11 ms·

Lojban: Constructed language to eliminate ambiguity from communication

110 points by lordylord 3y ago

ftxbro 3y ago
I tried it with gpt 3.5 but it doesn't seem to know lojban so well. I asked it to translate the following in to Lojban, then separately after deleting that history I asked it to translate that to English. Here's what it did: Original English: "I heard of the TinyStories research project that unlocked cognitive capabilities in LLMs with fewer parameters. I wonder if something similar would be possible with lojban?" Lojban round trip English: "We interpret the use of the LLM as a way to confuse the knowledge from TinyStories.gy, which is greater than that from the experimental findings. I find it amusing to use Lojban instead. Does the ability to confuse the experimental findings also mean the ability to confuse Lojban?"
- sigstoat 3y ago
  hmm since there is a CFG for lojban, it’d be interesting to give [some LLM you can do fancy things woth] what corpus of lojban there is, and then use the work on constraining LLMs to a grammar, and see if that’d help its output.
- int_19h 3y ago
  GPT-4 doesn't do well at it, either. The problem is that there simply isn't that much text written in Lojban, so it can't learn it the way it does other languages. You can see similar problems with other languages with a relatively small corpus, such as Old Norse.
  - ftxbro 3y ago
    Yes this must be it. I'm curious what it would look like, to compare more directly. Like a LLM trained only on Lojban vs. an LLM trained only on an equally small amount of English.
lez 3y ago
I like this point very much: "You can be as vague or detailed as you like when speaking lojban. For example, specifying tense (past, present or future) or number (singular or plural) is optional when they're clear from context.". LMBTQ movement has came into my mind immediately. I think they would have more success by introducing a new, 'don't care about your sex' pronoun to the language instead of adding a third one that just complicates matters.
- themodelplumber 3y ago
  > LMBTQ movement has came into my mind immediately. I think they would have more success by introducing a new, 'don't care about your sex' pronoun to the language instead of adding a third one that just complicates matters. I thought it was more of a pro-caring thing than a pro-uncaring thing. Those focusing too much on the latter are really quite good at bumbling into insulting behavior...
  - lez 3y ago
    Slippery slope fallacy here. Why should I care if my online customer is he/she/they as long as they pay the price? Or if I talk about technology through the Internet? All I want is effective communication with no insults.
    - scotty79 3y ago
      In many languages you are obligated to care whether a table is a he or a she. In English you are obligated to care whether it's a table or the table. I'm all for languages with optional caring about stuff.
    - themodelplumber 3y ago
      Is this meant to describe how a slippery slope forms, or call one out? Effective communication involving other people only sometimes overlaps with this subjective-first, transactional approach, the: I care about, I talk about, or All I want is. (It's a great view for designing a specific type of community, with business-transactional perspective + valuing of tightening subjective logic and avoiding fallacy. Well, this will always result in a specific type of tech community with some special focus on business and economic factors as combined with tech.) And I guess if we effectively cherry-pick the premise, with straw people assuming their various positions on the stage, we can easily justify a more anonymizing, transactional view of language. "This is how I prefer to view my life, therefore the needed language looks like..." In this view, language doesn't need all that extra stuff. Keep it direct, functional, logical, and avoid fallacies especially. But this isn't really getting at the relating capabilities of language for example, like when there are bigger problems at hand and diverse mindsets working together to solve them. Instead, it ends up bringing minds into silos, as broader personality theory predicts. So the I, I, me, is really more about self-support. With this view, individuals effectively make relating all about the subjective-them 24/7, and about whether they are good/not good. They see relating with others as "we will each be our best and get what we each want" as if it's a universal creed. The one philosophy that just makes sense. Do others even exist? And is this actively mean? No. Perhaps a better term is effectively: "passively careless," to borrow a phrase from Sabine Hossenfelder. And even in the subjective-first view it still does not make for a high-quality individual, which is also bad news. It's unfortunate, because not only does this view raise the stakes on their end with no good reason, creating things like martyrdom issues, but it also perpetuates the avoidance of effective contact with others, effectively making "others" the persistent bête noir.
sowbug 3y ago
If memory serves, Attempto was recently on HN. It's another natural (invented) language that parses unambiguously. https://en.wikipedia.org/wiki/Attempto_Controlled_English https://en.wikipedia.org/wiki/Attempto_Controlled_English
yosito 3y ago
I don't know if eliminating ambiguity is even possible, much less a key feature of the language. There's even prominent ambiguity on this page: > Lojban means different things to different people
- vidarh 3y ago
  A lot of communication is intentionally ambiguous at that. Be it diplomatic texts written to let both sides present a victory, or comedy intentionally obscuring key facts for a later reveal. Having the ability to be unambiguous is great, but human communication is full of situations where that's not what we want.
- jmillikin 3y ago
  The goal of Lojban is to eliminate syntactic/grammatical ambiguity: Time flies like an arrow, fruit flies like a banana. In this example, "flies" could be either a noun or verb depending on context. That sort of construct isn't possible in Lojban. Of course, it's still possible to use Lojban to express statements that have an ambiguous or unclear meaning.
  - djtango 3y ago
    Ah, classic HN Saturday morning - finding myself dreaming up Hungarian notation in English to disambiguate grammatical types nTime vFlies avLike an nArrow, nFruit vFlies avLike a nBanana It's ugly but if we all adhere to this convention and haze our juniors, we'll eliminate most our problems
    - prerok 3y ago
      avLike still has two different meanings ;) Actually, I am not sure that having different meanings of the same word in different contexts is really the problematic part causing ambiguity in conversations. For example, sarcasm is a much more difficult problem. The words are the same, the meaning of the sentences is the same but the author's intent, the meaning behind those sentences is opposite if they are being sarcastic or not.
    - 3y ago
wirrbel 3y ago
I spent a few weeks with Lojban around 2006. imho it felt really dense as a language. Definitely interesting. Back then there wasn’t as much material in Lojban as is now available.
tagyro 3y ago
