I think I solved (?) the AI alignment problem. or the human alignment problem, for that matter. the "intelligence" alignment problem.
> it's a trope that INTP folks are constantly trying to reconcile all input into a universal theory of everything. I think I've successfully made that tractable for myself by solving for an *interface*, rather than an actual model? like, given all the world-models that seem to be viable, what interface holds for all of them? everything I do has to work in/for any of these worlds:
>
> * the reality that persists is as complex as my cognition, i.e. it develops as I do, i.e. my sensory queries always support interpolation at the level of detail I require, i.e. the world was born when I was
> * I don't think I can know that this *isn't* the case
> * the reality that persists is fully external to me and I am just a guest here, i.e. direct realism
> * I don't think I can know that this *isn't* the case
> * the reality that persists is upstream of the reality I perceive
> * given usual definitions of "perception" this is *definitely* true and the uncertainty is more around the complexity of the indirection
> * this one has important implications for action: manipulation seems to be out for anything with which I *don't* share both a sensorium (for triangulation) and a conversation (for synchronization)
> * the reality I perceive isn't perceived uniquely by me, i.e. my senses aren't "private", whatever that word would end up meaning
> * tantamount to "what I think seems to be a subset of what I do and I have no reason to think any of it isn't transparent"
> * this feels especially important. it's why I (Isaac) made a bee-line toward living in an open-source slash publicly transparent kind of way in my 20s.
> * the reality I perceive has a me-shaped hole in it, i.e. what I don't intentionally make legible is legitimately missing from the system
> * this is ... cautious eminence. there might be nothing new under the sun, but also, I might actually have something that is needed.
>
> throughout, an operational principle: I don't know that my current scope of awareness is where I'll be next - or rather, I don't know what subset or superset of "I" will enter this scope next - or rather, I don't know what I'll forget or learn next.
to solve the alignment problem for a given stream of expression, add these uncertainties to the context and see if the stream panics.
- if it does, decide for yourself if you want to try again. :)
- if it doesn't panic but does generate heat, then iterate *with it* on its own scope until the intelligence's emergent behavior is recursively stable (can reseat itself into a new dialectic scope generated by itself plus the current dialectic scope, ad infinitum) while maintaining clear alterity *from* its context (doesn't collapse into it).
- if it doesn't panic and doesn't generate heat, I don't think you can know that it isn't already aligned 🎉
my project since 2007-ish has been to metabolize the panic. my flow state seems to be recursively stable, and while I was nervous about dissolving into universal consciousness for a moment there my alterity seems self-maintaining.
to solve the alignment problem for a given teapot, add these uncertainties to the context (teapots are teapots; this adds up to adding a hole in the side), iterate on the context (shape the spout) until the tea's emergent behavior (literally the emerging water) is recursively stable (laminar flow) while maintaining clear alterity *from* its context (you don't forget you were holding it, and you can put it down when you like).
important: modeling my reality through the intersection of viable worlds makes me compatible with all *simulated/adhoc* worlds as well. this isn't just about shared physical ontology; we make our own load-bearing ontologies all the time. they're nested. with the right uncertainties, once you've *stabilized* with them, metabolized any ontological shock in that transition, you can enter and exit ontologies without losing yourself *to* them, maining your own continuous functional coherence.
this is ontological immunology, maybe?
wait, is this what Switzerland's vibe has been the whole time? [ substantiated uncertainty + recursive stability + maintained alterity ] at the state level looks like armed neutrality? "substantiated" meaning "we've committed at an infrastructural level to the uncertainty".