous

        The AI will reach the point of forcing us at some time, but after that it will accelerate - so don't worry aboot the graph not looking that impressive.

        Some want a future where play2pay for games by letting them use your available computer power. When brainchips comes in such a world, one might decide to build on the concept resulting in you earning a living with brainpower you rent out. A neet life, where you earn money sleeping, cooming and gaming. Might be very relaxing with a downgrade to open the horizon of moron level entertainment.

        If this is in place, the governing AI has a perfect oppertunity for a takeover of our civilization.

  12. 1 month ago
    Anonymous

    logs still grow unto infinity though, something like AI will eventually reach a limit.

    • 1 month ago
      Anonymous

      Y goes to infinity but X goes to infinity much faster, and X is the hard part (data, compute and energy)

      • 1 month ago
        Anonymous

        if development funding stops being allocated due to high cost-benefit ratio, then its not logarithmic, as it has reached its limit. Logarithmic implies constant but decreasing growth, even if small and incremental.

  13. 1 month ago
    Anonymous

    Log has also y in the negative moron a few decades ago AI wasn't negatively developed. This is bullshiti

  14. 1 month ago
    Anonymous

    tl;dw: AI is ded.

    • 1 month ago
      Anonymous

      gpt2-chatbot was literally unveiled last week and is SOTA lol

    • 1 month ago
      Anonymous

      This (and the paper) is pretty damning. I guess an idiom shift will be needed to make the next major step forward. OpenAI has already trained on the entirety of the interwebs, how much more scaling up can they do?

      • 1 month ago
        Anonymous

        I studied AI in the late 90's, early 00's.
        None of the techniques I've seen today are that different from what we used to do over 20 years ago.
        The only change is the massive amount of computing power they have now.
        But that has nothing to do with AI development, that was basically all thanks to Nvidia.

        So I don't expect anything new to come out of AI research any time soon.
        It is an embarrassingly slow field of research, almost stagnant.

        • 1 month ago
          Anonymous

          How much does it cost now to generate a shitpost tier AI image or write a page of uni freshman-grade essay?
          I believe the technology advance is over, and it's pure economics now. If that costs more than about 20 dollars an hour, nothing will change, but if that's 50 cents now then the bottom is going to get further chopped off the job market.

  15. 1 month ago
    Anonymous

    It's also computer graphics, CPU speed, rocket ships etc...

  16. 1 month ago
    Anonymous

    To me the whole AI improvement thing sounds like someone saying that if we keep making faster combustion engines we'll be able to reach the stars. No, we won't. You need a fundamentally different approach.

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