I vaguely remember a TNG* episode in which humans encounter an alien race that found our language too ...inaccurate, and spent a lot of time going over each word and sentence, to precisely define its meaning. * Was it TNG or am I dreaming? *edit: Thank you everyone, but it wasn't "Darmok". The humans and the aliens were working, for a long time, on a treaty of sorts because the human language was so confusing and ambiguous and they had to go over every small details, over and over ... Need to re-watch TNG
- sgtnoodle 3y ago
  I believe I've seen every TNG episode at least once, and that's not ringing any bells. Maybe it was an original series episode? Despite my best efforts, I haven't been able to make it through more than one or two episodes of the original.
- pwnmonkey 3y ago
  Shaka, when the walls fell?
  - sgtnoodle 3y ago
    I recall that episode, but I'm not sure that's what they were referencing.
- numpad0 3y ago
  “Darmok” is Internet favorite, about a species which language is exclusively in analogy that the translator could not be set up, and they decide to kidnap Picard to establish vocabulary over a campfire. There were couple episodes in Voyager in which aliens deem humans interesting but lesser, and tries to experiment or study on crews. One that I remember ended up in a ship self destruction, in another the doctor leaving his copy. As to why they didn’t simply ask Borg for a zipped archive, I don’t know. The Dominion commanders throughout DS9 must have made similar remarks, and so should Cardassians have. Q would outright say so in spite, corrupted Data might. Maybe combination of those? Or maybe I’m just a suboptimal Trekkie :p
  - 3y ago
Decabytes 3y ago
I wonder if LLM entirely in lobjan would require less text to train to be just as good as the models we have today due to its higher clarity in expression. I also feel the same way about Esperanto due to its high regularity in syntax. Part of me also wonders if an ai model could be optimized by training it on a language specifically invented to be both easy for the ai to understand and the human to communicate in. Maybe the language was made up of lots of 4 letter words bringing the number of words to tokens closer to 1:1, or injecting more context per word so sentences could be shorter and use less tokens
- ftxbro 3y ago
  you should check TinyStories if you didn't know of it already https://arxiv.org/abs/2305.07759 https://arxiv.org/abs/2305.07759
adastra22 3y ago
I like the idea of Lojban. I hate the implementation. It’s basically spoken math notation for predicate logic, which falls flat well short of the goal. There are much better logical frameworks that have been developed for computational semantics (e.g. Davidsonian event semantics). Lojban’s formalism was out of date already when it was proposed as Loglan half a century ago. If there is anyone out there who likes the idea of Lojban but want something more practical and with a firmer foundation to boot, contact me. I’ve got ideas and I’d love to find some people to work on this with.
- Beltiras 3y ago
  I wonder if you could get the intended result for any language with a simple ruleset. You don't need a language to reduce/eliminate miscommunications, you need intent and understanding.
  - msla 3y ago
    Yes, you could, because any natural language can be used to teach mathematics. Lojban is, like any notation, a convenience for people who understand the concepts, but the concepts can be expressed in a natural language just as precisely, albeit less concisely. In fact, as Florian Cajori stated in his book "A History of Mathematical Notations", the rise of mathematical notation was opposed by people who preferred the older, natural language style of doing mathematics, what Cajori termed the "struggle between symbolists and rhetoricians." So, yes, logic is mathematics, and humans did mathematics in natural language for a very long time before we invented the conlang of mathematical notation. Moving more of the ideas into what is, ultimately, a more expressive notation is not a fundamental shift.
  - jpk 3y ago
    This is kinda what we call "legalese". It's a sort of formalized subset of English (or whatever language) that leans on standardized turns of phrase, it tends to set up definitions for terms that are then used throughout a document, etc. All in order to reduce misunderstanding and (hopefully) be easy to interpret in the event of a dispute. However, we have whole judicial systems that spend a non-trivial fraction of their time interpreting legal verbiage. So clearly it falls short at least some of the time, otherwise courts could be, in part, automated away. Maybe that's because it's too hard or not possible with natural languages? Or the legalese ruleset just isn't refined enough?
baby 3y ago
After learning toki pona (the simplest conlang perhaps?) I got attracted by Lojban due to the machine-parsable aspect of the conlang, but I found it really tedious and not fun to learn. I started working on a conlang with similar ideas a while back (emojis in the language, machine parsable) but less strict and nicer sounding. Never ended up with anything though :)
- adastra22 3y ago
  You might like Mini / Mini Kore. I no joke believe that the eventual end state of linguistic evolution is an ideographic written language with emoji as characters.
  - BiteCode_dev 3y ago
    That would be terrible. Ideographic written languages are hard to learn (you need to memorize much more to get basic skills) and hard to parse (tokenization is more difficult). Of course, if that means the entire world speaks the same language, the price would be worth it. But I doubt we will converge on anything that is difficult to master, since water takes the path of least resistance.
    - adastra22 3y ago
      That's not actually true. When MacArthur tried to get rid of the use of Chinese characters in Japanese writing during the post-WW2 occupation, he backed off when it was embarrassingly pointed out that Japan had a higher literacy rate than the US, despite having less of an investment in state schools at the time. Studies have shown that it does not take more time or effort for a child to learn to read and write Chinese or Japanese (assuming those are their native spoken languages) than for an American student to learn to read/write English. It may defy what you think you know as common sense, but it is a replicable observation. The reason for this is that nobody except little children actually read by sounding out individual letters of words. That would be too much for the human mind to process in real time. You learn to read by recognizing particular clusters of lines on a paper or screen as having semantic meaning, and your brain translates that directly. The imlpiactoin is the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a wlohe. (Did you read that OK?) So essentially, you and a Chinese person both put in the same amount of effort to learn the same number of squiggly lines on paper as having semantic meaning, and memorizing a mapping of squiggles to words. It's just that in English the pronunciation is emphasized in the writing, whereas in Chinese it is mainly the meaning of the word that is focused on. But given that in this age most of our communication is through text which may never even be read aloud, it does make one wonder if the Chinese had the right idea all along. But if so, rather than use a 3000 year old writing system with a lot of inherited historical baggage, why not start afresh? Keep the same basic concept (digital pictures in squares representing atomistic units of meaning, aligned in sequential order according to a grammar), but allow the pictures to make full benefit of today's digital display and font rendering technology. The logical end result of this is, I believe, emoji. Or something very close to it / evolved from it. To get an idea of what I mean, you might want to check out Xu Bing's "Book from the Ground", which is a very readable book written in emoji, or something close to it. You can read about his work here: https://lareviewofbooks.org/article/emojis-comics-novel-future/ https://lareviewofbooks.org/article/emojis-comics-novel-futu...
