|
|
|
|
|
|
by PaulHoule
118 days 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. |
|
|
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."