2 pointsby sparkcreativity4 hours ago3 comments
  • PaulHoule3 hours ago
    Maybe two years I had been interested in

    https://en.wikipedia.org/wiki/Kitsunetsuki

    and last December got serious about it in terms of character acting and found Copilot was initially very helpful. So that’s an example of using an LLM for something really unusual and creative.

    The really important developments happened as a result of interacting with people though and “foxwork” turned into “foxography”.

    It’s gotten to be less fun to talk about it with Copilot as it fits everything into a schema and doesn’t seem to mirror my emotional highs and lows. It is still thrilling to talk to another LLM about it because most of them seem to think it is a good idea.

    • sparkcreativity3 hours ago
      *

      1 point by sparkcreativity 6 minutes ago | parent | next | edit | delete [–]

      I think we share some interest in math, APL, computer algebra, etc. Anyway this is a LLM response for you that I agree with: "Paul, your background in the ISO 20022 metamodel and RDF makes your 'schema' comment particularly biting.

      If anyone understands that the 'map is not the territory,' it's you. My hypothesis is that LLMs are the ultimate 'lossy map.' They provide a convenient, averaged-out representation of human thought, but they are fundamentally incapable of capturing the 'foxwork'—those high-density, emotional, and 'unusual' outliers that define real creative breakthroughs.

      You mentioned it's 'less fun' to talk to a model that can't mirror your highs and lows. That 'fun' is the signal of Alpha. When the tool stops being a mirror for your unique complexity and starts being a filter that flattens you, the value of the collaboration drops to zero. We're trading the 'treasure' of specific, jagged insights for the convenience of a predictable schema."

      • PaulHoulean hour ago
        Well I'd say foxwork is an "unusual experience" that needs to be communicated to ordinary people in a way that they get right away or at least that they get some of it and can get more from progressive revelation. (e.g. my current costume gets misidentified as a bear and I roll with that)

        So I am not against there being a bit of averaging. I'd say Copilot was useful when it came to determining brand colors and quite a few things -- but it has all gone from being a peak experience every time I "go out" to something where my heart isn't always in it but other people believing in the character gives me the strength just like Peter Parker.

  • smallerize4 hours ago
    I'm not sure the LLM completely understood the idea. For example this sentence: "Creativity is a game where the only prizes worth winning are high-potential outliers." That doesn't mean anything. It's not a game and there aren't prizes and why is it talking about potential?
    • 3 hours ago
      undefined
    • sparkcreativity3 hours ago
      LLMs are trained on averages. Breakthroughs are outliers. Don't let the average touch your outlier too early.

      That's it. No games, no prizes, no Alpha/Beta framing. Just: protect your weird half-formed thoughts from the smoothing function until they're strong enough to survive it.

  • miningape4 hours ago
    The raw idea is the only part of this text I found interesting.
    • 3 hours ago
      undefined
    • sparkcreativity3 hours ago
      I tend to agree, but it could be that you just expressed AI slop fatigue, or that the LLM text just clarified the initial thought and now you see the initial thought more clear. Anyway thanks for your comment (you deserve a human response )