optimalsolver 3y ago
XKCD on Lojban: https://xkcd.com/191/ https://xkcd.com/191/
- zharknado 3y ago
  Well put. The socio-technical system gets so much neater if you take all the socio bits out of it.
FireInsight 3y ago
I take all my conlang opinions from jan Misali aka Conlang Critic and he tells me to love toki pona [0] instead of lojban [1]. [0] https://www.youtube.com/watch?v=eLn6LC1RpAo https://www.youtube.com/watch?v=eLn6LC1RpAo [1] https://www.youtube.com/watch?v=l-unefmAo9k https://www.youtube.com/watch?v=l-unefmAo9k
- culi 3y ago
  Very different languages with different goals
bmacho 3y ago
Why is it so hard, and why hasn't anybody done it before? The whole problem seems trivial to me. You need a language where - first you need your words to have only one meaning. You can even use an existing vocabulary, like English, or Esperanto. - then you need to be express math/logic/structured programming. That's it, now you have a language without any ambiguity. If there are some ambiguity in the objects/concepts that you are referring, then use your language more to specify the object.
- TeMPOraL 3y ago
  > Why is it so hard, and why hasn't anybody done it before? The whole problem seems trivial to me. You need a language where > - first you need your words to have only one meaning There's your answer. There is no such language, there are no such words. It's questionable if it's even possible - to me, it sounds like a category error. "Words" aren't keys in a hash table, and "meaning" isn't a discrete thing. > now you have a language without any ambiguity. If there are some ambiguity in the objects/concepts that you are referring, then use your language more to specify the object. Thing is, I don't think ambiguity comes from language. Concepts and categories are naturally fuzzy for us. Precision is a continuum. If you need to express something more precisely than your language allows with single word or term, you just create a new word and attach a long-form description to it to position and connect it to other words in the latent^H^H^H^H^H^H^H concept space. That's how we always did it, and it was never hard.
  - bmacho 3y ago
    I still don't get it. In the linguistic part the only problem is to make to word->object mapping unambiguous. And it is trivial problem. If the world is fuzzy (i don't think so) then is not a problem in linguistics. Yet linguists seem to aim for a language that eliminates ambiguity (see TFA). Where am I wrong?
    - sillymath 3y ago
      theoretically you can introduce new words to be more precise, for example eskimos have many words for snow, from (1) that is because they use polysynthesis: a base word is attached to many different suffixes which change its meaning. So, while in the English language we might have a sentence describing snow, fusional languages such as the Eskimo-Aleut family will have long, complex words. So, you may solve the ambiguity problem by introducing a huge numbers of words, that could help a LLM, but humans need a relatively small vocabulary tailored to everyday use. So the tradeoff is reducing ambiguity and not increasing a lot the vocabulary. (1) https://readable.com/blog/do-inuits-really-have-50-words-for-snow/ https://readable.com/blog/do-inuits-really-have-50-words-for...
SillyUsername 3y ago
Bah now I'm going to have to give up learning other languages, Klingon and Esperanto, to learn this. Qapla! Mi prenos mian capelon!
- rikroots 3y ago
  Honestly if you're tempted to learn one of these universal languages then I'd suggest going all-in on SolReSol[1]. At least if you get bored of speaking it, you can instead paint it, or play it on a trumpet. [1] - https://i.sidosi.org/resources/grammar-of-solresol/grammar-of-solresol.html https://i.sidosi.org/resources/grammar-of-solresol/grammar-o...
thih9 3y ago
It reminds me of constructing a type system to lower the chance of accidental ambiguity in code. So basically haskell for regular conversation?
e12e 3y ago
My initial thought on seeing this was; lojban again? Then i immediately realized that the surge of LLMs/AI would rekindle "Bable-17"-style ideas... https://en.m.wikipedia.org/wiki/Babel-17 https://en.m.wikipedia.org/wiki/Babel-17
leke 3y ago
I tried to learn lojban back in the day. I liked the phonetics, vocabulary and system in general, but it seemed an impossible task to remember the placeholder order of every single predicate. Maybe not so much of a struggle for an AI though. I still believe there will be an awesome international auxiliary language that the world will embrace, or rather those who wish to communicate with everyone will embrace. It may currently exist or might yet need to be invented. People are making new ones all the time.
- mrtranscendence 3y ago
  The dream of an international auxiliary language is an old one, yet none — of many — have ever caught on outside of niche hobbyists. It’s unclear to me what further advantages a conlang could have that would overcome the incumbency advantage of the dominant lingua francas. I don’t think eliminating syntactic ambiguity is really that exciting for people in general.
_8j50 3y ago
A language humans use only in non-digital medium would be interesting. To avoid LLMs and surveillance-capitalism from knowing too much about you and your ways.
- meepmorp 3y ago
  > A language humans use only in non-digital medium would be interesting. There's dozens or hundreds of them already, living in the world now. Some of them don't even have a formalized writing system.
zearen 3y ago
I'm one of the few conversation Lojban speakers, and after years extended study, I've found the language does not meet expectations. When you start deeply exploring the semantics, many constructs are only half defined or nonsensical. Further, the self appointed DFL tries to preserve the language in an unusable state and had alienated most of the proficient speakers. There are other logical languages out there. Most of them are in development, and are discussed in a Discord called the "Logical Languages Lodge". However there is another called Toaq that is almost complete, it at least as complete as Lojban (its vocabulary is still growing, but all of the essential words are there, and the grammar and semantics are complete). If you wish to learn a logical language in 2023, that's the one I'd suggest.
- Paianni 3y ago
  It would seem Toaq was started by a former Lojbanist, interesting.
