Hacker News Reader

Comment by dchuk | Hacker News Reader

I’ve always been fascinated by the idea of speaking multiple languages, but the basic fact is that technology is clearly going to make the idea of speaking different languages completely irrelevant in the next few years. We’ll all have real time Rosetta stones in our pockets, at which point why not just everyone speak the same language going forward?

I get that there’s something romantic about the idea of regional dialects, but practically speaking in a global world with advanced tech, it’s simply only a point of friction.

View on HN

I agree with the basic premise that having multiple languages is an annoying point of friction, and it would be great if it could be removed (I have lived in five countries and dealing with government, insurance, doctors, etc. is always a big roadblock).

However, if you have any sort of international exposure beyond holidays, it is abundantly clear that you cannot extricate the language that people speak with the culture and history of a region. Having everyone speak the same language would mean throwing away much of what makes countries unique, because the old language would be forgotten over time. And people won't like that. Politicians won't like that either, as it becomes much harder to push the narrative that "we" are superior and "they" are barbarians, when everything is so uniform.

> you cannot extricate the language that people speak with the culture and history of a region.

This, to me, is the reason for the diversity of language (and the reason for learning foreign languages, and for conserving your own language).

Whether or not a speaker is familiar with his own literary and historical traditions, he relies on them in the idioms he uses, the jokes he makes, and in his understanding of the actions and utterances of others. If everyone in France spoke only English, they'd struggle to understand themselves.

This is the main reason I consider myself a prescriptivist, in lexicographic terms. If you strip your language of what some people consider archaisms, you lose the ability to read the old plays and poetry, the soil from which your culture germinated.

But a universal translator makes it irrelevant if you learn a common language, it’ll allow people to speak their native language without material friction. IMO that’ll be the best thing for preserving a rich linguistic body of languages and dialects.

> a universal translator makes it irrelevant if you learn a common language

We are nowhere close to a proper universal translator. Simply transcribing words from one language to another is lossy.

> technology is clearly going to make the idea of speaking different languages completely irrelevant

Only in certain contexts. Business, transactions, news and, to some degree, science, can all be done with machine translation already. You can literally walk into a store right now somewhere that you don't speak the language and have a full conversation with the staff using Google Translate. But once you move beyond straightforward transactional interactions, start caring about expressing feelings and experiences, or understanding diverse cultures, no translation app is going to help you. I'm not sure how relevant this is in West Virginia, because it's such a culturally homogeneous and isolated place. But in multicultural places and places near cultural borders, it's absolutely essential for harmonious relationships between cultures. What happens when people from West Virginia move to Texas or Florida and start to encounter Spanish in their daily lives? Until they learn Spanish, they're going to build cultural walls that prevent them from seeing and understanding the Spanish speaking culture around them. And that will have profound implications on society at scale.

Personally, I think language learning is important if not for practical reasons, for reasons of personal development. Everyone should become fluent in at least one language other than their native one simply for the ability to see and express different cultural perspectives.

> start caring about expressing feelings and experiences, or understanding diverse cultures

I'd add that speaking in a different language, once you get good enough at it, is just plain fun. Google translate isn't fun, and I don't think it ever will be.

Protip: these magic Rosetta stones you hold in such high regard are trained on corpora of texts translated by professional translators with degrees in translation.

If you think these corpora we have today are enough to power the ML forever and ever until the end of time, then you should have another and better thought instead.

That's not correct.

Consider a human who can speak both spanish and english. They can translate intrinsically. Why? Because language is an interface to their knowledge. You do not require a special intermediary language that only trained professionals can teach.

Language -> knowledge (embeddings) -> Language is fine

> They can translate intrinsically.

Unfortunately no. If you speak two languages natively you know how hard translating can be unless you have training.

As the ML improves the algorithms themselves can be used to generate more training data.

Believe it or not, the First Law of Thermodynamics seems to apply here!

No, they cannot. (At least not in any meaningful sense.) This will not create new information you can draw upon to improve translations models and keep them up to date.

The model can generate translations and humans can verify that they're accurate enough to be used for training, making corrections if needed. Verifying work is much faster than creating it.

a) You will definitely need trained and credentialled translators to do the verifying.

b) The claim that verifying is faster than creating is unsupported. (E.g. reviewing programmer code is tougher than writing from scratch.)

Try reading a book vs writing one

Language eradication was and still is one of the major methods of genocide. Of the remaining 8000 or so human languages still extent, one or another dies every few weeks... and those are just the ones we know about.

There's nothing romantic about this. The proper genre is horror or maybe post-apocalyptic dystopia.

There's something of a control freak thing going on when one who would never want to speak to the people who use these minority languages still insists that they speak the same language as he does.

I think that’s not at all the point being made. In fact it’s quite the opposite. With a real time bi directional universal language translator pervasively available and ergonomically adapted it would be relatively irrelevant which language you spoke. This would in fact be a boon to minority languages or languages outside the language of business and science (today English). It would allow minority languages to flourish without extrinsic pressure to assimilate into more influential languages because its relatively toll free to be monolingual. Most of the pressure outside of planned cultural eradication is economic, and making the economic cost of speaking your native language close to zero is the best way to preserve native languages. Dreaming of mass literacy of minority languages won’t work - its impractical to expect everyone to learn hundreds of languages to meet people at their native tongue, especially in a multicultural society like modern western societies. That’s not a horror; that’s human nature. But technology has the chance to make it irrelevant, allowing people to communicate without assimilation.

That said people will still learn languages to do comparative analysis and out of personal drive. That’s great too.

> With a real time bi directional universal language translator

Those aren’t being developed for most languages of the world, only the fairly large languages where corporations see financial interest. Moreover, due to poverty or censorship that has hindered publication of books and internet content in most languages, most languages do not have a sizeable electronic corpus with which to train MT.

