> Moreover, in line with this, Quijada has stated he does not believe a speaker would think necessarily any faster because even though Ithkuil is terse, a single word requires a lot more thought before it can be spoken than it would in a natural language.
I can't imagine what this is like to think before I have words to describe them. My current thought process is an inner monologue and I can't imagine how I'd think if I couldn't put words to my ideas immediately.
It happens in any other language too, at least for me. Sometimes you have a thought but can't quite put it into words. My guess is that it's simply more pronounced in Ithkuil.
Also, anecdotally, my inner monologue is sometimes behind my "real" thoughts. I can think faster than I can speak, even in my own head.
Do you ever think in images or actions? For example when I'm navigating in a new place I sometimes have this "reorientation moment" when I thought something is forward, but I realize it's actually to the right. At the moment it feels like I have a map of the world in my mind and it rotates around me :) I get slight vertigo even :)
As for actions you probably don't think to yourself "I should kick the ball in 3.. 2.. 1.. now! with THIS amount of spin and THAT power" - you just kick it when it's appropriate. Inner monologue would be too slow and too imprecise to think about these things. The same mode of thinking works for other things too. For example bad habits are automatic - you don't need to think "I should alt-tab to browser and open hackernews" - it's one thought expressed without language.
I mean, how do I even know what I say before I said it? Speaking is such a social act, you can definitely pan out and say a few words (even if it's fillers) before your brain catches up and you create a plan. And then you adjust your plan based on audience response. The obsession with precision is strange. Vagueness in spoken language is a feature, not a bug. It makes social situations easier to navigate if you have an "out".
Different languages are precise and vague in different domains.
For example English requires articles. You can have a book or the book. You cannot just have book. In my native language we don't have articles. If I want to specify it's "a" book I can (with the word to the effect of "some"), if I want to specify it's "the" book I can (with the equivalent of "this" or "that"). But by default we don't care.
On the other hand every verb encodes gender in my language. So when author wrote in English "Somebody in the audience said >I was with you<" - the translator has to guess what was the gender of that person. Because "was" is a different word depending on the gender of that person :)
EDIT: I'm -> I was, because in present tense verbs can be the same :)
I'd guess it's one of the non-Southern Slavic languages (e.g. Polish or Russian): those languages repurposed what used to be past participles (and so, just like adjectives, had genders) to be past tense verbs themselves. The Southern ones (e.g. Bulgarian) did not do this.
Or hand/arm, or finger/thumb: for some reason, while in Polish na ręce jest 5 palców, in English a hand has 4 fingers and 1 thumb on it, what's the deal with it?
That's totally wild. A huge part of my experience is trying to cram my thoughts into words that so rarely seem to fit. Makes texting (or any text communication without other context) stressful
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.
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.
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.
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?
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.
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.
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).
This makes no sense to me. What does computability have to do with semantic expressiveness? You've demonstrated that English can generate every sequence of tokens, not that it can express every concept.
If a concept can be expressed as a sequence of tokens, then bam, we're done.
If a concept can't be expressed as a sequence of tokens, then it doesn't really matter, because no language can, since some concepts exist which are not computable, and all languages have the same drawback.
Either way, "English is the linguistic equivalent of Turing complete" is valid, since it is at the top of the hierarchy, and as powerful as any other language (eg everything Ithkuil can express, so can English).
Imagine there is a concept that no language can express.
Now Ithkuil introduces a new word that is defined as that concept.
But English (and all other languages) still cannot express that concept. So the idea that all/most languages are expressively equivalent appears to be false.
The Chomsky hierarchy is about grammatical expressiveness, not about semantic expressiveness. I see no reason to believe that all languages have the same semantic expressiveness, which is what I understood the above claim to mean.
>Now Ithkuil introduces a new word that is defined as that concept.
You are assuming the consequent. How exactly did Ithkuil 'define' a new word as that concept if the concept can't be expressed in Ithkuil ("any language")?
Even assuming you could do such a thing, why couldn't every other language define a word to mean that concept in the same way? Why wouldn't English - in particular - just make the new Ithkuil word also an English word?
That's what the original form of the Sapir-Worf hypothesis was about. IIRC they were studying indigenous languages which lacked a word for numerical concepts. E.g. there's a bunch of languages/cultures which allegedly only have a words for "one", "two", and "many." Do people in these cultures not have a concept of "three"? Early studies suggested that might be the case. Sapir and Worf wondered if perhaps these people would be incapable of understanding the notion of "three", as appeared to be the case, because they lacked the word for it.