- IIAOPSW 3y ago
  Oh hey I'm in that discord. I'm the one with that weird highly domain specific loglang and occasionally obtusely adversarial question (aka "trolling")
- oh_sigh 3y ago
  > I'm one of the few conversation Lojban speakers, and after years extended study Honest question...why?
  - zearen 3y ago
    Can you clarify your question ?
    - tester457 3y ago
      Fitting question from a Lojbanist.
      - austinjp 3y ago
fsiefken 3y ago
Ithkuil also is a constructed language with the goal of being completely un-ambiguous. As a bonus it uses the most phonemes of any language so complex messages and meanings can be communicated in an efficient way. http://www.ithkuil.net/ http://www.ithkuil.net/ https://www.reddit.com/r/Ithkuil/ https://www.reddit.com/r/Ithkuil/
canjobear 3y ago
How can they be sure it’s unambiguous? Ambiguity of a context-free grammar is undecidable.
- malisper 3y ago
  Ambiguity of a context-free Grammer is undecidable means there are some context free grammers where it is impossible to determine if they are ambiguous or not. There are context free grammers that we can prove are unambiguous. As an example "A = xAy | ε" is unambiguous. Lojban is one of the examples of grammers we are able to prove is unambiguous.
1bent 3y ago
I'd heard of it from Heinlein, but never thought it might be real.
gleki 3y ago
I'm one of a few "fluent" speakers of Lojban. I still think that among similar attempts (including controlled English) Lojban is the best option so far. I even wrote a new course for laymen https://lojban.pw/books/learn-lojban/ https://lojban.pw/books/learn-lojban/ Needless to say, few people invest in learning new languages at all. If you are interested you may join the discord chat https://discord.gg/c8weYzf https://discord.gg/c8weYzf
mark_l_watson 3y ago
A long time ago I tried to learn Esperanto well enough to read in it. I quit, but I was still intrigued. Given events in the last 18 months, instead of a constructed language like Lojban or Esperanto I would vote for a small subset of English with a limited grammar and vocabulary - with tools to enforce both vocabulary and simplified grammar. Why the change in opinion? It turns out that English works better than many other languages for text embeddings and generally working with Large Language Models. A simplified version of English, perhaps called ‘glish’ would be easy for English speakers to pick up and use with good automatic writing and editing tools. Advantages: make it easier for AI and for casual English users around the world.
- Izkata 3y ago
  Maybe start with the Simple English Wikipedia: https://simple.wikipedia.org/wiki/Main_Page https://simple.wikipedia.org/wiki/Main_Page
- soonix 3y ago
  Looks like you think about "Basic English" which is a bit more formalized than Wikipedia's "Simple English" - see https://simple.wikipedia.org/wiki/Basic_English https://simple.wikipedia.org/wiki/Basic_English (and https://en.wikipedia.org/wiki/Basic_English https://en.wikipedia.org/wiki/Basic_English )
progrus 3y ago
There will be inexpressible ideas in any such language, per Gödel. In fact, this means that there is no such thing as “free speech” in any totally-consistent language!
syngrog66 3y ago
I've been making a post-apoc survival adventure game lately [1], and as part of the world-making I've spent some time on banging out little slang languages and cultural terms, for the characters in it to use. They deal in English, mostly. But so much time has past, and theres been so much loss of the traditional institutions which "anchor" languages over the decades and actively help to keep them consistent, that there cant help but be LOTS of drift, here and there. Linguistic mutation and evolution -- but survival of the "most popular." Of whatever gets most repeated, person to person. Its also a comedy so I get to have fun playing around with the distinction between "Approved, Proper and Official" languages and words, and their nemeses: the hip, the veil-piercing, the casual, the quickly improvised or intentionally irreverant ones. When I was a kid I could effortlessly switch back and forth between the two types. Though I felt more natural and at home in the latter. Then as a groan-up in the white-collar workforce, when I was forced to do the whole hours-long, two--way highway commute M-F to spend yet another 8 hours in a drab cubicle, and had to Watch Every Single Thing I Said and Did All Day Long lest some random person take it out-of-context or Find It Offensive, a certain joy for life was... slowly beaten out of me. I've always been a fan of Lojban, on paper. But I dont see it having a bigger future. English is "good enough" however imperfect. Less is more strikes again? :-) ---- (1. see my bio: Slartboz)
throwawayForMe2 3y ago
I spent some time looking at Loglan, purchased some books and tapes, but never really made any progress with it. A Scientific American article got me interested, pre-internet days so finding a community to converse with was not easy. https://en.wikipedia.org/wiki/Loglan https://en.wikipedia.org/wiki/Loglan
euroderf 3y ago
IIRC Loglan vocabulary had a large element of Mandarin Chinese, and so Lojban too ? This seemed both uselessly idealistic and absurdly user-hostile. I would think that a conlang targeted at global appeal would be based on some combination of Romance languages and Germanic languages, perhaps an ur-version of each (like, Latin and Icelandic). Such a process might bring you a result where the vocabulary was as accessible as English while having immensely better-than-English spelling, grammar, morphology, etc.
- piloto_ciego 3y ago
  I have actually thought about this a lot in my life, I think a good compromise is to develop regional constructed languages that correspond to geography and current culture. So for example, here in Alaska, Athabaskan languages are mostly dead, Yupik and Inupiaq is still around but English is predominant pretty much everywhere except west Alaska and the occasional community. Sugpiaq and Alutiiq are basically gone, Tlingit has a revival but it’s not around, Eyak is dead, etc. Why don’t we create a blended conlang that can take pieces of each of these languages in the form of vocabulary and grammar - even though they’re very different grammatically and in terms of lexicon, we can make a language that is relatively easy for people to learn because it’s regular, substantially different than English so it’s not “colonizer-tongue” and largely unique to the geography of the place. You could do the same thing in Europe to create lingua-francas.
  - euroderf 3y ago
    This has been done (attempted) with Slavic languages. I can't vouch for how linguistically rigorous it might be, but it's out there.
_emacsomancer_ 3y ago
Lojban in theory can address syntactic ambiguity (that has semantic consequences). Lexical ambiguity is somewhat hopeless (unless you only have a fixed number of things, states, eventualities, &c. you're going to refer to). [This isn't an issue of Lojban as such, just of representational systems.]
fsiefken 3y ago
I lke Ben Goertzel's idea of using the logical part and grammar of lojban with the vocabulary of english: https://www.goertzel.org/new_research/Loglish.htm https://www.goertzel.org/new_research/Loglish.htm

