4 ms·

"People will be fluent in every language." Given the current state of language translating software, I think this is highly unlikely in the next 10 years. Anyo

"People will be fluent in every language." Given the current state of language translating software, I think this is highly unlikely in the next 10 years. Anyone who is bilingual and has used a translation app knows how woefully inadequate it is for actual communication. The nuances of language are far too complex, not to mention the importance of non-verbal communication. How can software correctly translate the answer "Well, I'm not sure..." (which could signify hesitation to accept, passive disagreement, a pause for time to think, etc.) without understanding the context of the answer, and the tone of voice and facial expression of the speaker? Not in 10 years...

sixQuarks 14y ago
I think it's possible. Human brains are not wired to understand exponential growth. I don't think it will happen within 5 years, but 10 = very possible.
- dalke 14y ago
  You've just said that human brains "are not wired to understand exponential growth" but then made a prediction based on understanding exponential growth. Plus, you haven't established the doubling factor for the growth - it might be exponential but needing 20 years to get twice as good. "Perfect instant translation" requires full AI. Consider the Borges story "Hombre de la esquina rosada". That translates directly to "The Man of the Pink Corner." But the "corner" here means also the corner store, so it's not just the physical corner but also the "hangout for the low-life of the barrio." (I'm quoting from the translator's notes here.) Moreover, the "pink" refers to the pastel hues popular in earlier decades, so the pink implies an older neighborhood "populated by toughs and knife fighters." If you knew Buenos Aires of 100 years ago then you would infer that directly from the title. But without the context you just know it as "the pink corner." I've thought that "The Hood of the Pink Corner" would be a better translation, given the double or triple meaning of "hood" to mean "hoodlum", "neighborhood", and more specifically "poor neighborhood." It's not as direct, but it might be better. How do you explain those nuances to a computer? Douglas Hofstader gives more examples of the difficulties in natural language to natural language translation. How does one translate a nonsense poem like "Jabberwocky"? One French translation starts "Il brilgue: les tôves lubricilleux. Se gyrent en vrillant dans le guave..." Or, and I quote from his Gödel, Escher, Bach: Take a look at the first sentence of Dostoevsky's novel 'Crime and Punishment' in Russian and then in a few different English translations... The first sentence employs the street name "S. Pereulok" (as transliterated). What is the meaning of this? A careful reader of Dostoevsky's work who knows Leningrad ... can discover by doing some careful checking of the rest of the geography in the book ... that the street must be 'Stoliaryny Peteulok". .. should we keep the initial so as to reproduce the aura of semi-mystery which appears already in this first sentence of the book? We could get to "S. Lane", ("lane" being the standard translation of "pereulok"). None of the translators took this tack. However, one chose to write "S. Place". [The second] gave the translation as "Stoliarny Place". What about number 3? That's the most interesting of all. The translation says "Carpenter's Lane." And why not, indeed? After all, "stoliar" means "carpenter" and "ny" is an adjectival ending. He then goes on to argue that perhaps we should just read Dickens instead, since Dickens is, at some level, a "translation" of the Dosteovsky novel. BTW, Google Translate gives "S. Lane", which none of the three translators used. As that's the obvious and most direct translation, which the humans didn't choose, I assume that we can say it's not the perfect translation. How would an automatic translation system, without AI, look at "С. Переулок", figure out that it's supposed to be "Столярный Переулок", and translate it correctly?
  - sixQuarks 14y ago
    Just because human brains are not wired to understand exponential growth doesn't mean you can't understand it. I'm talking about human intuition. I understand the problems inherent in translation. But again, I still feel it's possible to have instant translation within 10 years. The only way to find out is to wait and see. I'll give you a dollar if I'm wrong in 2023.
    - dalke 14y ago
