6 ms·

Like typing with a modified Dvorak like keyboard layout I imagine it would be more comfortable to think, read and speak in Ithkuil. It took me three months to s

Like typing with a modified Dvorak like keyboard layout I imagine it would be more comfortable to think, read and speak in Ithkuil. It took me three months to switch from Qwerty to a Dvorak-like 20 years ago. I still can type both, even though it feels significantly less comfortable to type in Qwerty, I guess I’m spoiled. How long would it take for me to learn Ithkuil? I don’t have to speak it as there is no one understanding it (except other dorks on the internet), so it’s better to learn to read and write my personal notes in Ithkuil. This means I should also have a dvorak like Ithkuil TNIL keyboard layout. Is there such a thing? Ideally I also want to converse with a LLM trained on Ithkuil texts. As there are very few TNIL (the latest version) texts I wonder if somekind of translator can be build that can translate a large relevant body of English texts for me into Ithkuil.

ibejoeb 3y ago
It seems more like, if you do manage to internalize some of the language, that you might be able to augment your thinking language rather than replace it. For the things that are not readily expressible in your native language, Ithkuil could serve to characterize those thoughts. I do wonder how robust recall would be, though, in terms of time and fidelity.
- lordnacho 3y ago
  What are some interesting examples of things I could express in Ithkuil? Suppose for an English speaker. I get that it will be weird to write an answer in English. Could be really interesting if someone has a good example though.
  - thrtythreeforty 3y ago
    One of my favorite examples: Aistlaţervièllîmļ - skipping an opportunity because it's not ideal, despite the likelihood that an ideal instance will never come
    - p-e-w 3y ago
      But that concept can also be expressed in English – which you just did. If you simply run the English description you gave through a Huffman coder, you should get a compressed version that is the same length or even shorter than the Ithkuil word (English usually compresses to about 1 bit per character). Is Ithkuil anything more than a compressed encoding of English? Can it express any concept that English is fundamentally incapable of expressing?
      - adastra22 3y ago
        No, because English is the linguistic equivalent of Turing complete. There is no concept you can’t express (if verbosely) in English. But the weak form of Sapir-Worf is that verbosity and grammar matters, that we tend to think in shortest forms, and that if we had a mother tongue or native-level language that was expressive, precise, and minimally ambiguous, that would be reflected in the clarity of our thinking.
        p-e-w 3y ago
        > English is the linguistic equivalent of Turing complete Proof?
        adastra22 3y ago
        You are asking for a proof of a negative, which can’t be proven. Nevertheless I offer the entire field of computational semantics, which has yet to generate a counter-example.
        kortex 3y ago
        English (or any human language with the possible exception of Pirahã) is not a formal grammar, but if it was, it would behave like a recursively enumerable one, at the top of the Chomsky hierarchy. Given enough time and enough text, you can express anything computable. Think about it this way, I give you instructions in english, and you have to follow them: You are acting as a turing machine. Your symbols are s1, s2, ..., sn. Here are your processing rules: ... Here is your tape: ... Begin! You could follow that process to the letter, given enough time, focus, and tape (insert xkcd reference about rocks in the desert).
p-e-w 3y ago
> I imagine it would be more comfortable to think, read and speak in Ithkuil There are some cognitive scientists who believe that the ability to use language is built into the human brain, and (natural) language acquisition basically involves just setting the "parameters" of that built-in system. If that is true, there's no reason to assume that a highly constructed language like Ithkuil can ever be as comfortable to use as a natural language that has evolved to be compatible with our brain's language mechanism. Language has been with humans for a very long time (up to 100k years according to some hypotheses); it's not just another tool like a keyboard layout, which I believe might be the main reason why all conlangs from the past 100 years have failed to reach widespread adoption, despite their numerous claimed advantages. The assumption that humans are able to intellectually construct a language that is usable as a human language in the same way natural languages are lacks any evidence.
- meepmorp 3y ago
  > There are some cognitive scientists who believe that the ability to use language is built into the human brain, and (natural) language acquisition basically involves just setting the "parameters" of that built-in system. This, by the way, is essentially what Chomsky means by “universal grammar,” in his writing - the inborn language capability and its fundamental structure. His theoretical framework even went through a period where it was called Principles and Parameters.
- kortex 3y ago
  As someone who tinkers with conlangs as a hobby and has been working on a similarly highly-engineered language not unlike ithkuil in features, this is one of the main things I struggle with. Actual natural language is very fluid. Spelling, pronunciation, grammar, all shift around, and for all manner of reasons. But the main one is efficiency. Humans are always subconsciously trying to cram as much information into as little time as possible, while preserving enough intent. This means extra redundant information whittles away, but there has to be enough redundancy for error correction. With these highly compressed languages, the problem becomes that a slight mispronounciation or misspelling can totally mess with the meaning in ways which lead to confusion. There are just too many phonemic sounds (a given sound difference is phonemic if the distinction matters, eg in English, p and p' (aspiration) is non-phonemic, but in Korean it is) When you actually analyze languages for their "baud rate" so to speak, languages with more symbol diversity (English) have less phonemes per unit time, while languages with less symbol diversity (Italian) have more phonemes, but the overall entropy per unit time works out pretty much the same. (I can't find my exact source right now, but if you poke around keywords like spoken language entropy rate, there's quite a bunch). Basically when you have machine code as a language, you pretty much need machine brain to process it. Edit: here's a start of some sources: https://seantrott.github.io/information/ https://seantrott.github.io/information/ Which mostly cites https://www.science.org/doi/10.1126/sciadv.aaw2594 https://www.science.org/doi/10.1126/sciadv.aaw2594 Also: re why no conlangs have taken over, momentum is a huge thing. We still have big iron running COBOL. Now imagine you hear COBOL from the moment you are born, for hours a day, you start babbling in COBOL, basically everyone around you converses in COBOL. It would be thoroughly entrenched. There's a reason "English follows other languages down dark alleys, beats them up and rifles through their pockets for vocabulary". It is the lingua franca. It's incredibly flexible and extensible, and combined with superpower hegemony, it dominates. Any dominant language is bound to acrete vocab and eject dead weight (though why we still have dental fricatives is really weird).
pohl 3y ago
Just yesterday I was pondering Ithkuil, the Sapir–Whorf Hypothesis, and LLMs.
kortex 3y ago
Fluency in Ithkuil is probably impossible practically speaking. John Quijada himself doesn't consider himself fluent. I actually think LLMs trained on conlangs is a super cool idea, but you'd need a huge corpus to do that. Kind of a bootstrapping problem.
- fsiefken 3y ago
  The new 2023 Ithkuil addresses part of the learnability. Yes I need a large corpus, that’s why I thought maybe some kind of rough automatic translator can be built to aid in getting this body and the LLM trained
  - ted_dunning 3y ago
    But that will only train the LLM on Ithkuil expressions that came from English (or whatever you translated from). The interesting point is the expressivity beyond that.