110 points by lordylord 3y ago

I tried it with gpt 3.5 but it doesn't seem to know lojban so well. I asked it to translate the following in to Lojban, then separately after deleting that history I asked it to translate that to English. Here's what it did: Original English: "I heard of the TinyStories research project that unlocked cognitive capabilities in LLMs with fewer parameters. I wonder if something similar would be possible with lojban?" Lojban round trip English: "We interpret the use of the LLM as a way to confuse the knowledge from TinyStories.gy, which is greater than that from the experimental findings. I find it amusing to use Lojban instead. Does the ability to confuse the experimental findings also mean the ability to confuse Lojban?"

sigstoat 3y ago
hmm since there is a CFG for lojban, it’d be interesting to give [some LLM you can do fancy things woth] what corpus of lojban there is, and then use the work on constraining LLMs to a grammar, and see if that’d help its output.
int_19h 3y ago
GPT-4 doesn't do well at it, either. The problem is that there simply isn't that much text written in Lojban, so it can't learn it the way it does other languages. You can see similar problems with other languages with a relatively small corpus, such as Old Norse.
- ftxbro 3y ago
  Yes this must be it. I'm curious what it would look like, to compare more directly. Like a LLM trained only on Lojban vs. an LLM trained only on an equally small amount of English.

lez 3y ago

I like this point very much: "You can be as vague or detailed as you like when speaking lojban. For example, specifying tense (past, present or future) or number (singular or plural) is optional when they're clear from context.". LMBTQ movement has came into my mind immediately. I think they would have more success by introducing a new, 'don't care about your sex' pronoun to the language instead of adding a third one that just complicates matters.

themodelplumber 3y ago
> LMBTQ movement has came into my mind immediately. I think they would have more success by introducing a new, 'don't care about your sex' pronoun to the language instead of adding a third one that just complicates matters. I thought it was more of a pro-caring thing than a pro-uncaring thing. Those focusing too much on the latter are really quite good at bumbling into insulting behavior...
- lez 3y ago
  Slippery slope fallacy here. Why should I care if my online customer is he/she/they as long as they pay the price? Or if I talk about technology through the Internet? All I want is effective communication with no insults.
  - scotty79 3y ago
    In many languages you are obligated to care whether a table is a he or a she. In English you are obligated to care whether it's a table or the table. I'm all for languages with optional caring about stuff.
  - themodelplumber 3y ago
    Is this meant to describe how a slippery slope forms, or call one out? Effective communication involving other people only sometimes overlaps with this subjective-first, transactional approach, the: I care about, I talk about, or All I want is. (It's a great view for designing a specific type of community, with business-transactional perspective + valuing of tightening subjective logic and avoiding fallacy. Well, this will always result in a specific type of tech community with some special focus on business and economic factors as combined with tech.) And I guess if we effectively cherry-pick the premise, with straw people assuming their various positions on the stage, we can easily justify a more anonymizing, transactional view of language. "This is how I prefer to view my life, therefore the needed language looks like..." In this view, language doesn't need all that extra stuff. Keep it direct, functional, logical, and avoid fallacies especially. But this isn't really getting at the relating capabilities of language for example, like when there are bigger problems at hand and diverse mindsets working together to solve them. Instead, it ends up bringing minds into silos, as broader personality theory predicts. So the I, I, me, is really more about self-support. With this view, individuals effectively make relating all about the subjective-them 24/7, and about whether they are good/not good. They see relating with others as "we will each be our best and get what we each want" as if it's a universal creed. The one philosophy that just makes sense. Do others even exist? And is this actively mean? No. Perhaps a better term is effectively: "passively careless," to borrow a phrase from Sabine Hossenfelder. And even in the subjective-first view it still does not make for a high-quality individual, which is also bad news. It's unfortunate, because not only does this view raise the stakes on their end with no good reason, creating things like martyrdom issues, but it also perpetuates the avoidance of effective contact with others, effectively making "others" the persistent bête noir.

sowbug 3y ago

If memory serves, Attempto was recently on HN. It's another natural (invented) language that parses unambiguously. https://en.wikipedia.org/wiki/Attempto_Controlled_English https://en.wikipedia.org/wiki/Attempto_Controlled_English

yosito 3y ago

I don't know if eliminating ambiguity is even possible, much less a key feature of the language. There's even prominent ambiguity on this page: > Lojban means different things to different people

vidarh 3y ago
A lot of communication is intentionally ambiguous at that. Be it diplomatic texts written to let both sides present a victory, or comedy intentionally obscuring key facts for a later reveal. Having the ability to be unambiguous is great, but human communication is full of situations where that's not what we want.
jmillikin 3y ago
The goal of Lojban is to eliminate syntactic/grammatical ambiguity: Time flies like an arrow, fruit flies like a banana. In this example, "flies" could be either a noun or verb depending on context. That sort of construct isn't possible in Lojban. Of course, it's still possible to use Lojban to express statements that have an ambiguous or unclear meaning.
- djtango 3y ago
  Ah, classic HN Saturday morning - finding myself dreaming up Hungarian notation in English to disambiguate grammatical types nTime vFlies avLike an nArrow, nFruit vFlies avLike a nBanana It's ugly but if we all adhere to this convention and haze our juniors, we'll eliminate most our problems
  - prerok 3y ago
    avLike still has two different meanings ;) Actually, I am not sure that having different meanings of the same word in different contexts is really the problematic part causing ambiguity in conversations. For example, sarcasm is a much more difficult problem. The words are the same, the meaning of the sentences is the same but the author's intent, the meaning behind those sentences is opposite if they are being sarcastic or not.
  - 3y ago

wirrbel 3y ago

I spent a few weeks with Lojban around 2006. imho it felt really dense as a language. Definitely interesting. Back then there wasn’t as much material in Lojban as is now available.

