The Word Made Lifeless. Are we becoming stochastic parrots?

(hedgehogreview.com)

57 points | by pseudolus 2 hours ago

14 comments

  • Avicebron
    1 hour ago
    This is a very interesting way of saying that by offloading the procession of thoughts we take to come to a conclusion individually we become more dull and homogenized as a whole when using ai "shortcuts".
    • rustystump
      1 hour ago
      I think this is true with many tools but pronounced here with how ubiquitous language is. For most of human history most people could not read and write. Given llm interaction is almost exclusively written we may take a small step back where a mass of people are lazy using shortcuts and others are not.
      • Avicebron
        54 minutes ago
        It's going to be depressing if "reading is hard" is the Great Filter
        • mlinhares
          48 minutes ago
          It already is, most of my friends don't read anything beyond social networks anymore. This is already happening and was in progress long before LLMs, they are mostly fast tracking the process.
  • tolerance
    21 minutes ago
    I get the point that the author is trying to make here and to an extent I think it’s prudent. If only it wasn’t bogged down by allusions that are withering in cultural relevancy.

    AI is not a faithful representation of human intelligence—or the human essence at all, for that matter—and prolonged dependence on its technologies will subdue human expression and the means in which we come to know about ourselves and life around us.

    He is exercising his point through his prose. A cursory glance at his Wikipedia bio and this passage gives insight into the objective of this article:

    > As Socrates sees things, the proper use of logos is to work toward, and to be transformed by, an increasingly clear grasp of the good. He regards this as an erotic undertaking. The more clearly we see the good, the more we long to bring ourselves into closer proximity to it. And the most promising path to the apprehension and internalization of the good is prolonged union and thoughtful conversation with a worthy lover.

    The problem is that he’s attempting to display the art of literary lovemaking in the age of hook-up culture.

    Hint: It may serve this crop of readers to start immediately from the section titled “Another Creation Story”.

    Aside: I’d like to see a study that scores contemporary literacy rates using articles from the Hedgehog Review instead of Jane Austen.

  • ggm
    1 hour ago
    Could be said more succinctly: use it or lose it
  • conartist6
    6 minutes ago
    But everyone is so thrilled about not having to think again any more ever!

    It would be so sad to disappoint

  • shnpln
    1 hour ago
    What happens when/if the machines do everything? We will have no more problems to solve.
    • AyyEye
      54 minutes ago
      I don't believe in digital gods.

      But also the machine stops[1] and idiocracy both address this.

      [1] https://www.gutenberg.org/cache/epub/72890/pg72890-images.ht...

    • shoobiedoo
      59 minutes ago
      What is life without struggle, though? Reading through a long tech manual, overcoming failure dozens of times, telling myself "just one more try, you're so close" and the ecstatic feeling of success that follows as I see my project come to fruition. Without it, how do we reaffirm ourselves...?
      • anhyz
        28 minutes ago
        Digital gods can't take away your ability to struggle purely for the enjoyment of it.

        Think of all the things we consider fun or worthwhile that are purely unnecessary struggles. Ever play sports? Build a side project just for the fun of it? Video games? Music? Art?

      • marcus_holmes
        46 minutes ago
        If you've ever used an LLM to try and build something, you know there's still plenty of struggle involved ;)
      • gdulli
        52 minutes ago
        I think it's even more than affirmation and pride, I think the process is the most critical part of learning. And now we're indulging people who want shortcuts that are going to significantly dull them.
    • pshc
      39 minutes ago
      We can spend more time on art, music, camaraderie, love, play, and the pursuit of grace.
      • sublinear
        27 minutes ago
        None of which is possible without the lived experiences of problem solving.
    • esafak
      1 hour ago
      Paying your bills and finding something worthwhile to do with your life may be problems enough.
    • throw_m239339
      40 minutes ago
      We will always have a problem to solve. Isn't it the point of Asimov's story, "The Last Question"?
    • kjkjadksj
      32 minutes ago
      Bold of you to assume the machine gods would allocate resources for our survival.
    • antisthenes
      54 minutes ago
      Disease? Mortality? Finding a good partner?

      How do you imagine a robot solving those?

      • shnpln
        46 minutes ago
        I cant. I think of a species that evolved at the time soonest after the big bang and is still alive today. Maybe they have? idk Billions of years is a long time.
  • frogger42
    1 hour ago
    The text on this site is too small. When you spend so much time on the design of your blog and don’t optimize for viewing on smaller mobile devices, it’s like painting the Mona Lisa and leaving it in your garage.
    • nightpool
      57 minutes ago
      It's about 49 characters, which is maybe a little small but much closer to the size I normally read my ebooks at then most website these days. Most recommendations I see online for digital prose line length recommend around 55 characters at the lowest, so I think it's doing okay.
      • bburris
        39 minutes ago
        I can’t read the Hedgehog Review on my phone either. It’s the combination of the small letters and font. HN has a similar width of characters, but the font is legible at that size.
    • sublinear
      23 minutes ago
      It's funny that I had the exact opposite reaction. I was happy to get more information with less eye movement. Large text is everywhere these days and causes me some eye strain after a while.
    • g_host
      1 hour ago
      Reader mode for the win
    • inquirerGeneral
      1 hour ago
      [dead]
  • wzdd
    52 minutes ago
    Quite an astonishing amount of quotes -- I counted Heidegger, Nietzsche, Murdoch, the Bible, and Plato as key players -- to contribute very little of substance. It feels like the same kind of hand-wringing that you might have read about the invention of the typewriter, the telegraph, or recorded music. Humans are complex creatures! We can be "in the process of becoming" about a great many things along a great many axes. Using ChatGPT to draft a letter of complaint to your local government isn't going to change that.
    • conartist6
      5 minutes ago
      The basilisk consumes the minds of those who worship it
    • add-sub-mul-div
      50 minutes ago
      Invoking the cliched comparison of AI to past inventions in a discussion about parroting is brilliant satire.
      • logicprog
        31 minutes ago
        So is writing an ultimately empty essay mostly composed of other people's quotes, thoughts, and concerns
      • cgio
        35 minutes ago
        While I agree, making the assumption that our GP was not indeed using an LLM, that has a lot to say about the essence of the article’s argument. We were maybe already stochastic parrots and are just building tools to be more efficient at it.
  • Liftyee
    1 hour ago
    This highlights an interesting point that I hadn't examined before... the thoughts behind the words, and the experiences linked to those thoughts / the moment for which the words are needed.