Turns out the results didn't replicate though, for a variety of reasons. Even in cases where the source languages didn't have higher numbers (which was much less common than originally thought), the speakers of course had a way of referencing numbers. In at least one case I believe the language had no unique number words at all, and what the linguist had thought was the word for "one" was in fact "alone", and the word for two was "couple" (as in "married couple"), and so on. They did indeed have such figurative words for larger numbers, and a place system for speaking of even larger amounts (e.g. a dozen-dozen people).
The Sapir-Worf hypothesis also falls flat in another way: if we cannot conceive of a concept without a word for it, then how did language develop? Clearly there must be a way to learn a new concept, think about it, and then be the first to coin a new word for it.
The original, strong form of the Sapir-Worf hypothesis is largely discredited today. But not the weak form, which is expressed thusly: "The language we speak influence the ways in which we think." This does seem to be largely true.
There are studies but I'd rather point to some anecdotal examples that illustrate more why I find this interesting. Learning math, critical thinking, and engineering has enabled me to reason much more precisely about things than I did as a teenager. Having been through the community college system, I also have observed the great difficulty a lot of people less interested in scientific thinking have had in adopting various concepts in critical thinking. Like the difference between conditional implication (if) and biconditional statements (if and only if). In everyday English we often say "if", when in fact "if and only if" is the more correct statement. Or we say "A or B" when we mean "A or B, but not both" (exclusive-or).
In English these longer forms are cumbersome, and the other alternative, "if" or "or", are in fact abbreviations of the longer forms "if and only if" and "A or B, but not both". So we decide that the shorter forms can mean either form, and resort to cumbersome language when clarity is required. If I tell you "I wear a hat if it is sunny" and you see me come indoors with my hat on, can you infer that it is sunny outside? No, because I said nothing about what I do when it is cloudy, not sunny outside. Maybe I wear a hat then too. If I had said "I wear a hat if and only if it is sunny" it'd be a different story. But that doesn't feel right because we know that people often shorten "if and only if" to just "if", and so we might feel justified in assuming the speaker meant they wear a hat if and only if it is sunny outside.
But in Lojban, Ithkuil, and Toaq, each truth table possibility is assigned a unique, approximately equal-length and unambiguous word/morpheme to represent it. It's like we coined the word "ba" to mean "if and only if." So the two statements become "I wear a hat if it is sunny" and "I wear a hat ba it is sunny". They are equal length so why wouldn't I pick the one that I mean? Then "if" can go back to meaning "if" exclusively, and never "if and only if".
Even the weak Sapir-Worf hypothesis is quite tricky to test in practice. It's likely true, but quite hard to prove. Guy Deutscher's excellent book Through the Language Glass documents much of the work that's been done in that area. Mostly, what's amenable to testing is colour perception and absolute direction, more than anything else.
That’s the Chuch-Turing thesis, combined with computational semantics. Computational semantics gives us formal logical statements for anything you might want to say, the Church-Turing thesis states that computational processes are equivalent to these logical formalisms, and you’ve accepted that any computational process can be represented in English (if verbosely).
The only exception would be things not express able on our semantic models which you might want to say. No one has ever been able to offer up an example of something that couldn’t be represented in a logical formalism, so…
By construction. Assume language L has concept C expressed by term T, while English does not have a simple term for C. Further, assume that English speakers have a frequent need to express C.
English will simply absorb T and trundle off like the demented lunatic it is. Examples include remuda, schadenfreude, and obi, all of which are copacetic English words after a remarkably short period of time.
One example where natural language is clearly suboptimal is math. Most people can do simple transformations on equations with small polynomials in memory. But that's thinking in formulas.
When you try to do it in plain English it becomes very hard.
I agree. Wouldn’t it be interesting then to see what we could accomplish if we unlocked similar formalisms for critical thinking and daily life?
I’m not a fan of Ithkuil actually, as I think it takes it too far. But I do wonder how much clearer our thinking would be if we had a logical / rational language as our mother tongue.
It's suboptimal but not impossible. As a child I won a prize (not first place, just a mention) in a national mathematical competition despite the fact no-one had taught me algebra yet. The judges found my reasoning "interesting" and I got the right answers to the puzzles.
Of course when, a couple of years later, I was taught algebra, I was wondering why I hadn't been told before!
Does Ithkuil have any provisions to express power dynamics between speaker and listener, say politeness levels? For instance all languages that I know have constructions to soften up requests, or make statements less assertive.
> 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.
> 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.
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.
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).