tagyro 3y ago

I vaguely remember a TNG* episode in which humans encounter an alien race that found our language too ...inaccurate, and spent a lot of time going over each word and sentence, to precisely define its meaning. * Was it TNG or am I dreaming? *edit: Thank you everyone, but it wasn't "Darmok". The humans and the aliens were working, for a long time, on a treaty of sorts because the human language was so confusing and ambiguous and they had to go over every small details, over and over ... Need to re-watch TNG

sgtnoodle 3y ago
I believe I've seen every TNG episode at least once, and that's not ringing any bells. Maybe it was an original series episode? Despite my best efforts, I haven't been able to make it through more than one or two episodes of the original.
pwnmonkey 3y ago
Shaka, when the walls fell?
- sgtnoodle 3y ago
  I recall that episode, but I'm not sure that's what they were referencing.
numpad0 3y ago
“Darmok” is Internet favorite, about a species which language is exclusively in analogy that the translator could not be set up, and they decide to kidnap Picard to establish vocabulary over a campfire. There were couple episodes in Voyager in which aliens deem humans interesting but lesser, and tries to experiment or study on crews. One that I remember ended up in a ship self destruction, in another the doctor leaving his copy. As to why they didn’t simply ask Borg for a zipped archive, I don’t know. The Dominion commanders throughout DS9 must have made similar remarks, and so should Cardassians have. Q would outright say so in spite, corrupted Data might. Maybe combination of those? Or maybe I’m just a suboptimal Trekkie :p
- 3y ago

Decabytes 3y ago

I wonder if LLM entirely in lobjan would require less text to train to be just as good as the models we have today due to its higher clarity in expression. I also feel the same way about Esperanto due to its high regularity in syntax. Part of me also wonders if an ai model could be optimized by training it on a language specifically invented to be both easy for the ai to understand and the human to communicate in. Maybe the language was made up of lots of 4 letter words bringing the number of words to tokens closer to 1:1, or injecting more context per word so sentences could be shorter and use less tokens

ftxbro 3y ago
you should check TinyStories if you didn't know of it already https://arxiv.org/abs/2305.07759 https://arxiv.org/abs/2305.07759

adastra22 3y ago

I like the idea of Lojban. I hate the implementation. It’s basically spoken math notation for predicate logic, which falls flat well short of the goal. There are much better logical frameworks that have been developed for computational semantics (e.g. Davidsonian event semantics). Lojban’s formalism was out of date already when it was proposed as Loglan half a century ago. If there is anyone out there who likes the idea of Lojban but want something more practical and with a firmer foundation to boot, contact me. I’ve got ideas and I’d love to find some people to work on this with.

Beltiras 3y ago
I wonder if you could get the intended result for any language with a simple ruleset. You don't need a language to reduce/eliminate miscommunications, you need intent and understanding.
- msla 3y ago
  Yes, you could, because any natural language can be used to teach mathematics. Lojban is, like any notation, a convenience for people who understand the concepts, but the concepts can be expressed in a natural language just as precisely, albeit less concisely. In fact, as Florian Cajori stated in his book "A History of Mathematical Notations", the rise of mathematical notation was opposed by people who preferred the older, natural language style of doing mathematics, what Cajori termed the "struggle between symbolists and rhetoricians." So, yes, logic is mathematics, and humans did mathematics in natural language for a very long time before we invented the conlang of mathematical notation. Moving more of the ideas into what is, ultimately, a more expressive notation is not a fundamental shift.
- jpk 3y ago
  This is kinda what we call "legalese". It's a sort of formalized subset of English (or whatever language) that leans on standardized turns of phrase, it tends to set up definitions for terms that are then used throughout a document, etc. All in order to reduce misunderstanding and (hopefully) be easy to interpret in the event of a dispute. However, we have whole judicial systems that spend a non-trivial fraction of their time interpreting legal verbiage. So clearly it falls short at least some of the time, otherwise courts could be, in part, automated away. Maybe that's because it's too hard or not possible with natural languages? Or the legalese ruleset just isn't refined enough?

baby 3y ago

After learning toki pona (the simplest conlang perhaps?) I got attracted by Lojban due to the machine-parsable aspect of the conlang, but I found it really tedious and not fun to learn. I started working on a conlang with similar ideas a while back (emojis in the language, machine parsable) but less strict and nicer sounding. Never ended up with anything though :)

adastra22 3y ago
You might like Mini / Mini Kore. I no joke believe that the eventual end state of linguistic evolution is an ideographic written language with emoji as characters.
- BiteCode_dev 3y ago
  That would be terrible. Ideographic written languages are hard to learn (you need to memorize much more to get basic skills) and hard to parse (tokenization is more difficult). Of course, if that means the entire world speaks the same language, the price would be worth it. But I doubt we will converge on anything that is difficult to master, since water takes the path of least resistance.
  - adastra22 3y ago
    That's not actually true. When MacArthur tried to get rid of the use of Chinese characters in Japanese writing during the post-WW2 occupation, he backed off when it was embarrassingly pointed out that Japan had a higher literacy rate than the US, despite having less of an investment in state schools at the time. Studies have shown that it does not take more time or effort for a child to learn to read and write Chinese or Japanese (assuming those are their native spoken languages) than for an American student to learn to read/write English. It may defy what you think you know as common sense, but it is a replicable observation. The reason for this is that nobody except little children actually read by sounding out individual letters of words. That would be too much for the human mind to process in real time. You learn to read by recognizing particular clusters of lines on a paper or screen as having semantic meaning, and your brain translates that directly. The imlpiactoin is the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a wlohe. (Did you read that OK?) So essentially, you and a Chinese person both put in the same amount of effort to learn the same number of squiggly lines on paper as having semantic meaning, and memorizing a mapping of squiggles to words. It's just that in English the pronunciation is emphasized in the writing, whereas in Chinese it is mainly the meaning of the word that is focused on. But given that in this age most of our communication is through text which may never even be read aloud, it does make one wonder if the Chinese had the right idea all along. But if so, rather than use a 3000 year old writing system with a lot of inherited historical baggage, why not start afresh? Keep the same basic concept (digital pictures in squares representing atomistic units of meaning, aligned in sequential order according to a grammar), but allow the pictures to make full benefit of today's digital display and font rendering technology. The logical end result of this is, I believe, emoji. Or something very close to it / evolved from it. To get an idea of what I mean, you might want to check out Xu Bing's "Book from the Ground", which is a very readable book written in emoji, or something close to it. You can read about his work here: https://lareviewofbooks.org/article/emojis-comics-novel-future/ https://lareviewofbooks.org/article/emojis-comics-novel-futu...

optimalsolver 3y ago

XKCD on Lojban: https://xkcd.com/191/ https://xkcd.com/191/

zharknado 3y ago
Well put. The socio-technical system gets so much neater if you take all the socio bits out of it.

