Time to stop studying languages?

[lavengro]

It is a solid lock that well within the lifetime of most of us, technology will come to pass that will allow one to plug headphones into an iPhone or whatever magic box is in vogue at the time and have it translate adequately in real time a foreign language speaker`s jibberish.

I am so certain of this that I would bet up to a gazillion dollars in support of this occurring, if I could find a bookie who would take the opposite end of this action (my own bookie won`t, and he has already bought his Lamborghini its own Maserati off of many of my foolish bets in the past - that shows how certain he also is that this will occur, and not terribly far off either).

[Iversen]

I have written a whole thread about things that could help Google translate to bring down the number of totally ridiculous mistranslations - simple things like filling out holes in the vocabulary by importing whole dictionaries and lists of idiomatic phrases with meaning based translations, both of which can be done without changing the 'automatic learning by discovery' algorithms in any way, plus things like supplying words with morphological markers and using simple rules of thumb (transformations, field models) to stick them together - a method which according to my tests apparently already is in use in isolated cases, but not generalized as a way to do things. The final thing that would make automatic translations really useful would be to connect the words with facts about the world, maybe by introducing semantic fields based on some kind of factor analysis of texts. In short, I see no reason why automated translations shouldn't improve a lot in the coming years.

I haven't done nearly as much to study speech recognition (and character recognition), but the necessary ingredient would probably be to switch from simple comparisons to some kind of pattern analysis - which is how we analyse speech ourselves. If you try to understand a person with a lisp or a heavy accent you can't just compare the general wawe shape of a word said by that person with shapes you have heard before, and the proof that something more is going on is that you may have to substitute one sound with another in all contexts to identify the intended words. Or if you read an advertisement with some weird font you may notice that the horisontal lines have been dropped from all A's and derivatives thereof (like Æ and Å and Ä). If you just relied on simple comparison the machine would conclude that Λ wasn't the same letter as A, and there its analysis would grind to a screeching halt. So pattern analysis is crucial - and hopefully already employed, but as I said: I haven't studied this subject in detail.

Summa summarum: we can expect some progress, but maybe not to level where the machines can compete with well educated humans. Instead something else may happen: they could become so good at understanding each other that they will cut US out of the communication lines. Why should a machine bother with a meatblob that can't even pronounce a simple Czech sentence like "Věta vedlejší může vyjadřovat kterýkoliv člen své řídící věty" correctly?

[Cavesa]

I wholeheartedly agree with what has been said. Not only some language pairs are gonna suck for one or two more centuries in the automated translation but even the good ones will be far from precision. EMK's and Iversen's posts are great. And it is very true we hear about just a few mainstream languages of the rich europeans and americans. Adrianslont is right that such a service won't target the languages that could use it the most.

For basic transactions, I agree with Reineke, there is no point. Most people can use neanderthalian English or just point at things. It works. And if someone is embarassed due to this, they should either learn a language or stop being embarassed, no need to change the world for that.

But there is one more aspect not taken into account. Authenticity. It has been mentioned in various threads over the years. Not only as precision of the message but also jokes, a particular style or register you choose to use, all that combined with non verbal communication. And simply the feeling of direct access to someone or something, like Jean-Luc describes. Even though I would highlight the emotional aspect of this.

I know it sounds absolutely stupid and worthless, compared to having a digital babble fish for doing business with anyone. But I think people are getting tired of the technologies in some ways already. Sure, they are and will always be a part of our lives and in some ways much more than now. But we are at times hitting the ceiling of what is comfortable (I am not going into the ethical aspects and possible abuse of the technologies here, I am really talking about the emotions tied to this). Many people, including the young ones that cannot be suspected of nostalgia, are craving authentic experiences. People are giving each other experience vouchers for an afternoon of driving an expensive car or digging a hole with the real machine. Gardening gets new waves of popularity even in the cities. There are tons of websites and videos about making something yourself instead of buying an objectively better product.

