I previously have had no experience with IFS, Focusing or Felt sense, but it seems to absolutely click with my worldview and thoughts I’ve been having about the mind and the self for a long time. Still reading through several LW articles about it, but it gave me an idea. I have a creative project that I have a general ‘vibe’ for what I want it to be, but have no idea what I actually want out of it. So, aiming as much as possible to simply point as much at ‘the feeling’ or ‘felt sense’ it had in my mind, I wrote/dictated a few paragraphs of text about the work, much of which was literally just free association of words and vibes that got me closer to what I was feeling.
Then, I pasted it, verbatim, into GPT4. And I got one of the best prompt results I’ve ever gotten, it effortlessly translated my ramblings and vibes into a title, genre, and solid rundown of near-exactly what I had in mind, far better than I’ve had in the past when I’ve tried to just ask directly for creative advice. It didn’t ask me for specification, explain what I wanted. It just understood.
This is really interesting to me, especially given what you’ve said here about emotional flavors and what I know about how tokens operate in vector space by way of their relative meaning. If the human brain is a vector space of concepts, with certain neurons related to others based on their literal distance both semantically and physically (which I’m pretty sure it does, given what I’ve heard about different parts of the brain ‘lighting up’ on an mri when experiencing different things) then what is the difference, effectively, between our brains and this vector space of tokens that LLMs operate on?
I previously have had no experience with IFS, Focusing or Felt sense, but it seems to absolutely click with my worldview and thoughts I’ve been having about the mind and the self for a long time. Still reading through several LW articles about it, but it gave me an idea. I have a creative project that I have a general ‘vibe’ for what I want it to be, but have no idea what I actually want out of it. So, aiming as much as possible to simply point as much at ‘the feeling’ or ‘felt sense’ it had in my mind, I wrote/dictated a few paragraphs of text about the work, much of which was literally just free association of words and vibes that got me closer to what I was feeling.
Then, I pasted it, verbatim, into GPT4. And I got one of the best prompt results I’ve ever gotten, it effortlessly translated my ramblings and vibes into a title, genre, and solid rundown of near-exactly what I had in mind, far better than I’ve had in the past when I’ve tried to just ask directly for creative advice. It didn’t ask me for specification, explain what I wanted. It just understood.
This is really interesting to me, especially given what you’ve said here about emotional flavors and what I know about how tokens operate in vector space by way of their relative meaning. If the human brain is a vector space of concepts, with certain neurons related to others based on their literal distance both semantically and physically (which I’m pretty sure it does, given what I’ve heard about different parts of the brain ‘lighting up’ on an mri when experiencing different things) then what is the difference, effectively, between our brains and this vector space of tokens that LLMs operate on?
Would you be willing to share the prompt and result?