Where would you say Esperanto sits on the "natural" scale, since it's design goal is to be easy to learn and use as opposed to being highly compressed? As I wouldn't say its "[s]pelling, pronunciation, grammar, all shift around, and for all manner of reasons".
> has evolved to be compatible with our brain's language mechanism.
Under the hypothesis that there actually is such a "language mechanism": There would be a host of other mechanisms a language needs to compatible with - such as spoken and written word - that might have contributed adversarial constraints along the way (from the perspective of optimal adjustment of a language with respect to this mechanism).
There’s a whole strand of work in phonology (and other linguistics areas) called optimality theory, and the basic idea is that what you see in surface-level language is the the result of solving a competing constraints problem.
> which I believe might be the main reason why all conlangs from the past 100 years have failed to reach widespread adoption
Given how much people hate learning much more basic things—new programming languages, new operating systems, new tools—it seems hard to come to any conclusion about conlangs based on popularity! Languages are hard to learn, have immense cultural significance and are dominated by network effects. The potential advantages of a constructed language are, at best, illegible. What would push anybody to learn a brand new language well enough to actually use it?
Whether or not a language becomes popular tells us little to nothing about the language itself.
Generative grammar is extremely nativist. The truth is, people will learn anything society forces them to learn. In Piraha society nobody is forced to learn recursion and consequently nobody thinks recursively or in numbers.
The global languages of A.D. 2023 (English, French, etc...) are probably stuck in a similar "Chomsky tarpit" as Piraha, but because we're so globalized and together nobody is brave enough to look out of this tarpit.
I am constructing a conlang codename LXX (because it has 70 phonemes, 15 vowels and 55 consonants, but they were carefully chosen to be learnable to pronounce unlike Ithkuil) for the purpose of getting at least one human society out of this tarpit.
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
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.
> The Sapir–Whorf hypothesis postulates that a person's language influences their perceptions and cognitive patterns. Stanislav Kozlovsky proposed[8] in the Russian popular-scientific magazine Computerra that a fluent speaker of Ithkuil, accordingly, would think "about five or six times as fast" as a speaker of a typical natural language.
This seems to assume that one is thinking in their spoken language - but is that generally the case?
Outside of reading, writing and communicating, I damned near never think in natural language.
I don't know about other people, for anything technical I'm generally thinking in structures, not words.
The idea that language and thought are the same thing was discarded decades ago, and no linguist uses the pop term "Sapir-Whorf hypothesis" anymore (though linguistic relativity is still sometimes discussed).
If the cited claim was even remotely true, there would be significant differences in the "speed of thought" between speakers of natural languages also (which differ in semantic density). Such cognitive superiority would have obvious real-world consequences, but no such phenomenon appears to exist.
Precisely. Languages have more or less information density ("bits per symbol"), but when you look at information rate per unit time, it all works out to pretty similar.
Kind of like how QAM256 can convey more bits per symbol than QAM16, but it needs a lower baud for a given noise level, channel bandwidth, and error rate. You can squeeze out efficiency of a channel, but you can't overcome the noisy channel coding theorem max rate.
I don't think that language and thought are the same thing, and I certainly don't think language affects speed of thought, but I think in English all the time. It depends on the person, really. Some people have a strong internal monologue, others have a weak one, and others don't have one at all. Its a weird phenomenon, and AFAIK nobody knows why people have varying levels of internal monologuing.
I fucking love ithkuil, in the same way I love marble machines and Rube Goldberg devices. Like as a concept, in practice, how it sounds actually spoken, the writing system. Every thing about it is just so extra. It has its goals and design criteria, and it just goes full send on them, fuck everything else, like ease of pronunciation, and I respect that.
It's... interesting... to note that the example of a text in ithkuil takes up about as much space as the normal English text, due to how large ithkuil has to be written compared to regular text.
Spoken language tends to converge at ~40 bits per second for all natural human languages.[1] I wouldn't be surprised that if separate scripts were developed independently for each language they would all converge on a certain rate of information per square centimeter as well.
If you look at the v4 example sheet liked to at ithkuil.place you see that it is roughly the same length as English, keep in mind it has a larger alphabet so to really compare one needs to transcribe to the ascii adaptation
There was an article about Ithkuil a while back in the New Yorker [0]. It's a funny read. The most successful constructed language remains Esperanto, I believe.
My father is a fluent esperanto speaker. He's in the 10k group. I can understand the language a bit passively. I'm among 1 million people who can. There are 10 million people who have learned a bit of esperanto in their lives.
