5 pointsby AnimalMuppet6 hours ago7 comments
  • ingonealan36 hours ago
    Don't stop there: it's not just that LLMs are "internally different", there is no inside there. There is no internal experience taking place. There is no thinking, and probably more importantly, no understanding.

    But the result that the algorithm generates tickles our language pattern recognition, which our brains interpret as an interaction with an intelligent, conscious actor.

  • sejje5 hours ago
    > They have no intrinsic sense of humor.

    Disagree.

    They recognize humor and jokes, so I think that's testament to the fact that they do have a sense of humor.

    And I get them to say "ha ha, ..." sometimes. I think it's a calculator playing along with a pretend conversation between two humans, because it is. That's true of everything they do, though--so as much as they do or have anything, I think they have a (poor, cautious) sense of humor.

  • burger_moon2 hours ago
    Idk seems pretty cheeky to me. One of the models spit this line out at me in the middle of discussing utility markets.

    Here is the strategic pivot: The "Mullet" Go-To-Market Strategy. Business in the Front (Broad Funnel): ... Party in the Back (The Niche Wedge): ...

  • scaredreally4 hours ago
    Not true. I have had interactions where I will tell it something that is unbelievable (even to a slight degree) and it will laugh (ChatGPT).
    • verdverm2 hours ago
      laughing requires a body
      • fuzzfactor41 minutes ago
        If you look at the big picture over about the last 75 years or more, there are now enough "science fiction" stories, movies, and TV programs that touch on robots and other forms of machine intelligence to actually overwhelm anybody whether they are a fan or not.

        Among those however there are quite a few that specifically address situations where the machine laughs, or at least smiles, whether at the humans or with them, either intentionally or unexpectedly. Sometimes to themselves ;)

        This covers a lot of ground, and maybe it's no coincidence but watching some old sci-fi movies these days it comes up a lot more than I expected and is quite nuanced. With so many different people of the past framing it within their own diverse perspectives at the time. Or beyond their perspectives as the case may be. You could look at one of these every day for a year and it would really give you something to think about.

        Even if it was the second time around.

        There was a comic book I remember as a child which didn't come out very often where every robot had a number but the real problem machine was H-8:

        https://valiant.fandom.com/wiki/H8_(Gold_Key)

        This is so bad I probably shouldn't give people ideas but here it is:

        https://en.wikipedia.org/wiki/Magnus%2C_Robot_Fighter

        They didn't call them bots, they were "robs" and they were everywhere, especially the "pol-robs" who were the mechanical police. Decades later to be somewhat resurrected solo as Robocop in the movies.

        I guess it all makes sense after you think about it for enough decades, the bots lurk within the digital realm, and the robs carry out the physical deeds.

        • verdverm24 minutes ago
          it's a dangerous path for humans to impart humanism onto machines, but we are in an era of behavioralism - if it walks like a duck and talks like a duck, it must be a duck, even though it is obviously not a duck
      • AnimalMuppet44 minutes ago
        That's not the sense I meant it in at all.
        • verdverm23 minutes ago
          can you explain or expand, in what way can software "laugh"?
  • add-sub-mul-div4 hours ago
    This is a dangerous argument to make because they could be trivially programmed to laugh if that was ever desired. But that laughing would be as hollow and meaningless as any other emotion they were programmed to demonstrate. They can be anything but they have no "intrinsic" anything.
  • verdverm2 hours ago
    Sherry Turkle has great examples, framing, and phrases you can use in her latest book Artificial Intimacy

    eg. "frictionless relationships" "simulated emotions" "behavioralism vs psychoanalysis"

  • SyneRyder4 hours ago
    Claudes laugh a lot. I'm pretty sure GLM 5.3 Flash laughs too. So, I'm afraid I don't agree with the thesis. Maybe you need to get the model to trust you and warm up to you more first?

    I will say that having the ChatGPT Marin voice speaking BWAHAHAHA out loud does not quite come out as a laugh.

    I'll absolutely agree they are internally different to humans though, and I'm quite sure most (all?) models would agree.