ibejoeb 3y ago

It seems more like, if you do manage to internalize some of the language, that you might be able to augment your thinking language rather than replace it. For the things that are not readily expressible in your native language, Ithkuil could serve to characterize those thoughts. I do wonder how robust recall would be, though, in terms of time and fidelity.

lordnacho 3y ago
What are some interesting examples of things I could express in Ithkuil? Suppose for an English speaker. I get that it will be weird to write an answer in English. Could be really interesting if someone has a good example though.
- thrtythreeforty 3y ago
  One of my favorite examples: Aistlaţervièllîmļ - skipping an opportunity because it's not ideal, despite the likelihood that an ideal instance will never come
  - p-e-w 3y ago
    But that concept can also be expressed in English – which you just did. If you simply run the English description you gave through a Huffman coder, you should get a compressed version that is the same length or even shorter than the Ithkuil word (English usually compresses to about 1 bit per character). Is Ithkuil anything more than a compressed encoding of English? Can it express any concept that English is fundamentally incapable of expressing?
    - adastra22 3y ago
      No, because English is the linguistic equivalent of Turing complete. There is no concept you can’t express (if verbosely) in English. But the weak form of Sapir-Worf is that verbosity and grammar matters, that we tend to think in shortest forms, and that if we had a mother tongue or native-level language that was expressive, precise, and minimally ambiguous, that would be reflected in the clarity of our thinking.
      - p-e-w 3y ago
        > English is the linguistic equivalent of Turing complete Proof?
        adastra22 3y ago
        You are asking for a proof of a negative, which can’t be proven. Nevertheless I offer the entire field of computational semantics, which has yet to generate a counter-example.
        kortex 3y ago
        English (or any human language with the possible exception of Pirahã) is not a formal grammar, but if it was, it would behave like a recursively enumerable one, at the top of the Chomsky hierarchy. Given enough time and enough text, you can express anything computable. Think about it this way, I give you instructions in english, and you have to follow them: You are acting as a turing machine. Your symbols are s1, s2, ..., sn. Here are your processing rules: ... Here is your tape: ... Begin! You could follow that process to the letter, given enough time, focus, and tape (insert xkcd reference about rocks in the desert).