I don't think kingon numbers are nowhere close to that, but I only found a reference to the fluent klingon speakers (the 20-50 of them)
Also Esperanto has native speakers, i.e. children who grew up speaking Esperanto from an early age. It even has second generation native speakers: native speakers whose parents were also native speakers. Klingon has nothing like that.
A month ago, a father shared a very compelling story about how his young son started his own conlang and how ChatGPT was able to learn it from a few examples and converse into it.
Has that trend taken over more established conlang? Would it make sense for those communities to train an LLM to speak that language as fluently as possible to help materialize it: build lessons, have a conversation partner at any time, and possibly discuss grammar?
Even GPT-4 is not actually that good at it, and breaks down past a few simple rules. If you want an LLM to actually learn a language, today, you have to include enough of that language in its training data to begin with (I'm not sure if anyone tried it with fine-tuning).
Why is it that sci-fi authors trick otherwise smart people into thinking they have the key to some kind of universal truth as opposed to just being skilled prose stylists with a sense of drama?
Nah. Compare it with fantasy authors. Nobody thinks that Tolkien or Robert Jordan was putting forward some kind of aspirational project for better living through whatever. But people read Asimov and take psychohistory way too seriously, or read Heinlein and become a libertarian crank or invent a language. And let's not dwell too much on the extreme end of that scale, a certain writer with some funny ideas about mainstream psychology.
I think Tolkien in particular felt he had a moral and spiritual message, although the way to act on it, if any, presumably wouldn't be to pretend to be living in his Third Age.
> Nobody thinks that Tolkien ... was putting forward some kind of aspirational project for better living through whatever.
Actually I think that's a fair argument to make about LOTR. The simple, rustic wholesomeness of the Hobbits, the valor and respect of the Men, and even the friendship of a Dwarf and an Elf destroyed rapacious greed multiple times. I think even the Scouring episode at the end of the novels, while it may seem unimportant to the plot, can be considered important to the message of resisting industrialization. But I'm not a Tolkien scholar myself so take this all with a grain of salt.
There's a difference between a work of fiction having compelling themes and a political message and the way that some sci-fi fans seem to treat their favorite novels. There's almost a sense that the work is predictive rather than just prescriptive.
Tolkien is very obviously putting forward some kind of aspirational project for better living through whatever. Namely that small ordinary people (hobbits) who come from the Shire (the green pastures of England) and live in Hobbit holes (quant English cottages and farmhouses) are more powerful than kings, wizards, and unspeakable evil (Hitler), despite the fact that they live in the shadow of a golden age (the British Empire) that will never happen again. Plus the evils of industrialization, in case you got to the end without spotting the obvious. But social class matters intensely, and it's a shame that people are forgetting that. (The great virtue of Samwise Gamgee being not just that he is loyal, but most importantly that he knows his place in the order of things).
What is it about creative projects that tricks otherwise rational people into inventing unfounded characterizations of their creators? There's not one sentence in the entire article to suggest Quijada thinks that way.
Interesting that GPT 4 will happily translate to Esperanto, Klingon, Elvish, Aramaic, and Babylonian, but throws its up when asked to translate to Ithkuil!
> The Sapir–Whorf hypothesis postulates that a person's language influences their perceptions and cognitive patterns.
While this might be true to a certain extent [I personally doubt this hypothesis], the real limiter is how we can efficiently express or share the language with others. We are social creatures, after all, and we cannot thrive in a vacuum. If the language is too hard to learn among the general population - the ideas cannot be spread out and disseminated by others. Thus we can't engage in a positive feedback loop where ideas are also shared back to us (supposing as individuals, we only knew this language).
That's my theory, anyways, and one of the reasons why English has been so successful - not because it's the "best" language [spoiler: it's not], but for all its flaws - it is relatively easy to learn and communicate (and express ideas) in.
96 comments
[ 3.6 ms ] story [ 139 ms ] threadI can't imagine what this is like to think before I have words to describe them. My current thought process is an inner monologue and I can't imagine how I'd think if I couldn't put words to my ideas immediately.
Also, anecdotally, my inner monologue is sometimes behind my "real" thoughts. I can think faster than I can speak, even in my own head.
As for actions you probably don't think to yourself "I should kick the ball in 3.. 2.. 1.. now! with THIS amount of spin and THAT power" - you just kick it when it's appropriate. Inner monologue would be too slow and too imprecise to think about these things. The same mode of thinking works for other things too. For example bad habits are automatic - you don't need to think "I should alt-tab to browser and open hackernews" - it's one thought expressed without language.