      ha-ha on myself! I read "With DARPA and Google racing to perfect instant translation" as "PERfect" meaning 100% perfect quality, and not "perFECT" as in to smooth out the details. Well, there goes my PERfect understanding skills. That's not a bet I want to take. We have "instant translation" now - I've seen demos. It's just rather crappy. What level do you expect that we'll have in 10 years? My challenge would be to take chapter 12 of Gödel, Escher, Bach and translate it into Swahili then ask the Swahili speaker if it seems stilted or normal. Well, compared to what a human translator would do. Chapter 12 contains a mixture of English, French, German, and transliterated Russian, but is perfectly understandable to someone who doesn't know the non-English language, because the text describes how to understand the missing parts. It also uses "ASU" as an inversion of "USA", where "ASU" is defined in the text as "Alternative Structure of the Union." Will the translation into Swahili handle that in-text complexity?
      - sixQuarks 14y ago
        I don't think 90% of the population can even understand Godel, Escher, Bach, let alone translate it in any meaningful way. You're asking too much from a computer. Perhaps in 30 years that can happen. I think computers will be able to translate very accurately at a middle-school level within 10 years.
        dalke 14y ago
        This wasn't a hard chapter of GEB. :) Okay, what about Asterix and Obelisk? Still, can you quantify what "very accurately" and "middle-school level" means? Do you think in 10 years that machine translation will be able to translate the following two (independent) paragraphs correctly? Hikaru liked the old English teacher more than the new Scottish one. He still found it hard to understand the replacement's strange dialect as she droned away about the Battle of Agincourt, the Treaties of Aix-la-Chapelle, or how Paris 'deserves a mass', while the boyish Londoner had made far-away European history seem exciting. Hikaru liked the old English teacher most of all. He moved to London only a year ago, and Mrs. Bane had helped him fall in love with the language. Her 40 years of service had not stunted her enthusiasm one bit, as she taught him and her other charges about the subtleties in Shakespeare, the humor in Carroll, and the epic power of Morrison. In the first, "old English teacher" refers to the previous history teacher, who was from London. In the second, it refers to the teacher of English, who is at least 60 years old. In some Germanic languages, the first is written as something akin to "english teacher" and the second "englishteacher." To translate it correctly, the translator has to read ahead to disambiguate the meaning. Which isn't hard, even for middle-school students. But it is hard for computers. In other languages, gender is more important. In the first paragraph, it's very likely that the "old English teacher" is a man, and the new Scottish one is a woman. In the second paragraph, the "old English teacher" is a woman. Is your claim that an automatic translation system will be able to translate these correctly to (say) Swedish or French, within 10 years? I don't see it happening. To do so requires some quite good AI, while much of the progress over the last 10 years has been by rejecting traditional AI and collecting large statistical samples. BTW, the current Swedish translations from Google translate for these are "Hikaru gillade gamla engelsklärare mer än den nya skotska ett." and "Hikaru gillade gamla engelsklärare mest av allt." Both are wrong. The first should be more like "Hikaru gillade den förra engelsk lärare mer än den nya skotska." and the second "Hikaru gillade den gamla engelsklärare mest av allt.
ChuckMcM 14y ago
One wonders if we'll converge on a single common language. But in the case of the article they were implying your phone would translate for you.
- nemo1618 14y ago
  What about a common "intermediary" language? Instead of writing n^2 translation algorithms, you'd only need 2n: one to translate into the intermediary, and one to translate out. Such a language would have to be incredibly flexible though; my vote is for lojban ;). Done correctly, such a system could also teach the intermediary language to the speaker during translation. Eventually people would start to forgo the translation process and simply speak the common tongue.
  - robryan 14y ago
    Which would already be English, I assume google translate works a lot better from and to English than other language combinations.