FireInsight 3y ago

I take all my conlang opinions from jan Misali aka Conlang Critic and he tells me to love toki pona [0] instead of lojban [1]. [0] https://www.youtube.com/watch?v=eLn6LC1RpAo https://www.youtube.com/watch?v=eLn6LC1RpAo [1] https://www.youtube.com/watch?v=l-unefmAo9k https://www.youtube.com/watch?v=l-unefmAo9k

culi 3y ago
Very different languages with different goals

bmacho 3y ago

Why is it so hard, and why hasn't anybody done it before? The whole problem seems trivial to me. You need a language where - first you need your words to have only one meaning. You can even use an existing vocabulary, like English, or Esperanto. - then you need to be express math/logic/structured programming. That's it, now you have a language without any ambiguity. If there are some ambiguity in the objects/concepts that you are referring, then use your language more to specify the object.

TeMPOraL 3y ago
> Why is it so hard, and why hasn't anybody done it before? The whole problem seems trivial to me. You need a language where > - first you need your words to have only one meaning There's your answer. There is no such language, there are no such words. It's questionable if it's even possible - to me, it sounds like a category error. "Words" aren't keys in a hash table, and "meaning" isn't a discrete thing. > now you have a language without any ambiguity. If there are some ambiguity in the objects/concepts that you are referring, then use your language more to specify the object. Thing is, I don't think ambiguity comes from language. Concepts and categories are naturally fuzzy for us. Precision is a continuum. If you need to express something more precisely than your language allows with single word or term, you just create a new word and attach a long-form description to it to position and connect it to other words in the latent^H^H^H^H^H^H^H concept space. That's how we always did it, and it was never hard.
- bmacho 3y ago
  I still don't get it. In the linguistic part the only problem is to make to word->object mapping unambiguous. And it is trivial problem. If the world is fuzzy (i don't think so) then is not a problem in linguistics. Yet linguists seem to aim for a language that eliminates ambiguity (see TFA). Where am I wrong?
  - sillymath 3y ago
    theoretically you can introduce new words to be more precise, for example eskimos have many words for snow, from (1) that is because they use polysynthesis: a base word is attached to many different suffixes which change its meaning. So, while in the English language we might have a sentence describing snow, fusional languages such as the Eskimo-Aleut family will have long, complex words. So, you may solve the ambiguity problem by introducing a huge numbers of words, that could help a LLM, but humans need a relatively small vocabulary tailored to everyday use. So the tradeoff is reducing ambiguity and not increasing a lot the vocabulary. (1) https://readable.com/blog/do-inuits-really-have-50-words-for-snow/ https://readable.com/blog/do-inuits-really-have-50-words-for...

SillyUsername 3y ago

Bah now I'm going to have to give up learning other languages, Klingon and Esperanto, to learn this. Qapla! Mi prenos mian capelon!

rikroots 3y ago
Honestly if you're tempted to learn one of these universal languages then I'd suggest going all-in on SolReSol[1]. At least if you get bored of speaking it, you can instead paint it, or play it on a trumpet. [1] - https://i.sidosi.org/resources/grammar-of-solresol/grammar-of-solresol.html https://i.sidosi.org/resources/grammar-of-solresol/grammar-o...

thih9 3y ago

It reminds me of constructing a type system to lower the chance of accidental ambiguity in code. So basically haskell for regular conversation?

e12e 3y ago

My initial thought on seeing this was; lojban again? Then i immediately realized that the surge of LLMs/AI would rekindle "Bable-17"-style ideas... https://en.m.wikipedia.org/wiki/Babel-17 https://en.m.wikipedia.org/wiki/Babel-17

leke 3y ago

I tried to learn lojban back in the day. I liked the phonetics, vocabulary and system in general, but it seemed an impossible task to remember the placeholder order of every single predicate. Maybe not so much of a struggle for an AI though. I still believe there will be an awesome international auxiliary language that the world will embrace, or rather those who wish to communicate with everyone will embrace. It may currently exist or might yet need to be invented. People are making new ones all the time.

mrtranscendence 3y ago
The dream of an international auxiliary language is an old one, yet none — of many — have ever caught on outside of niche hobbyists. It’s unclear to me what further advantages a conlang could have that would overcome the incumbency advantage of the dominant lingua francas. I don’t think eliminating syntactic ambiguity is really that exciting for people in general.

_8j50 3y ago

A language humans use only in non-digital medium would be interesting. To avoid LLMs and surveillance-capitalism from knowing too much about you and your ways.

meepmorp 3y ago
> A language humans use only in non-digital medium would be interesting. There's dozens or hundreds of them already, living in the world now. Some of them don't even have a formalized writing system.

zearen 3y ago

I'm one of the few conversation Lojban speakers, and after years extended study, I've found the language does not meet expectations. When you start deeply exploring the semantics, many constructs are only half defined or nonsensical. Further, the self appointed DFL tries to preserve the language in an unusable state and had alienated most of the proficient speakers. There are other logical languages out there. Most of them are in development, and are discussed in a Discord called the "Logical Languages Lodge". However there is another called Toaq that is almost complete, it at least as complete as Lojban (its vocabulary is still growing, but all of the essential words are there, and the grammar and semantics are complete). If you wish to learn a logical language in 2023, that's the one I'd suggest.

