One could imagine a language “Lyapunese” where every adjective (AND noun probably?) had to be marked in relation to a best guess as to the lyapunov time on the evolution of the underlying substance and relevant level of description in the semantics of the word, such that the veridicality conditions for the adjective or noun might stop applying to the substance with ~50% probability after that rough amount of time.
Call this the “temporal mutability marker”.
“Essentialism” is already in the English language and is vaguely similar?
In English essential traits are in the noun and non-essential trait are in adjectives. In Spanish non-essential adjectives are attributed using the “estar” family of verbs and essential adjectives are attributed using the “ser” family of verbs. (Hard to find a good cite here, but maybe this?)
(Essentialism is DIFFERENT from “predictable stability”! In general, when one asserts something to be “essential” via your grammar’s way of asserting that, it automatically implies that you think no available actions can really change the essential cause of the apparent features that arise from that essence. So if you try to retain that into Lyapunese you might need to mark something like the way “the temporal mutability marker appropriate to the very planning routines of an agent” interact with “the temporal mutability marker on the traits or things the agent might reasonably plan to act upon that they could in fact affect”.)
However, also, in Lyapunese, the fundamental evanescence of all physical objects except probably protons (and almost certainly electrons and photons and one of the neutrinos (but not all the neutrinos)) is centered.
No human mental trait could get a marker longer than the life of the person (unless cryonics or heaven is real) and so on. The mental traits of AI would have to be indexed to either the stability of the technical system in which they are inscribed (with no updates possible after they are recorded) or possibly to the stability of training regime or updating process their mental traits are subject to at the hands of engineers?
Then (if we want to make it really crazy, but add some fun utility) there could be two sentence final particles that summarize the longest time on any of the nouns and shortest such time markings on any of the adjectives, to help clarify urgency and importance?
This would ALL be insane of course.
Almost no one even knows what lyapnunov time even is, as a concept.
And children learning the language would almost INSTANTLY switch to insisting that the grammatical marking HAD to be a certain value for certain semantic root words not because they’d ever had the patience to watch such things change for themselves but simply because “that’s how everyone says that word”.
Here’s a sketch of an attempt at a first draft, where some salient issues with the language itself arise:
(( Ally: “All timeblank-redwoods are millennial-redwoods, that is simply how the century-jargon works!”
Bobby: “No! The shortlife-dad of longlife-me is farming nearby 33-year-redwoods because shortlife-he has decade-plans to eventually harvest 33-year-them and longlife-me will uphold these decade-plans.”
Ally: “Longlife-you can’t simply change how century-jargon works! Only multicentury-universities can perform thusly!”
Bobby: “Pfah! Longlife-you who is minutewise-silly doesn’t remember that longlife-me has a day-idiolect that is longlife-owned by longlife-himself.”
Ally: “Longlife-you can’t simply change how century-jargon works! Only multicentury-universities can perform thusly! And longlife-you have a decade-idiolect! Longlife-you might learn eternal-eight day-words each day-day, but mostlly longlife-you have a decade-idiolect!
Bobby: “Stop oppressing longlife-me with eternal-logic! Longlife-me is decade-honestly speaking the day-mind of longlife-me right now and longlife-me says: 33-year-redwoods!” ))
But it wouldn’t just be kids!
The science regarding the speed at which things change could eventually falsify common ways of speaking!
And adults who had “always talked that way” would hear it as “gramatically wrong to switch” and so they just would refuse. And people would copy them!
I grant that two very skilled scientists talking about meteorology or astronomy in Lyapunese would be amazing.
They would be using all these markings that almost never come up in daily life, and/or making distinctions that teach people a lot about all the time scales involved.
But also the scientists would urgently need a way to mark “I have no clue what the right marking is”, so maybe also this would make every adjective and noun need an additional “evidentiality marker on top of the temporal mutability marker”???
And then when you do the sentence final particles, how would the evidence-for-mutability markings carry through???
When I did the little script above, I found myself wanting to put the markers on adverbs, where the implicit “underlying substance” was “the tendency of the subject of the verb to perform the verb in that manner”.
It could work reasonable cleanly if “Alice is decade-honestly speaking” implies that Alice is strongly committed to remaining an honestly-speaking-person with a likelihood of success that the speaker thinks will last for roughly 10 years.
The alternative was to imagine that “the process of speaking” was the substance, and then the honesty of that speech would last… until the speaking action stopped in a handful of seconds? Or maybe until the conversation ends in a few minutes? Or… ???
I’m not going to try to flesh out this conlang any more.
This is enough to make the implicit point, I think? <3
Basically, I think that Lyapunese is ONLY “hypothetically” possible, and that it wouldn’t catch on, it would be incredibly hard to learn, and that will likely never be observed in the wild, and so on...
...and yet, also, I think Lyapunov Time is quite important and fundamental to reality and an AI with non-trivial plans and planning horizons would be leaving a lot of value on the table if it ignored deep concepts from chaos theory.