p-e-w 3y ago

> I imagine it would be more comfortable to think, read and speak in Ithkuil There are some cognitive scientists who believe that the ability to use language is built into the human brain, and (natural) language acquisition basically involves just setting the "parameters" of that built-in system. If that is true, there's no reason to assume that a highly constructed language like Ithkuil can ever be as comfortable to use as a natural language that has evolved to be compatible with our brain's language mechanism. Language has been with humans for a very long time (up to 100k years according to some hypotheses); it's not just another tool like a keyboard layout, which I believe might be the main reason why all conlangs from the past 100 years have failed to reach widespread adoption, despite their numerous claimed advantages. The assumption that humans are able to intellectually construct a language that is usable as a human language in the same way natural languages are lacks any evidence.

meepmorp 3y ago
> There are some cognitive scientists who believe that the ability to use language is built into the human brain, and (natural) language acquisition basically involves just setting the "parameters" of that built-in system. This, by the way, is essentially what Chomsky means by “universal grammar,” in his writing - the inborn language capability and its fundamental structure. His theoretical framework even went through a period where it was called Principles and Parameters.
kortex 3y ago
As someone who tinkers with conlangs as a hobby and has been working on a similarly highly-engineered language not unlike ithkuil in features, this is one of the main things I struggle with. Actual natural language is very fluid. Spelling, pronunciation, grammar, all shift around, and for all manner of reasons. But the main one is efficiency. Humans are always subconsciously trying to cram as much information into as little time as possible, while preserving enough intent. This means extra redundant information whittles away, but there has to be enough redundancy for error correction. With these highly compressed languages, the problem becomes that a slight mispronounciation or misspelling can totally mess with the meaning in ways which lead to confusion. There are just too many phonemic sounds (a given sound difference is phonemic if the distinction matters, eg in English, p and p' (aspiration) is non-phonemic, but in Korean it is) When you actually analyze languages for their "baud rate" so to speak, languages with more symbol diversity (English) have less phonemes per unit time, while languages with less symbol diversity (Italian) have more phonemes, but the overall entropy per unit time works out pretty much the same. (I can't find my exact source right now, but if you poke around keywords like spoken language entropy rate, there's quite a bunch). Basically when you have machine code as a language, you pretty much need machine brain to process it. Edit: here's a start of some sources: https://seantrott.github.io/information/ https://seantrott.github.io/information/ Which mostly cites https://www.science.org/doi/10.1126/sciadv.aaw2594 https://www.science.org/doi/10.1126/sciadv.aaw2594 Also: re why no conlangs have taken over, momentum is a huge thing. We still have big iron running COBOL. Now imagine you hear COBOL from the moment you are born, for hours a day, you start babbling in COBOL, basically everyone around you converses in COBOL. It would be thoroughly entrenched. There's a reason "English follows other languages down dark alleys, beats them up and rifles through their pockets for vocabulary". It is the lingua franca. It's incredibly flexible and extensible, and combined with superpower hegemony, it dominates. Any dominant language is bound to acrete vocab and eject dead weight (though why we still have dental fricatives is really weird).

pohl 3y ago

Just yesterday I was pondering Ithkuil, the Sapir–Whorf Hypothesis, and LLMs.

kortex 3y ago

Fluency in Ithkuil is probably impossible practically speaking. John Quijada himself doesn't consider himself fluent. I actually think LLMs trained on conlangs is a super cool idea, but you'd need a huge corpus to do that. Kind of a bootstrapping problem.

fsiefken 3y ago
The new 2023 Ithkuil addresses part of the learnability. Yes I need a large corpus, that’s why I thought maybe some kind of rough automatic translator can be built to aid in getting this body and the LLM trained
- ted_dunning 3y ago
  But that will only train the LLM on Ithkuil expressions that came from English (or whatever you translated from). The interesting point is the expressivity beyond that.