    - ChuckMcM 14y ago
      Android may have more Chinese speakers than English speakers here pretty quickly.
      - robryan 14y ago
        I am fairly ignorant on languages. Does Chinese keep up with the fast pace of innovation these days by creating new words for new inventions from the English speaking world? One of the reasons I think English does make a good exchange language is because it generally has a word for anything that comes along. Also it is the easiest to deal with in terms of building software and existing natural language processing research.
tokenadult 14y ago
From the article: "People will be fluent in every language. With DARPA and Google racing to perfect instant translation, it won't be long until your cellphone speaks Swahili on your behalf." That was suggested as a plausible prediction for 2022. The comment to which I reply says, Given the current state of language translating software, I think this is highly unlikely in the next 10 years. Anyone who is bilingual and has used a translation app knows how woefully inadequate it is for actual communication. The last time a video demonstration of software in development for language translation was posted here on HN, http://news.ycombinator.com/item?id=4759620 http://news.ycombinator.com/item?id=4759620 I commented, To someone who spent years learning Chinese as a second language, and then made my living for years as a Chinese-English interpreter, that was pretty impressive. The economics of the issue is that a machine interpreter just has to be as good as a human interpreter at the same cost. That's a reachable target with today's computer technology. EVERY time I've heard someone else interpreting English or Chinese into the other language, I have heard mistakes, and I am chagrined to remember mistakes that I made over the years. We can't count on error-free machine interpretation between any pair of languages (human language is too ambiguous in many daily life cases for that), but if companies develop tested, validated software solutions for consecutive interpreting (what I usually did, and what is shown in the video) or simultaneous interpreting (the harder kind of interpreting in demand at the United Nations, where even in the best case it is not always done well), then those companies will be able to displace a lot of human professionals who rely on their language ability to make a living. And relating that comment to a timeline appropriate for today's thread, I think ten years out is about right for a mobile app widely available at reasonable cost that will displace human interpreters for most use cases that now require human interpreters. As I further commented on the the earlier thread 37 days ago, Right now a lot of interpreters in the United States make a lot of part-time income from gigs that involve suddenly getting telephone calls and joining in to interpret a telephone conversation in two languages. This is often necessary, for example, for physician interviews of patients in emergency rooms or pharmacist consultations with patients buying prescribed drugs (where I last saw a posted notice on how to access such an interpretation service). The IBM Watson project is already targeted at becoming an expert system for medical diagnosis, and patient care markets will surely provide a lot of income for further development of software interpretation between human languages. It's still good for human beings to spend the time and effort to learn another human language (as so many HN participants have by learning English as a second language). That's a broadening experience and an intellectual delight. But just as riding horses is more a form of recreation these days than a basis for being employed, so too speaking another language will be a declining factor in seeking employment in the next decade. If there is economic demand for interpreting between a given language pair, there will be great economic incentive to develop software-as-a-service to do interpreting for that language pair. I would expect Japanese-English and Spanish-English and Chinese-English interpreting all to decline as a human paying occupation by ten years from now, even though international communication will surely increase in the same time frame.
SiVal 14y ago
Sure, people with a pocket translator will be "fluent" in every language just like people who don't know the difference between "divide by one-half" and "divide in half" but have a calculator app on their phone are now brilliant at math.