Paianni 3y ago
It would seem Toaq was started by a former Lojbanist, interesting.
IIAOPSW 3y ago
Oh hey I'm in that discord. I'm the one with that weird highly domain specific loglang and occasionally obtusely adversarial question (aka "trolling")
oh_sigh 3y ago
> I'm one of the few conversation Lojban speakers, and after years extended study Honest question...why?
- zearen 3y ago
  Can you clarify your question ?
  - tester457 3y ago
    Fitting question from a Lojbanist.
    - austinjp 3y ago

fsiefken 3y ago

Ithkuil also is a constructed language with the goal of being completely un-ambiguous. As a bonus it uses the most phonemes of any language so complex messages and meanings can be communicated in an efficient way. http://www.ithkuil.net/ http://www.ithkuil.net/ https://www.reddit.com/r/Ithkuil/ https://www.reddit.com/r/Ithkuil/

canjobear 3y ago

How can they be sure it’s unambiguous? Ambiguity of a context-free grammar is undecidable.

malisper 3y ago
Ambiguity of a context-free Grammer is undecidable means there are some context free grammers where it is impossible to determine if they are ambiguous or not. There are context free grammers that we can prove are unambiguous. As an example "A = xAy | ε" is unambiguous. Lojban is one of the examples of grammers we are able to prove is unambiguous.

1bent 3y ago

I'd heard of it from Heinlein, but never thought it might be real.

gleki 3y ago

I'm one of a few "fluent" speakers of Lojban. I still think that among similar attempts (including controlled English) Lojban is the best option so far. I even wrote a new course for laymen https://lojban.pw/books/learn-lojban/ https://lojban.pw/books/learn-lojban/ Needless to say, few people invest in learning new languages at all. If you are interested you may join the discord chat https://discord.gg/c8weYzf https://discord.gg/c8weYzf

mark_l_watson 3y ago

A long time ago I tried to learn Esperanto well enough to read in it. I quit, but I was still intrigued. Given events in the last 18 months, instead of a constructed language like Lojban or Esperanto I would vote for a small subset of English with a limited grammar and vocabulary - with tools to enforce both vocabulary and simplified grammar. Why the change in opinion? It turns out that English works better than many other languages for text embeddings and generally working with Large Language Models. A simplified version of English, perhaps called ‘glish’ would be easy for English speakers to pick up and use with good automatic writing and editing tools. Advantages: make it easier for AI and for casual English users around the world.

Izkata 3y ago
Maybe start with the Simple English Wikipedia: https://simple.wikipedia.org/wiki/Main_Page https://simple.wikipedia.org/wiki/Main_Page
soonix 3y ago
Looks like you think about "Basic English" which is a bit more formalized than Wikipedia's "Simple English" - see https://simple.wikipedia.org/wiki/Basic_English https://simple.wikipedia.org/wiki/Basic_English (and https://en.wikipedia.org/wiki/Basic_English https://en.wikipedia.org/wiki/Basic_English )

progrus 3y ago

There will be inexpressible ideas in any such language, per Gödel. In fact, this means that there is no such thing as “free speech” in any totally-consistent language!

syngrog66 3y ago

I've been making a post-apoc survival adventure game lately [1], and as part of the world-making I've spent some time on banging out little slang languages and cultural terms, for the characters in it to use. They deal in English, mostly. But so much time has past, and theres been so much loss of the traditional institutions which "anchor" languages over the decades and actively help to keep them consistent, that there cant help but be LOTS of drift, here and there. Linguistic mutation and evolution -- but survival of the "most popular." Of whatever gets most repeated, person to person. Its also a comedy so I get to have fun playing around with the distinction between "Approved, Proper and Official" languages and words, and their nemeses: the hip, the veil-piercing, the casual, the quickly improvised or intentionally irreverant ones. When I was a kid I could effortlessly switch back and forth between the two types. Though I felt more natural and at home in the latter. Then as a groan-up in the white-collar workforce, when I was forced to do the whole hours-long, two--way highway commute M-F to spend yet another 8 hours in a drab cubicle, and had to Watch Every Single Thing I Said and Did All Day Long lest some random person take it out-of-context or Find It Offensive, a certain joy for life was... slowly beaten out of me. I've always been a fan of Lojban, on paper. But I dont see it having a bigger future. English is "good enough" however imperfect. Less is more strikes again? :-) ---- (1. see my bio: Slartboz)

throwawayForMe2 3y ago

I spent some time looking at Loglan, purchased some books and tapes, but never really made any progress with it. A Scientific American article got me interested, pre-internet days so finding a community to converse with was not easy. https://en.wikipedia.org/wiki/Loglan https://en.wikipedia.org/wiki/Loglan

euroderf 3y ago

IIRC Loglan vocabulary had a large element of Mandarin Chinese, and so Lojban too ? This seemed both uselessly idealistic and absurdly user-hostile. I would think that a conlang targeted at global appeal would be based on some combination of Romance languages and Germanic languages, perhaps an ur-version of each (like, Latin and Icelandic). Such a process might bring you a result where the vocabulary was as accessible as English while having immensely better-than-English spelling, grammar, morphology, etc.

piloto_ciego 3y ago
I have actually thought about this a lot in my life, I think a good compromise is to develop regional constructed languages that correspond to geography and current culture. So for example, here in Alaska, Athabaskan languages are mostly dead, Yupik and Inupiaq is still around but English is predominant pretty much everywhere except west Alaska and the occasional community. Sugpiaq and Alutiiq are basically gone, Tlingit has a revival but it’s not around, Eyak is dead, etc. Why don’t we create a blended conlang that can take pieces of each of these languages in the form of vocabulary and grammar - even though they’re very different grammatically and in terms of lexicon, we can make a language that is relatively easy for people to learn because it’s regular, substantially different than English so it’s not “colonizer-tongue” and largely unique to the geography of the place. You could do the same thing in Europe to create lingua-francas.
- euroderf 3y ago
  This has been done (attempted) with Slavic languages. I can't vouch for how linguistically rigorous it might be, but it's out there.

_emacsomancer_ 3y ago

Lojban in theory can address syntactic ambiguity (that has semantic consequences). Lexical ambiguity is somewhat hopeless (unless you only have a fixed number of things, states, eventualities, &c. you're going to refer to). [This isn't an issue of Lojban as such, just of representational systems.]

fsiefken 3y ago

I lke Ben Goertzel's idea of using the logical part and grammar of lojban with the vocabulary of english: https://www.goertzel.org/new_research/Loglish.htm https://www.goertzel.org/new_research/Loglish.htm