7 ms·
Can one of the LLM systems be trained to translate to or from this? Is that useful?
by Animats 3y ago
Can one of the LLM systems be trained to translate to or from this? Is that useful?
- int_19h 3y agoYou can see what happens when you ask an LLM (in this case I believe it's GPT-3.5) to use Toki Pona as the primary language while also providing English translations: https://chirper.ai/tokipona?tab=activity https://chirper.ai/tokipona?tab=activity Deciding on usefulness of this experiment is left as an exercise for the reader.
- andsoitis 3y agoI just tried say "hello what is your name" in toki pona in ChatGPT and the response was: In toki pona, "hello what is your name" can be expressed as "toki! nimi sina li seme?"
- Aachen 3y agoGPT3(?) is not that good, making quite basic mistakes, as in, I can pick them out and I'm an absolute noob at toki pona. GPT4 link that someone posted in a sibling comment gives 404 for me so I can't judge that, please include versions when saying that some LLM can do it :)
- cosmojg 3y agoHere's an excerpt from the linked ChatGPT conversation: > [User]: Please translate this Toki Pona into English: > [User]: o! mi sona e toki pona. mi lukin sona e toki mute... taso mi toki ike e toki ale. toki mute li jo e nimi mute. nimi mute li pona ala tawa mi. toki pona li pona e mi tan ni. toki pona li jo nimi lili. nimi lili li pona e me > [GPT-4]: Here's a translation of the provided Toki Pona text into English: > [GPT-4]: "Oh! I know Toki Pona. I have studied many languages... but I speak all languages poorly. Many languages have many words. Many words are not good for me. Toki Pona is good for me because of this. Toki Pona has few words. Few words are good for me." > [GPT-4]: Note: Toki Pona is a minimalistic constructed language with a small vocabulary. The translation may not capture all the nuances, but this is a close representation of the message in English.
- bfung 3y agoLol nice. Reminds me of that one HN post a few years ago… https://news.ycombinator.com/item?id=31840331 https://news.ycombinator.com/item?id=31840331 > grug brain developer not so smart, but grug brain developer program many long year and learn some things although mostly still confused
- coder543 3y agoYou assume that existing ones can’t? Using the example provided in the main link, ChatGPT-4 had no trouble at all: https://chat.openai.com/share/3b8faf2e-4eca-4d52-97fb-ac89a4923955 https://chat.openai.com/share/3b8faf2e-4eca-4d52-97fb-ac89a4...
- devit 3y agoSeems ineffective with abstract concepts. "In mathematics, a real number is a number that can be used to measure a continuous one-dimensional quantity such as a distance, duration or temperature. Here, continuous means that pairs of values can have arbitrarily small differences. Every real number can be almost uniquely represented by an infinite decimal expansion." -> "toki pi nanpa, nanpa lon li nanpa li ken lukin e suli tawa wan. lon ni la, nanpa tu li ken ala li poka li ken lili. nanpa lon ali li ken lukin kepeken nimi nanpa mute tawa pini ala" -> "Talking about numbers, when a number is present, it can be seen to be important or significant to an individual." "In this context, the number two might not be significant or might be seen as small or lesser." "All numbers can be seen using many number names endlessly or without an end." ChatGPT 4, obviously in separate chats: as you can see a complete failure to convey and roundtrip the original message.
- coder543 3y ago> Seems ineffective with abstract concepts I’ve only heard of Toki Pona for about 20 minutes, but everything I’m reading says that’s just a normal property of the language, not a limitation of ChatGPT. As a human, which of the 120 Toki Pona words would you confidently and correctly translate as “continuous” or “real” or any of these other linguistically uncommon concepts? Based on my own chat with ChatGPT, I feel almost certain that it warned you about this exact limitation of the language, but you ignored that?
- devit 3y ago"continuous" (in this actually incorrect usage, since the correct term is "dense") -> said of an algebraic structure such that for every x != y there exists z such that x < z and z < y -> characteristic ("nimi"?) of a set ("poki") where for ("tawa?") all ("ale") thing ("ijo") different ("ante" + "lon"?) than other ("ante" + "nimi"?) thing ("ito") there exists ("lon") a new ("sin") thing ("ijo") such that the first ("wan" + "nimi" + "ito"?) thing is less ("lili") than the new thing ("sin" + "ito"?) and the new thing ("sin" + "ito"?) is less ("lili") that the second thing ("tu" + "nimi" + "ito"?) Would need some way to join into a grammatical structure and probably a preamble that establishes that the context is mathematics.
- weinzierl 3y agoI always wanted to train an LLM from scratch. There is an interesting project that does just that with Esperanto and my immediate thought was I'll do it with Toki Pona. Unfortunately, I had vastly underestimated the size of the existing Esperanto corpus and overestimated size of the Toki Pona corpus. So, I gave up on that idea because I believe that there is not enough training data, but I'd love to be proven wrong.
- bravura 3y agoHere was GPT4 giving me a full literal and then figurative translation of the sample text: https://chat.openai.com/share/50fd27af-7515-4f89-b0e1-22850130306c https://chat.openai.com/share/50fd27af-7515-4f89-b0e1-228501...
- neontomo 3y agoYou can think of most common languages as "lossless" (at least in comparison), whereas Toki Pona is "lossy". Once you translate English to Toki, it loses a lot of granularity. For example... Translating English to Toki Pona: Talk -> Toki Translating Toki Pona to English: Toki -> To speak, Communicate, Use language, Think, Say "hello" Which means, it's now impossible to translate it back to English and know which one is referred to, without context. The language forces you to be less specific in your meaning, but also to focus on core human emotions instead of unnecessary complexity. Having a translator perfectly convert English to Toki Tona is gonna be a lot easier than the reverse, anyway, unless you have a lot of context to piece it together. Note: I tried learning this language many years ago and this is my personal opinion. It may not be shared by everyone.
- Aachen 3y ago> You can think of most common languages as "lossless" (at least in comparison), If only! For most words in most sentences, you can find direct translations, but between most language pairs, I'm pretty sure any given text (just one page or so, let alone books) will have at least one point where you have to interpret the text to translate it: choose what meaning the author intended or what saying/expression conveys the same meaning the closest without going into descriptive mode. I suppose that might be what you meant by "in comparison", but it seems misleading, at least to anyone not fluent/proficient enough in another language to know that lossless translations isn't how languages work But, yes, toki pona lifts this to a whole 'nother level. Does anyone happen to know how Esperanto compares? As far as I know, it should be quite similar to a "normal" (non-con)language in this regard (just simpler to learn, with bonus points if you speak a romance language afaik)
- shadowofneptune 3y agoEsperanto was intended as a sort of diplomatic language. It's got flaws, definitely. The sounds and spelling are very much from the creator's native Polish, a lot of important terms are rather obscure («Usono,» from "Usonia" is the word for the United States). That said, it is in the end relatively easy to learn, and it is easy to express the ideas of diplomacy, science, and civil society. China and Japan used to have a lot of Esperantists before WWII, for that reason. https://www.loc.gov/item/2021670575/ https://www.loc.gov/item/2021670575/ > After World War I, the League of Nations considered adopting Esperanto as a working language and recommending that it be taught in schools, but proposals along these lines were vetoed by France. It may be Eurocentric, but it's hell of a lot easier for diplomats to learn than English or French!