Foreign languages are like this and will be like this even more. They are a unique direct channel to other people, books, newspapers, street signs, radio, and lots of other stuff. They are a part of any travelling adventure (and people are craving an authentic adventure here too, more and more people are going abroad individually instead of an organised crowd). They are still a sign of education in many countries and spheres of the society. They are a way to break through any national propaganda of any government, that is why the totalitarian ones hate language learning so much (yes, speaking a western language could have brought you troubles during the communist era. I've recently read about a priest, who was punished not just for serving an illegal mass but doing so in Esperanto!). They feel different in music, they are a part of the whole listening experience. And let's not forget some countries are using their language as a natural protection of a part of their science (I've read so about China and I find it only logical), profiting from others not understanding until they choose to share their results in English. Or look at the heritage learners, it is becoming more and more common to crave the connection to one's roots, instead of striving to assimilate as perfectly as possible. And even in any professional setting, have you noticed how pleased many natives are, if you show honest interest in their culture by speaking the language? It makes the communication much simpler not just because of a practical language barrier. It also helps break the emotional one and makes people more open to you. That is something that shouldn't be underestimated. Noone will ever feel flattered that you'd bother to buy a translation device.

I think this authenticity is something that machines won't ever surpass. Sure, there will be fewer tourist learners. But perhaps the market won't be that much worse without tons of crappy "courses" for basic transactions. But I think we are likely to actually see a rise of interest in language learning for personal reasons. Impossible to tell when, how much, and in what languages. But I think it will happen.

[Stefan]

Somewhat related:



Speaking of Google Translate, maybe people demand a bit too much? 15 years ago we had to bring a regular dictionary and look up word for word. Now you can just take a photo or press a button and your phone will translate everything for you.

Users are uploading thousands of subtitles for YouTube every day, feeding Google with more data for these projects. A while back they launched auto generated subtitles and this year they enabled them for live broadcasts. As of July 2015, users uploaded 400 hours of new content, every minute. 300% more than two years earlier. Combine it with Google Assistant and their crawl engine and I wouldn't be surprised if they get more relevant data every year than all of the previous years combined.

Where will it be in another 15 years?

[iguanamon]

I read this article today in GlobalVoices The Crowd-Sourced Faroe Islands Answer to Google Translate. It dates from December 2017... I'm still catching up from my Hurricane Maria caused lack of internet.

What do you do when your national language doesn't show up in Google Translate? If you're the Faroe Islands, you just crowd-source your own solution.
Faroe Islands Translate is a fascinating and fun website created by Visit Faroe Islands, the tourism authority for the islands, and Atlantic Airways, the country's official airline. On the website, locals demonstrate how common (and not so common) words and phrases from a variety of languages — including English, Spanish and French- are spoken in Faroese, the country's official language

There are a lot of phrases translated already with videos of native-speakers and it looks like a fun way to learn a little Faroese with what's already there. Since it's GlobalVoices.org, the article is available in the following human-translated languages: Malagasy, Italiano, عربي, Français, Español, русский, 繁體中文, Català, English.

OLAC Faroese Resources

[Random Review]

At 40 years old and having lived almost my whole life in poverty until a year and a half ago, I take a more cynical view of this.
I can well believe that in a decade or two, real-time machine translation will render language learning redundant both for the rich worldwide and for professionals in rich countries; but for the average Joe in most countries and many of the poorest in rich countries, that will not be empowering.

Learning empowers people; products selectively empower some people (sometimes at the expense of others).
Learning and teaching languages will remain vitally important after machine translation hits these levels.

What will change is that teaching certain languages like English will probably cease to be a way for folk like me to make a comfortable living. I don't take that lightly, having been forced to earn a not-at-all comfortable living working my ass off washing dishes until a few years ago; but in the grand scheme of things, this is really not so important.

Finally, most people on here learn languages for the love of it or out of intellectual curiosity. This won't change anything for us either.

[Cavesa]

Speaking of Google Translate, maybe people demand a bit too much? 15 years ago we had to bring a regular dictionary and look up word for word. Now you can just take a photo or press a button and your phone will translate everything for you.

