Comment on Enshittification of ChatGPT

• Umbrias@beehaw.org ⁨4⁩ ⁨months⁩ ago

  Where is the world model you maintain? Can you point to it? You can’t - because the human mind is very much a black box just the same way as LLM’s are.

  something being a black box is not even slightly notable a feature of relation, it’s a statement about model detail; the only reason you’d make this comparison is if you want the human brain to seem equivalent to llm.

  for example, you didnt make the claim: “The inner workings of Europa are very much a black box, just the same way as LLM’s are”

  source

  • Opinionhaver@feddit.uk ⁨4⁩ ⁨months⁩ ago

    “The human mind is very much a black box just the same way as LLMs are” is a factually correct statement. You can’t look into a human brain for an exact explanation of why an individual did something any more than you can look into the inner workings of an LLM to explain why it said A rather than B. Claiming that my motive is to equate LLMs and human brains is not something I said - it’s something you imagined.

    source

    • Umbrias@beehaw.org ⁨4⁩ ⁨months⁩ ago

      It’s not really factually correct if you want to get pedantic, both brains and llms are called black boxes for different reasons, but this is ultimately irrelevant. Your motive may be here or there, the rhetorical effect is the same. You are arguing very specifically that we cant know llm’s dont hae similar features (world model) to human brains because “both are black boxes”, which is wrong for a few reasons, but also plainly an equivalence. It’s rude to pretend everyone in the conversation is as illiterate as wed need to be to not understand this point.

      source

      • Opinionhaver@feddit.uk ⁨4⁩ ⁨months⁩ ago

        A statement can be simplified down to the point that it borderlines on misinformation while still being factually correct. Another examples would be saying “photography is just pointing a camera and pressing a button” or “internet is just a bunch of computers talking to each other.” It would be completely reasonable for someone to take issue with these statements.

        You are arguing very specifically that we cant know llm’s dont hae similar features (world model) to human brains because “both are black boxes”

        At no point have I made such claim.

        source

        • -> View More Comments