Was musing about nyam-thogs as auto-complete token streams:
A nyam can refer to any experiential state (including both the nyam of formless bliss and the nyam of ordinary life.)
Thog can mean "arising"
So, nyam-thogs could be translated as "thought stream" or "state-arisings".
Translate "karma" as "habitual patterns of perception":
a: action
z: state (which conditions perception and thus conditions action)
k: karma (habitual patterns of perception [interpretation of sensation])
s: sensation (prior to manipulation by emphasis, de-emphasis, or putting aside)
p(a|z(k(s))): probability of action conditional upon state, which is a function of karma with respect to priorly unmanipulated sensation.
The available leverage for not getting trapped in conditioning is to relax k (karma), increasing its degree of injectivity, or what % of its inputs map to unique outputs.
- the function that maps history of observations into the state z [the interpretation of history into a concise state]
- and the parameters of the foundation model itself. If we assume that we are continuously training as we live, than the experience is continuously getting imprinted on the "auto-complete" function.
I think s in your notation is related to observations o in mine. Sensations or observations are inputs of the system. I guess, you could think of sensations as pre-sembolic level, while observations are their interpretations. Is that what k(s) does? Maps sensations to observations?
Raw training data = Sensation (which may already be somewhat conditioned, for example, by biology - not equal to "reality in itself", for sure...it's just the least processed input, however processed that is.)
Conceptuality/Symbolic = Observations/perceptions ("Conceptuality", here is used in the extremely broad Vajrayana usage, which means "pattern recognition", at whatever level. Traditionally, there are something like 84 levels of conceptuality, and an insect would be considered as more conceptual than a person, because its fixed pattern recognition is stronger, and a dog somewhere in between).
The pair of Conceptuality of Sensation, call them c and s:
Perhaps better, then, to expand:
z(m(o))
to
z(m(c(s))) (inner state based on memory based on conceptuality of sensation)
And "karma" is the composed function k = z ∘ m ∘ c
Was musing about nyam-thogs as auto-complete token streams:
A nyam can refer to any experiential state (including both the nyam of formless bliss and the nyam of ordinary life.)
Thog can mean "arising"
So, nyam-thogs could be translated as "thought stream" or "state-arisings".
Translate "karma" as "habitual patterns of perception":
a: action
z: state (which conditions perception and thus conditions action)
k: karma (habitual patterns of perception [interpretation of sensation])
s: sensation (prior to manipulation by emphasis, de-emphasis, or putting aside)
p(a|z(k(s))): probability of action conditional upon state, which is a function of karma with respect to priorly unmanipulated sensation.
The available leverage for not getting trapped in conditioning is to relax k (karma), increasing its degree of injectivity, or what % of its inputs map to unique outputs.
Another places that look "karmic":
- the function that maps history of observations into the state z [the interpretation of history into a concise state]
- and the parameters of the foundation model itself. If we assume that we are continuously training as we live, than the experience is continuously getting imprinted on the "auto-complete" function.
I think s in your notation is related to observations o in mine. Sensations or observations are inputs of the system. I guess, you could think of sensations as pre-sembolic level, while observations are their interpretations. Is that what k(s) does? Maps sensations to observations?
Yes.
I guess you could break the stream into:
Raw training data = Sensation (which may already be somewhat conditioned, for example, by biology - not equal to "reality in itself", for sure...it's just the least processed input, however processed that is.)
Conceptuality/Symbolic = Observations/perceptions ("Conceptuality", here is used in the extremely broad Vajrayana usage, which means "pattern recognition", at whatever level. Traditionally, there are something like 84 levels of conceptuality, and an insect would be considered as more conceptual than a person, because its fixed pattern recognition is stronger, and a dog somewhere in between).
The pair of Conceptuality of Sensation, call them c and s:
Perhaps better, then, to expand:
z(m(o))
to
z(m(c(s))) (inner state based on memory based on conceptuality of sensation)
And "karma" is the composed function k = z ∘ m ∘ c
(I appreciate math as colorful metaphor)
Delightful read, thank you. ✨🙏
Fantastic essay! I am going to re-read at least two more times.
I've also sent you a message via Substack with some more questions.
Thank you so much!