The OP is right, anyone active in this field knows that your MT dream would actually result in massive language loss, as bilingual users of MT technology feel forced to switch from their mother tongue to the regionally powerful language in order to access that MT technology.

Actually GPT-4 is at least marginally capable at almost all languages, and I expect LLMs being as nascent as they are, will become more proficient at more and more languages. The thing is that training a LLM is fundamentally simpler to train than prior NLP translation models. For example, I can carry on a reasonably good conversation with ChatGPT in the ancient dead language of Pali and have it translated into cuneiform, then Chinese, then Esperanto, and to English in a new session and the translation is almost flawless from the original. (I have in fact done this)

If it weren’t for this I would have agreed with you last year. But I see fairly clearly the way to preserve native tongues is to take a base LLM like Llama 2 and fine tune it with your native language. As people are invested in their native language this doesn’t seem unreasonable. As things develop and sharing LLM LoRa becomes easier, I think we will find that universal translator for all spoken languages forthcoming.

But I challenge the notion that not having a translation device assures native languages survive. The same pressures exist. You simply can’t tell the regionally powerful speakers to learn the minority languages and expect them to actually do it. And you can’t expect it to not be a stigma. Humans just don’t work that way, and never have. “When in Rome” applied in Roman times, in our time, and in all times. The only way to eliminate that is to make it not an issue.

> For example, I can carry on a reasonably good conversation with ChatGPT in the ancient dead language of Pali and have it translated into cuneiform

ChatGPT’s ability to translate into Pali or cuneiform is limited by the size of the corpus. As I said, most languages of the world do not have a sizeable electronic corpus, and what has appeared in writing is only a limited portion of those languages. ChatGPT cannot magically guess words or idioms that it was never trained on.

This is well known to anyone working in corpus linguistics. Do you have any formal background in the field?

> As people are invested in their native language this doesn’t seem unreasonable.

Outside a few relatively privileged languages, people are much less invested in their native language than you assume. Due to the pressures of poverty, political oppression, and social stigma, it can be difficult for linguists to even find speakers willing to answer some questions about their language, let alone train MT.

Then it’s up to those that care to preserve languages they care about. I for one am not an advocate that there be no one preserving languages or dedicating their lives to preserving languages. That’s a wonderful thing to do, and we have a tool that they can use to encode those languages for all time in a way that is functionally accessible to all for all time.

Reading the thread I don’t notice any proposal of an alternative. Without that I will still hold onto the idea that we can improve things with the miracles we create in our technologies. While corporations might not feel a profit potential here, surely the open source world has shown we don’t need corporations to do amazing things.

different languages come with entirely different ways of thinking and reasoning about the world - it would be a shame to throw that away

As someone who learnt a foreign to fluency as an adult, I am skeptical of the stronger interpretations of the Sapir Whorf hypothesis. I believe a second language is primarily a source of cultural richness (mutual understanding, interesting vocabulary, intriguing turns of phrase, and so forth) - but not a prerequisite for accessing any novel forms of thinking or reasoning, i.e. I can not recall ever feeling like my thought processes or understanding of the world were fundamentally different to my monolingual colleagues.

I think it depends on how close the two languages are, but also I think that the effect is much more powerful when you are a child and your brain is more malleable.

For example, compared to Western languages, Chinese has very little grammar and words that are more ambiguous and with several meanings. This translates to a way of thinking that is more fuzzy, holistic, and high level, relying more on connections (intuitively, when you don't have grammatical constructs telling you the connection between things, you are forced to seek/guess it by yourself)

Maybe that explains why it's almost impossible to get a straight "yes" or "no" to a direct question to Chinese native speakers.

Exactly. To go deeper in this topic, you may find this interesting: "The thinking styles of Chinese people", http://dx.doi.org/10.1093/oxfordhb/9780199541850.013.0012

> I can not recall ever feeling like my thought processes or understanding of the world were fundamentally different to my monolingual colleagues.

As a native German speaker I, for example, for consider thinking in German (regarding the case system that is deeply ingrained into this language) very much like thinking about types in a type system in a programming language; you have to pass an object [pun intended] of the correct type (genus) to a function/predicate so that the code compiles/the sentence is grammatically correct.

This kind of thinking does not work in English because of its different grammatical structure.

Another example: it is very common in German to analyze words part by part to find "sometimes hidden meanings"). Just one trivial example: "entschweben" [to waft away]: if you analyze the construction of this verb, you see that this is the verb "schweben" [to hover] that is modified with a prefix "ent-" that changes its meaning. This prefix thus works like some higher-order function that modifies the functionality of some existing function (in programming) - or in the given case of language, the meaning of a verb. So, one can really say that how I, as a native German speaker, think about my native language daily often corresponds to quite non-trivial programming concepts.

When I attempted a similar analysis of English words to English native speakers, they soon cut me off and explained to me that "this kind of analysis is not how English works or how English native speakers think about their language".

The compsci analogy that comes to mind for me is mesh generation:

https://en.m.wikipedia.org/wiki/Mesh_generation

There are many triangulations, but they approximate the same underlying model.

Think of it as an art form, sure language learning won't be financially necessary going forward but that doesn't mean you won't get personal benefits to learn languages. People never stopped learning to play an instrument even if pretty much anything can be generated by a computer nowadays.

And unlike what you think, this good translation tech will not make languages converge but diverge even more since there's no financial need to learn other ones anymore. Language learning is hard and less people will likely engage in the activity if it only becomes a hobby. The most affected languages negatively will likely be the most learned ones at the moment.

Reply on news.ycombinator.com