Login
You're viewing the front-end.social public feed.
  • poleguy looking for lost toolspoleguy@mastodon.social
    Oct 10, 2026, 4:39 AM

    @abhayakara @emilymbender I think if you read the google paper and understand it to some small degree, you will see that the logic is sound.

    What's a stochastic parrot? LLMs follow simple logic... they predict the next token and are simple models that produce outputs based on their training data. As such they have no capacity beyond that. They cannot possibly be conscious.

    They are very large, so they are very convincing models. But only because the training data is from human consciousness.

    💬 3🔄 0⭐ 5

Replies

  • Ted Lemonabhayakara@mastodon.nl
    Oct 10, 2026, 11:10 AM

    @poleguy

    In order to make this assertion, you first have to explain what you mean by "conscious." That's part of what I mean when I say that the assertion that something is conscious is not falsifiable:

    1. We haven't agreed on what "conscious" means.
    2. For meaningful values of "conscious" that I can think of, proving that a thing is conscious from some perspective outside of that thing is impossible: consciousness is necessarily subjective.

    💬 0🔄 0⭐ 0
  • Ted Lemonabhayakara@mastodon.nl
    Oct 10, 2026, 11:13 AM

    @poleguy

    Ironically, you can say the same thing about humans: we start out as babies, and interact with humans over time, and in the process learn how to be human.

    This is not to say that babies are stochastic parrots, or that LLMs are babies. The point is that this isn't a distinction that is useful in reasoning about whether or not LLMs are conscious.

    Again, I am not asserting (nor do I particularly believe) that they are. But I don't know of an intellectually honest way to _prove_ that they are not.

    💬 0🔄 0⭐ 0
  • Viraptorviraptor@cyberplace.social
    Oct 10, 2026, 11:37 AM

    @poleguy
    This falls into the good old: Why do you think humans are not next token predictors?
    Not that it's a great direction, but every time someone raises "LLMs can't ... because they're just predicting the next token", the whole premise hangs on - people are in some way different. Why?

    💬 1🔄 0⭐ 0
  • poleguy looking for lost toolspoleguy@mastodon.social
    Oct 10, 2026, 5:55 PM

    @viraptor that logic seems a tad shaky... I am not making a claim about humans.

    I'm making a claim about LLMs. The structure of how they work is documented in the 2017 Google paper.

    So my argument depends on exactly zero conclusions we could make about humans.

    💬 1🔄 0⭐ 0
  • Viraptorviraptor@cyberplace.social
    Oct 10, 2026, 6:18 PM

    @poleguy
    You said: "As such they have no capacity beyond that. They cannot possibly be conscious." - but there's an assumption that humans are conscious, (hopefully we agree on this) so that implies they have to be more than just next token predictors. That's where the "How/Why?" comes from.

    💬 1🔄 0⭐ 0
  • poleguy looking for lost toolspoleguy@mastodon.social
    Oct 10, 2026, 6:32 PM

    @viraptor Remember, I'm not making a claim about humans.

    It does imply this. And this is true. But that is a discussion about humans. I'm talking about LLM's. It is much simpler to debunk claims of consciousness in LLM's than to explain consciousness in humans.

    💬 0🔄 0⭐ 0