Users are uploading thousands of subtitles for YouTube every day, feeding Google with more data for these projects. A while back they launched auto generated subtitles and this year they enabled them for live broadcasts. As of July 2015, users uploaded 400 hours of new content, every minute. 300% more than two years earlier. Combine it with Google Assistant and their crawl engine and I wouldn't be surprised if they get more relevant data every year than all of the previous years combined.

Where will it be in another 15 years?

15 years ago, it was impossible to receive a business email someone "kindly" put into the google translate for you. Back then, you either learnt the language or you somehow got through it with a lot of dictionary use. And the person on the other side of the communication either wrote you in English (or another large language), not underestimating your intelligence and education, or they had someone translate it, which was respectful.

Now? It is sometimes horrible and I am saying that based on a rather small sample that confirms what I heard form more experienced people (what horrors do people with regular work get and how often, hard to imagine). Instead of getting a normal email in either langauge, you get a google translate result, that is impossible to decypher. Not only are some things completely opposite. But you usually need those "details" that get lost. And it is impossible to decypher the riddle without having the original text.

And now you are in an unpleasant social situation. You cannot just do your job and reply properly right away, and people dislike delays. And you look like a moron. The native of the other language cannot understand how stupid you must be that you cannot understand a text in your own language. (Nope, they don't know Google Translate is simply horrible for longer texts in general and for almost any language combination that is not En-FR or En-Es.) Obviously, you also don't appreciate their hyperkind gesture, how arrogant of you! And you want them to waste time looking for the original text, which they may not even have anymore. And you are obviously a neanderthalian far behind the 21st century, if you underestimate the power of the Google Translate. Someone should go to your second rate country and explain to you what it is, becasue an English native always knows the best. Not to mention that they were so kind to use the Translate because of your incompetence in the first place, because everyone knows people in your country are stupid and cannot speak any languages.

Yes, you are right the rate of data collection has skyrocketed in the last decade and that will help a lot. But still, I don't think the progress will be that fast to change everything in 15 or 20 years. Perhaps for the En-Fr and En-Es translation. But not for the rest.

At 40 years old and having lived almost my whole life in poverty until a year and a half ago, I take a more cynical view of this.
I can well believe that in a decade or two, real-time machine translation will render language learning redundant both for the rich worldwide and for professionals in rich countries; but for the average Joe in most countries and many of the poorest in rich countries, that will not be empowering.

Learning empowers people, products selectively empower some people (sometimes at the expense of others).
Learning and teaching languages will remain vitally important after machine translation hits these levels.

What will change is that teaching certain languages like English will probably cease to be a way for folk like me to make a comfortable living. I don't take that lightly, having been forced to earn a not-at-all comfortable living working my ass off washing dishes until a few years ago; but in the grand scheme of things, this is really not so important.

Finally, most people on here learn languages for the love of it or out of intellectual curiosity. This won't change anything for us either.

Thanks for you post, I think you bring in very important points!

I wouldn't be so sure about only the rich having access to some parts of those projects. Google Translate and similar huge ones will still need input from hundreds of millions of people to keep improving the tools. So, a mobile app for tourists (that means middle income and a part of the poorer people who simply prioritise travelling a lot and are still not at the absolute survival minimum income) is likely to happen. But anything more sophisticated will be just for someone, true.

And you are right that this is one of the areas where the machines will destroy jobs. It will happen to the English translation the first. Knowing English will not longer be an advantage on the job market in many countries and for a wide spectrum of people. Yes, the progress for other languages will be slower, but there are also fewer opportunities for their speakers. Also, we might see an unfortunate growth of two step translations. Instead of having something translated from a language X to Y, it will be more and more often translated from X to English and from English to Y, which is definitely not ideal.

This will replace part of the translators (not the high end ones competing for the few jobs in literature and similar stuff. But those tons of people translating tiny stuff for companies). And as a result the teachers too. The ESL business is huge. It will be a good and a bad thing in one, for many reasons. But for the individuals losing the jobs, it will be only the problem.