sixQuarks 14y ago

I think it's possible. Human brains are not wired to understand exponential growth. I don't think it will happen within 5 years, but 10 = very possible.

dalke 14y ago
You've just said that human brains "are not wired to understand exponential growth" but then made a prediction based on understanding exponential growth. Plus, you haven't established the doubling factor for the growth - it might be exponential but needing 20 years to get twice as good. "Perfect instant translation" requires full AI. Consider the Borges story "Hombre de la esquina rosada". That translates directly to "The Man of the Pink Corner." But the "corner" here means also the corner store, so it's not just the physical corner but also the "hangout for the low-life of the barrio." (I'm quoting from the translator's notes here.) Moreover, the "pink" refers to the pastel hues popular in earlier decades, so the pink implies an older neighborhood "populated by toughs and knife fighters." If you knew Buenos Aires of 100 years ago then you would infer that directly from the title. But without the context you just know it as "the pink corner." I've thought that "The Hood of the Pink Corner" would be a better translation, given the double or triple meaning of "hood" to mean "hoodlum", "neighborhood", and more specifically "poor neighborhood." It's not as direct, but it might be better. How do you explain those nuances to a computer? Douglas Hofstader gives more examples of the difficulties in natural language to natural language translation. How does one translate a nonsense poem like "Jabberwocky"? One French translation starts "Il brilgue: les tôves lubricilleux. Se gyrent en vrillant dans le guave..." Or, and I quote from his Gödel, Escher, Bach: Take a look at the first sentence of Dostoevsky's novel 'Crime and Punishment' in Russian and then in a few different English translations... The first sentence employs the street name "S. Pereulok" (as transliterated). What is the meaning of this? A careful reader of Dostoevsky's work who knows Leningrad ... can discover by doing some careful checking of the rest of the geography in the book ... that the street must be 'Stoliaryny Peteulok". .. should we keep the initial so as to reproduce the aura of semi-mystery which appears already in this first sentence of the book? We could get to "S. Lane", ("lane" being the standard translation of "pereulok"). None of the translators took this tack. However, one chose to write "S. Place". [The second] gave the translation as "Stoliarny Place". What about number 3? That's the most interesting of all. The translation says "Carpenter's Lane." And why not, indeed? After all, "stoliar" means "carpenter" and "ny" is an adjectival ending. He then goes on to argue that perhaps we should just read Dickens instead, since Dickens is, at some level, a "translation" of the Dosteovsky novel. BTW, Google Translate gives "S. Lane", which none of the three translators used. As that's the obvious and most direct translation, which the humans didn't choose, I assume that we can say it's not the perfect translation. How would an automatic translation system, without AI, look at "С. Переулок", figure out that it's supposed to be "Столярный Переулок", and translate it correctly?
- sixQuarks 14y ago
  Just because human brains are not wired to understand exponential growth doesn't mean you can't understand it. I'm talking about human intuition. I understand the problems inherent in translation. But again, I still feel it's possible to have instant translation within 10 years. The only way to find out is to wait and see. I'll give you a dollar if I'm wrong in 2023.
  - dalke 14y ago
    ha-ha on myself! I read "With DARPA and Google racing to perfect instant translation" as "PERfect" meaning 100% perfect quality, and not "perFECT" as in to smooth out the details. Well, there goes my PERfect understanding skills. That's not a bet I want to take. We have "instant translation" now - I've seen demos. It's just rather crappy. What level do you expect that we'll have in 10 years? My challenge would be to take chapter 12 of Gödel, Escher, Bach and translate it into Swahili then ask the Swahili speaker if it seems stilted or normal. Well, compared to what a human translator would do. Chapter 12 contains a mixture of English, French, German, and transliterated Russian, but is perfectly understandable to someone who doesn't know the non-English language, because the text describes how to understand the missing parts. It also uses "ASU" as an inversion of "USA", where "ASU" is defined in the text as "Alternative Structure of the Union." Will the translation into Swahili handle that in-text complexity?
    - sixQuarks 14y ago
      I don't think 90% of the population can even understand Godel, Escher, Bach, let alone translate it in any meaningful way. You're asking too much from a computer. Perhaps in 30 years that can happen. I think computers will be able to translate very accurately at a middle-school level within 10 years.
      - dalke 14y ago
        This wasn't a hard chapter of GEB. :) Okay, what about Asterix and Obelisk? Still, can you quantify what "very accurately" and "middle-school level" means? Do you think in 10 years that machine translation will be able to translate the following two (independent) paragraphs correctly? Hikaru liked the old English teacher more than the new Scottish one. He still found it hard to understand the replacement's strange dialect as she droned away about the Battle of Agincourt, the Treaties of Aix-la-Chapelle, or how Paris 'deserves a mass', while the boyish Londoner had made far-away European history seem exciting. Hikaru liked the old English teacher most of all. He moved to London only a year ago, and Mrs. Bane had helped him fall in love with the language. Her 40 years of service had not stunted her enthusiasm one bit, as she taught him and her other charges about the subtleties in Shakespeare, the humor in Carroll, and the epic power of Morrison. In the first, "old English teacher" refers to the previous history teacher, who was from London. In the second, it refers to the teacher of English, who is at least 60 years old. In some Germanic languages, the first is written as something akin to "english teacher" and the second "englishteacher." To translate it correctly, the translator has to read ahead to disambiguate the meaning. Which isn't hard, even for middle-school students. But it is hard for computers. In other languages, gender is more important. In the first paragraph, it's very likely that the "old English teacher" is a man, and the new Scottish one is a woman. In the second paragraph, the "old English teacher" is a woman. Is your claim that an automatic translation system will be able to translate these correctly to (say) Swedish or French, within 10 years? I don't see it happening. To do so requires some quite good AI, while much of the progress over the last 10 years has been by rejecting traditional AI and collecting large statistical samples. BTW, the current Swedish translations from Google translate for these are "Hikaru gillade gamla engelsklärare mer än den nya skotska ett." and "Hikaru gillade gamla engelsklärare mest av allt." Both are wrong. The first should be more like "Hikaru gillade den förra engelsk lärare mer än den nya skotska." and the second "Hikaru gillade den gamla engelsklärare mest av allt.

