(Mis?)applying Evolution 101 here, I'm wondering if misremembering (mutation) isn't as important as remembering (reproduction). Our fixation with remembering (aka accurately reproducing) comes from a pre-print age where, naturally, information retained with as less corruption was prized; Infact a lot of liturgy is insistent on exact reproduction going as far treating any deviation as unacceptable/unpardonable - here I'm thinking of the Vedic chants that, apparently, trace back to millennia of exact replication. However, what if the obsession with exact storage (which in itself raises a translation/ transcription problem from phenomena to encoding mechanism) is detrimental to evolving intelligence?
At a civilisation level, there is the ever-increasing baggage of memory that, unless constantly ETTO-simplified, slowly cleaves apart epistemology and ontology until the contradictions cause an implosion. At a more personal level, forgetting seems like an incredibly powerful mechanism, to be able to continue functioning. If indeed 'Action trumps Truth', then abstractions with inevitable corruptions that nevertheless respond quickly so that we can test are more useful than exact-but-slow recall. The Telugu writer Kasibhatla Venugopal, echoing your friend above, once said that we don't have a remembering problem but a recollection problem. His conjecture was that everything that enters your brain remains there, what you have trouble with is retrieving it when you ask for it. I want to extend that and say, 'in the form you want it'. What you write about Spaced Repetition sounds like Prompt Engineering, with shades of Venkat's 'To know is to stage' argument, where the functioning of the memory system is still a kind of black box but we've gotten better at creating the right environment and parameters to get what we want.
Which ties into another thread that I've been thinking about. I've always noted down thoughts but only in the last couple of years have I tried making notes while reading (book margins, Evernote, Roam etc.) but none have worked. Infact I feel like whatever I read slips through quicker than it would otherwise - ofcourse, there's always the risk of overestimating how much I remember when I don't have anything to verify against - but more importantly, I feel like my learning stays at a shallower level. Almost as if it is imperative to look past trees to see the woods. And tying back to the LLM Blackbox metaphor, something interesting seems to happen when we go beyond controlling procedures to giving feedback on performance.
(Mis?)applying Evolution 101 here, I'm wondering if misremembering (mutation) isn't as important as remembering (reproduction). Our fixation with remembering (aka accurately reproducing) comes from a pre-print age where, naturally, information retained with as less corruption was prized; Infact a lot of liturgy is insistent on exact reproduction going as far treating any deviation as unacceptable/unpardonable - here I'm thinking of the Vedic chants that, apparently, trace back to millennia of exact replication. However, what if the obsession with exact storage (which in itself raises a translation/ transcription problem from phenomena to encoding mechanism) is detrimental to evolving intelligence?
At a civilisation level, there is the ever-increasing baggage of memory that, unless constantly ETTO-simplified, slowly cleaves apart epistemology and ontology until the contradictions cause an implosion. At a more personal level, forgetting seems like an incredibly powerful mechanism, to be able to continue functioning. If indeed 'Action trumps Truth', then abstractions with inevitable corruptions that nevertheless respond quickly so that we can test are more useful than exact-but-slow recall. The Telugu writer Kasibhatla Venugopal, echoing your friend above, once said that we don't have a remembering problem but a recollection problem. His conjecture was that everything that enters your brain remains there, what you have trouble with is retrieving it when you ask for it. I want to extend that and say, 'in the form you want it'. What you write about Spaced Repetition sounds like Prompt Engineering, with shades of Venkat's 'To know is to stage' argument, where the functioning of the memory system is still a kind of black box but we've gotten better at creating the right environment and parameters to get what we want.
Which ties into another thread that I've been thinking about. I've always noted down thoughts but only in the last couple of years have I tried making notes while reading (book margins, Evernote, Roam etc.) but none have worked. Infact I feel like whatever I read slips through quicker than it would otherwise - ofcourse, there's always the risk of overestimating how much I remember when I don't have anything to verify against - but more importantly, I feel like my learning stays at a shallower level. Almost as if it is imperative to look past trees to see the woods. And tying back to the LLM Blackbox metaphor, something interesting seems to happen when we go beyond controlling procedures to giving feedback on performance.
A part of me chafes at this 'irrational' way of doing it, convinced that I'm rationalising my inabilities and pop Zen parables like these don't always help - https://www.facebook.com/groups/342329417071482/posts/1457168578920888/.