[rdearman]

And you are right that this is one of the areas where the machines will destroy jobs.

The reason new technology has not been the cause of mass unemployment is that new kit will only be used when it makes the productive process more profitable. Higher productivity frees up the resources to buy other goods and services. The rural workers that Thomas Hardy described in Tess of the D’Urbervilles found work in factories and offices. What’s more, it was better paid work, and so the upshot was an increase in living standards.
Business Today: sign up for a morning shot of financial news
Read more

Similarly, the age of robots will lead to more jobs. Kallum Pickering, analyst with Berenberg, says there is a big hole in the argument that artificial intelligence (AI) will lead to vast numbers of workers joining the dole queue.

“Producers will only automate if doing so is profitable. For profit to occur, producers need a market to sell to in the first place. Keeping this in mind helps to highlight the critical flaw of the argument: if robots replaced all workers, thereby creating mass unemployment, to whom would the producers sell? Because demand is infinite whereas supply is scarce, the displaced workers always have the opportunity to find fresh employment to produce something that satisfies demand elsewhere.”

[Cavesa]

And you are right that this is one of the areas where the machines will destroy jobs.

The reason new technology has not been the cause of mass unemployment is that new kit will only be used when it makes the productive process more profitable. Higher productivity frees up the resources to buy other goods and services. The rural workers that Thomas Hardy described in Tess of the D’Urbervilles found work in factories and offices. What’s more, it was better paid work, and so the upshot was an increase in living standards.
Business Today: sign up for a morning shot of financial news
Read more

Similarly, the age of robots will lead to more jobs. Kallum Pickering, analyst with Berenberg, says there is a big hole in the argument that artificial intelligence (AI) will lead to vast numbers of workers joining the dole queue.

“Producers will only automate if doing so is profitable. For profit to occur, producers need a market to sell to in the first place. Keeping this in mind helps to highlight the critical flaw of the argument: if robots replaced all workers, thereby creating mass unemployment, to whom would the producers sell? Because demand is infinite whereas supply is scarce, the displaced workers always have the opportunity to find fresh employment to produce something that satisfies demand elsewhere.”

You're absolutely right.

In general, I am all for increasing productivity and for implementing the modern technologies wherever it is useful and without too many negative side effects (some parts of the healthcare visions with the AI are pretty scary, for example). It is just important not to forget about a lot of individuals with problems during this change to more productivity in most areas of the economy and work life. And it is simply obvious that lots of the jobs with low entry requirements (a typical example would be B1 or B2 English for lots of people in various fields. Or CELTA, in case of teachers) are gonna dissappear or at least diminish.

It will be more important than ever before to spread language learning as a tool of personal development, a path to authentic experiences, and a fascinating thing to do that happens to be useful too despite the existing gadget alternative.

[Iversen]

I rarely watch TV with subtitles with the sound turned on, but a couple of days ago I was puzzled by the pauses and the stilted way the commentary to a program on Danish TV was spoken.

It was (as usual) a documentary bought from an Anglophone source, but they had cut out the English speaker and provided speech in Danish instead. And then I found the subtitles on text TV to confirm a hunch: yes, they had written the subtitles first, and then they let a speech synthethizer read themt aloud, word for word. Which solves several problems for language learners, namely that normal subtitles rarely follow the speech closely, and that modern professional actors and journalists speak in a hurried and muffled way which even native Danes sometimes have trouble understanding.

The day may come where the TV people let a program listen to the English sound, another program then (mis)translates the result, AI program will cut out anything it can't understand, and then a fourth program will read the patchy mishmash aloud. And then the humans in DR and TV2 can get even more leisure time to chat with each others while TV viewers pay their coffee.

Actually I thought the development would be to teach all Danes to understand the English programs without a translation, so now I wonder whether we'll see two diverging lines in 'Danish' TV production: 1) English speak, Danish subtitles, 2) subtitles in Danish, read aloud in Danish by a machine. If no. 2 happens then our advantage in acquiring English skills may be compromised.