    Although from a chemical perspective and learning perspective humans may seem like stochastic machines, I think that the capability for inner thought and emotion does differentiate us from genAI. After all, we would not ascribe sentience to such a "stochastic parrot", but in a similar way we see the text output that AI generates and wrongly assume (in a previously reasonable assumption) that there must have been "thoughts" behind said text.

    • kordlessagain
      59 minutes ago
      No different than a book. Maybe the places where the experiences were had were real and maybe they weren’t. Maybe we recall the emotion or it is projected from the story (or resonates in frequency with our own experiences).
    • timschmidt
      1 hour ago
      I liken LLMs to https://en.wikipedia.org/wiki/Broca%27s_area

      It's a functional subunit of the brain responsible for language interpretation and production.

  • brookst
    1 hour ago
    Hmm. What would Ian Betteridge say?
    • thfuran
      1 hour ago
      "What kind of parrot?"
      • pimlottc
        1 hour ago
        Norwegian blue. Beautiful plumage.
        • thfuran
          26 minutes ago
          Very low energy though.
  • 77pt77
    46 minutes ago
    [dead]
  • kordlessagain
    1 hour ago
    [dead]
  • robmccoll
    1 hour ago
    You are a state machine. You have finite internal state that roughly adheres to a particular structure, you take in input, and as a function of internal state and input, you produce output and a new state. Sufficiently large models are a rough approximation. We are perhaps different stochastic parrots than the models we create, but likely stochastic parrots none the less.
    • gyomu
      1 hour ago
      You are a hydraulic system. You are composed of interlinked pipes within which the pressure rises and lowers in order to produce all your thoughts and actions.

      You are a chemical soup. Your body is a closed system of proteins and amino acids reacting with each other, driving behavior in order to sustain the reactions.

      You are an electric grid. A system of interconnected wires where electric impulses respond to one another in a synchronized manner, from which your life force is derived.

    • OtherShrezzing
      35 minutes ago
      A state machine is a very specific thing in Comp Sci, and I’m not clear you have a strong grasp on it.

      You’re not a state machine. A state machine does one serial task, which is why the input+state can create a consistent and deterministic output+state. There are no secondary input streams or exogenous factors to consider for calculating a state machine transition.

      Humans create output from many streams of input, arriving at across many different time horizons. Because of this, you cannot create a deterministic model of a human’s state transition for a given input - a requirement of state machines.

      This isn’t philosophical or semantics. Mathematically, you’re not a state machine.

    • dekhn
      43 minutes ago
      Probably better to say that we contain multitudes of probabilistic graphs that resemble state machines, but those graphs do not make up our entire system. Further, those graphs interact constantly with stochastic systems (the environment, other graphs, etc) through couplings of varying degree.
    • AdieuToLogic
      54 minutes ago
      > You are a state machine.

      Humans are analog, state machines are not. And the analogue I will use here is that a model of anything is not the thing itself by definition.

      > We are perhaps different stochastic parrots than the models we create, but likely stochastic parrots none the less.

      To parrot is to "to repeat by rote"[0]. Algorithms, such as LLM's, do so as that is all they can do.

      I choose to not limit myself to being a parrot. Which is why I am not one.

      As Descartes proffers:

        The only thing we have power over in the universe is our 
        own thoughts.[1]
      
      0 - https://www.merriam-webster.com/dictionary/parrot

      1 - https://www.azquotes.com/quote/1521522

    • zabzonk
      1 hour ago
      > You have finite internal state

      As Terry Pratchett might have said - what about quantum?

      • Filligree
        1 hour ago
        What about it? That only exponentiates the state space, it doesn’t make it non-finite.
    • shortrounddev2
      1 hour ago
      I believe that the different topologies of the kind of "idea graph" in the human mind is becoming less diverse. That, as media consolidates and becomes more accessible in all parts of the globe, the diversity of modes of thought decreases as people become, more or less, Americanized
      • rustystump
        1 hour ago
        Perhaps Americanized isnt the right term. Fundamentally, there is something at play where the more “known” the less magical the brain is. That is, it doesnt have to think outside the box because the box is seemingly fully explored.

        Why do hard thing when everyone says hard thing is too hard. What if there was no everyone? Is thing that hard?

        I dont think original thinking is going to go away but i do think it will be owned by those who control the all thing which absorbs it from the mass of information.

  • marcus_holmes
    48 minutes ago
    I find it ironic, if that's the right word, that this essay quotes someone else in almost every paragraph and yet insists that humans are not stochastic parrots, and the essence of human communication is original thought.