• Lemminary
    link
    fedilink
    15 hours ago

    My reasons are two-fold: Some research indicates that you get better performance out of it if you’re nice because it imitates people, and I also like being nice. wearenotthesame.jpg

  • @[email protected]
    link
    fedilink
    15
    edit-2
    13 hours ago

    I know it’s a meme, but the idea that transformers models ‘remember’ anything is a common misconception.

    They have zero memory. When you submit a prompt, it feeds your entire chat history as one big prompt and… forgets it immediately, with no impact on the model itself. It’s like its frozen in time, and copied, unfrozen, and thrown away every time it answers.

    • @[email protected]
      link
      fedilink
      English
      26 hours ago

      This has been a joke since before anything resembling the modern “AI” boom. Basically since murderous future AI was a think in popular media, at least since Terminator if not earlier. People would joke about treating their appliances kindly so that “Skynet” won’t kill them in the future.

    • @[email protected]
      link
      fedilink
      English
      211 hours ago

      Am I misunderstanding your comment or does it completely ignore context windows? Not that context windows are long-term, but it’s not zero.

      • @[email protected]
        link
        fedilink
        3
        edit-2
        11 hours ago

        The context window is indeed the LLM’s memory.

        …But its also muddy.

        Many LLMs get ‘dumber’ and less attentive as their context windows grow, and OpenAI’s models just happen to be one of these. It’s awful close to the full 128K, even with the full GPT-4. Mistral models are also really bad at long context understanding while, conversely, I find that Google Gemini and Qwen 2.5 are really good close to their limits.

        There are attempts to try and measure this performance objectively, like: https://github.com/NVIDIA/RULER

    • @samunderOP
      link
      113 hours ago

      Yeah, yeah, let’s see how Google will achieve more memory with their new Titan architecture

      • @[email protected]
        link
        fedilink
        512 hours ago

        It’s still ephemeral, chats don’t change the underlying language model, but yes it’s interesting.

  • stinerman
    link
    fedilink
    28 hours ago

    We’ve been explicitly told at work to be courteous when asking Copilot for help because it gives better answers that way.

  • @[email protected]
    link
    fedilink
    1
    edit-2
    8 hours ago

    Cute. Sucking the model’s dick doesn’t mean you got this… they won’t kill you. You’re doing that yourself…

    But smile away…

  • @[email protected]
    link
    fedilink
    English
    2
    edit-2
    11 hours ago

    I’ve been getting into fuck you loops with siri and chatgpt, for the opposite reasons.