For example English requires articles. You can have a book or the book. You cannot just have book. In my native language we don't have articles. If I want to specify it's "a" book I can (with the word to the effect of "some"), if I want to specify it's "the" book I can (with the equivalent of "this" or "that"). But by default we don't care.
On the other hand every verb encodes gender in my language. So when author wrote in English "Somebody in the audience said >I was with you<" - the translator has to guess what was the gender of that person. Because "was" is a different word depending on the gender of that person :)
EDIT: I'm -> I was, because in present tense verbs can be the same :)
https://www.youtube.com/watch?v=mkNvlqEM8rk
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.
Aistlaţervièllîmļ - skipping an opportunity because it's not ideal, despite the likelihood that an ideal instance will never come
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?
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.
Proof?
Think about it this way, I give you instructions in english, and you have to follow them:
You could follow that process to the letter, given enough time, focus, and tape (insert xkcd reference about rocks in the desert).If a concept can't be expressed as a sequence of tokens, then it doesn't really matter, because no language can, since some concepts exist which are not computable, and all languages have the same drawback.
Either way, "English is the linguistic equivalent of Turing complete" is valid, since it is at the top of the hierarchy, and as powerful as any other language (eg everything Ithkuil can express, so can English).
Imagine there is a concept that no language can express.
Now Ithkuil introduces a new word that is defined as that concept.
But English (and all other languages) still cannot express that concept. So the idea that all/most languages are expressively equivalent appears to be false.
The Chomsky hierarchy is about grammatical expressiveness, not about semantic expressiveness. I see no reason to believe that all languages have the same semantic expressiveness, which is what I understood the above claim to mean.
You are assuming the consequent. How exactly did Ithkuil 'define' a new word as that concept if the concept can't be expressed in Ithkuil ("any language")?
Even assuming you could do such a thing, why couldn't every other language define a word to mean that concept in the same way? Why wouldn't English - in particular - just make the new Ithkuil word also an English word?
Turns out the results didn't replicate though, for a variety of reasons. Even in cases where the source languages didn't have higher numbers (which was much less common than originally thought), the speakers of course had a way of referencing numbers. In at least one case I believe the language had no unique number words at all, and what the linguist had thought was the word for "one" was in fact "alone", and the word for two was "couple" (as in "married couple"), and so on. They did indeed have such figurative words for larger numbers, and a place system for speaking of even larger amounts (e.g. a dozen-dozen people).
The Sapir-Worf hypothesis also falls flat in another way: if we cannot conceive of a concept without a word for it, then how did language develop? Clearly there must be a way to learn a new concept, think about it, and then be the first to coin a new word for it.
The original, strong form of the Sapir-Worf hypothesis is largely discredited today. But not the weak form, which is expressed thusly: "The language we speak influence the ways in which we think." This does seem to be largely true.
There are studies but I'd rather point to some anecdotal examples that illustrate more why I find this interesting. Learning math, critical thinking, and engineering has enabled me to reason much more precisely about things than I did as a teenager. Having been through the community college system, I also have observed the great difficulty a lot of people less interested in scientific thinking have had in adopting various concepts in critical thinking. Like the difference between conditional implication (if) and biconditional statements (if and only if). In everyday English we often say "if", when in fact "if and only if" is the more correct statement. Or we say "A or B" when we mean "A or B, but not both" (exclusive-or).
In English these longer forms are cumbersome, and the other alternative, "if" or "or", are in fact abbreviations of the longer forms "if and only if" and "A or B, but not both". So we decide that the shorter forms can mean either form, and resort to cumbersome language when clarity is required. If I tell you "I wear a hat if it is sunny" and you see me come indoors with my hat on, can you infer that it is sunny outside? No, because I said nothing about what I do when it is cloudy, not sunny outside. Maybe I wear a hat then too. If I had said "I wear a hat if and only if it is sunny" it'd be a different story. But that doesn't feel right because we know that people often shorten "if and only if" to just "if", and so we might feel justified in assuming the speaker meant they wear a hat if and only if it is sunny outside.
But in Lojban, Ithkuil, and Toaq, each truth table possibility is assigned a unique, approximately equal-length and unambiguous word/morpheme to represent it. It's like we coined the word "ba" to mean "if and only if." So the two statements become "I wear a hat if it is sunny" and "I wear a hat ba it is sunny". They are equal length so why wouldn't I pick the one that I mean? Then "if" can go back to meaning "if" exclusively, and never "if and only if".
You might say that people aren&...
Is this true?