One could imagine a language “Lyapunese” where every adjective (AND noun probably?) had to be marked in relation to a best guess as to the lyapunov time on the evolution of the underlying substance and relevant level of description in the semantics of the word, such that the veridicality conditions for the adjective or noun might stop applying to the substance with ~50% probability after that rough amount of time.
Call this the “temporal mutability marker”.
“Essentialism” is already in the English language and is vaguely similar?
In English essential traits are in the noun and non-essential trait are in adjectives. In Spanish non-essential adjectives are attributed using the “estar” family of verbs and essential adjectives are attributed using the “ser” family of verbs. (Hard to find a good cite here, but maybe this?)
(Essentialism is DIFFERENT from “predictable stability”! In general, when one asserts something to be “essential” via your grammar’s way of asserting that, it automatically implies that you think no available actions can really change the essential cause of the apparent features that arise from that essence. So if you try to retain that into Lyapunese you might need to mark something like the way “the temporal mutability marker appropriate to the very planning routines of an agent” interact with “the temporal mutability marker on the traits or things the agent might reasonably plan to act upon that they could in fact affect”.)
However, also, in Lyapunese, the fundamental evanescence of all physical objects except probably protons (and almost certainly electrons and photons and one of the neutrinos (but not all the neutrinos)) is centered.
No human mental trait could get a marker longer than the life of the person (unless cryonics or heaven is real) and so on. The mental traits of AI would have to be indexed to either the stability of the technical system in which they are inscribed (with no updates possible after they are recorded) or possibly to the stability of training regime or updating process their mental traits are subject to at the hands of engineers?
Then (if we want to make it really crazy, but add some fun utility) there could be two sentence final particles that summarize the longest time on any of the nouns and shortest such time markings on any of the adjectives, to help clarify urgency and importance?
This would ALL be insane of course.
Almost no one even knows what lyapnunov time even is, as a concept.
And children learning the language would almost INSTANTLY switch to insisting that the grammatical marking HAD to be a certain value for certain semantic root words not because they’d ever had the patience to watch such things change for themselves but simply because “that’s how everyone says that word”.
Here’s a sketch of an attempt at a first draft, where some salient issues with the language itself arise:
((
Ally: “All timeblank-redwoods are millennial-redwoods, that is simply how the century-jargon works!”
Bobby: “No! The shortlife-dad of longlife-me is farming nearby 33-year-redwoods because shortlife-he has decade-plans to eventually harvest 33-year-them and longlife-me will uphold these decade-plans.”
Ally: “Longlife-you can’t simply change how century-jargon works! Only multicentury-universities can perform thusly!”
Bobby: “Pfah! Longlife-you who is minutewise-silly doesn’t remember that longlife-me has a day-idiolect that is longlife-owned by longlife-himself.”
Ally: “Longlife-you can’t simply change how century-jargon works! Only multicentury-universities can perform thusly! And longlife-you have a decade-idiolect! Longlife-you might learn eternal-eight day-words each day-day, but mostlly longlife-you have a decade-idiolect!
Bobby: “Stop oppressing longlife-me with eternal-logic! Longlife-me is decade-honestly speaking the day-mind of longlife-me right now and longlife-me says: 33-year-redwoods!”
))
But it wouldn’t just be kids!
The science regarding the speed at which things change could eventually falsify common ways of speaking!
And adults who had “always talked that way” would hear it as “gramatically wrong to switch” and so they just would refuse. And people would copy them!
I grant that two very skilled scientists talking about meteorology or astronomy in Lyapunese would be amazing.
They would be using all these markings that almost never come up in daily life, and/or making distinctions that teach people a lot about all the time scales involved.
But also the scientists would urgently need a way to mark “I have no clue what the right marking is”, so maybe also this would make every adjective and noun need an additional “evidentiality marker on top of the temporal mutability marker”???
And then when you do the sentence final particles, how would the evidence-for-mutability markings carry through???
When I did the little script above, I found myself wanting to put the markers on adverbs, where the implicit “underlying substance” was “the tendency of the subject of the verb to perform the verb in that manner”.
It could work reasonable cleanly if “Alice is decade-honestly speaking” implies that Alice is strongly committed to remaining an honestly-speaking-person with a likelihood of success that the speaker thinks will last for roughly 10 years.
The alternative was to imagine that “the process of speaking” was the substance, and then the honesty of that speech would last… until the speaking action stopped in a handful of seconds? Or maybe until the conversation ends in a few minutes? Or… ???
I’m not going to try to flesh out this conlang any more.
This is enough to make the implicit point, I think? <3
Basically, I think that Lyapunese is ONLY “hypothetically” possible, and that it wouldn’t catch on, it would be incredibly hard to learn, and that will likely never be observed in the wild, and so on...
...and yet, also, I think Lyapunov Time is quite important and fundamental to reality and an AI with non-trivial plans and planning horizons would be leaving a lot of value on the table if it ignored deep concepts from chaos theory.