The only exception would be things not express able on our semantic models which you might want to say. No one has ever been able to offer up an example of something that couldn’t be represented in a logical formalism, so…
English will simply absorb T and trundle off like the demented lunatic it is. Examples include remuda, schadenfreude, and obi, all of which are copacetic English words after a remarkably short period of time.
When you try to do it in plain English it becomes very hard.
I’m not a fan of Ithkuil actually, as I think it takes it too far. But I do wonder how much clearer our thinking would be if we had a logical / rational language as our mother tongue.
Of course when, a couple of years later, I was taught algebra, I was wondering why I hadn't been told before!
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.
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.
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/
Which mostly cites 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).
Under the hypothesis that there actually is such a "language mechanism": There would be a host of other mechanisms a language needs to compatible with - such as spoken and written word - that might have contributed adversarial constraints along the way (from the perspective of optimal adjustment of a language with respect to this mechanism).
Given how much people hate learning much more basic things—new programming languages, new operating systems, new tools—it seems hard to come to any conclusion about conlangs based on popularity! Languages are hard to learn, have immense cultural significance and are dominated by network effects. The potential advantages of a constructed language are, at best, illegible. What would push anybody to learn a brand new language well enough to actually use it?
Whether or not a language becomes popular tells us little to nothing about the language itself.
The global languages of A.D. 2023 (English, French, etc...) are probably stuck in a similar "Chomsky tarpit" as Piraha, but because we're so globalized and together nobody is brave enough to look out of this tarpit.
I am constructing a conlang codename LXX (because it has 70 phonemes, 15 vowels and 55 consonants, but they were carefully chosen to be learnable to pronounce unlike Ithkuil) for the purpose of getting at least one human society out of this tarpit.
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.
This seems to assume that one is thinking in their spoken language - but is that generally the case?
Outside of reading, writing and communicating, I damned near never think in natural language.
I don't know about other people, for anything technical I'm generally thinking in structures, not words.
If the cited claim was even remotely true, there would be significant differences in the "speed of thought" between speakers of natural languages also (which differ in semantic density). Such cognitive superiority would have obvious real-world consequences, but no such phenomenon appears to exist.
Kind of like how QAM256 can convey more bits per symbol than QAM16, but it needs a lower baud for a given noise level, channel bandwidth, and error rate. You can squeeze out efficiency of a channel, but you can't overcome the noisy channel coding theorem max rate.
https://seantrott.github.io/information/
https://upload.wikimedia.org/wikipedia/commons/c/c9/Ithkuil_...
[1] https://www.science.org/content/article/human-speech-may-hav...
[0] https://www.newyorker.com/magazine/2012/12/24/utopian-for-be...
There are 10k fluent esperanto speakers and an order of magnitude more "active speakers".
According to http://klingon.wiki/En/FluentSpeakers there are ~30 fluent klingon speakers
My father is a fluent esperanto speaker. He's in the 10k group. I can understand the language a bit passively. I'm among 1 million people who can. There are 10 million people who have learned a bit of esperanto in their lives.
I don't think kingon numbers are nowhere close to that, but I only found a reference to the fluent klingon speakers (the 20-50 of them)
https://discord.com/invite/WgFrX8J
Has that trend taken over more established conlang? Would it make sense for those communities to train an LLM to speak that language as fluently as possible to help materialize it: build lessons, have a conversation partner at any time, and possibly discuss grammar?
Why is it that sci-fi authors trick otherwise smart people into thinking they have the key to some kind of universal truth as opposed to just being skilled prose stylists with a sense of drama?
Actually I think that's a fair argument to make about LOTR. The simple, rustic wholesomeness of the Hobbits, the valor and respect of the Men, and even the friendship of a Dwarf and an Elf destroyed rapacious greed multiple times. I think even the Scouring episode at the end of the novels, while it may seem unimportant to the plot, can be considered important to the message of resisting industrialization. But I'm not a Tolkien scholar myself so take this all with a grain of salt.
While this might be true to a certain extent [I personally doubt this hypothesis], the real limiter is how we can efficiently express or share the language with others. We are social creatures, after all, and we cannot thrive in a vacuum. If the language is too hard to learn among the general population - the ideas cannot be spread out and disseminated by others. Thus we can't engage in a positive feedback loop where ideas are also shared back to us (supposing as individuals, we only knew this language).
That's my theory, anyways, and one of the reasons why English has been so successful - not because it's the "best" language [spoiler: it's not], but for all its flaws - it is relatively easy to learn and communicate (and express ideas) in.