ChuckMcM 14y ago

One wonders if we'll converge on a single common language. But in the case of the article they were implying your phone would translate for you.

nemo1618 14y ago
What about a common "intermediary" language? Instead of writing n^2 translation algorithms, you'd only need 2n: one to translate into the intermediary, and one to translate out. Such a language would have to be incredibly flexible though; my vote is for lojban ;). Done correctly, such a system could also teach the intermediary language to the speaker during translation. Eventually people would start to forgo the translation process and simply speak the common tongue.
- robryan 14y ago
  Which would already be English, I assume google translate works a lot better from and to English than other language combinations.
  - ChuckMcM 14y ago
    Android may have more Chinese speakers than English speakers here pretty quickly.
    - robryan 14y ago
      I am fairly ignorant on languages. Does Chinese keep up with the fast pace of innovation these days by creating new words for new inventions from the English speaking world? One of the reasons I think English does make a good exchange language is because it generally has a word for anything that comes along. Also it is the easiest to deal with in terms of building software and existing natural language processing research.

tokenadult 14y ago

From the article: "People will be fluent in every language. With DARPA and Google racing to perfect instant translation, it won't be long until your cellphone speaks Swahili on your behalf." That was suggested as a plausible prediction for 2022. The comment to which I reply says, Given the current state of language translating software, I think this is highly unlikely in the next 10 years. Anyone who is bilingual and has used a translation app knows how woefully inadequate it is for actual communication. The last time a video demonstration of software in development for language translation was posted here on HN, http://news.ycombinator.com/item?id=4759620 http://news.ycombinator.com/item?id=4759620 I commented, To someone who spent years learning Chinese as a second language, and then made my living for years as a Chinese-English interpreter, that was pretty impressive. The economics of the issue is that a machine interpreter just has to be as good as a human interpreter at the same cost. That's a reachable target with today's computer technology. EVERY time I've heard someone else interpreting English or Chinese into the other language, I have heard mistakes, and I am chagrined to remember mistakes that I made over the years. We can't count on error-free machine interpretation between any pair of languages (human language is too ambiguous in many daily life cases for that), but if companies develop tested, validated software solutions for consecutive interpreting (what I usually did, and what is shown in the video) or simultaneous interpreting (the harder kind of interpreting in demand at the United Nations, where even in the best case it is not always done well), then those companies will be able to displace a lot of human professionals who rely on their language ability to make a living. And relating that comment to a timeline appropriate for today's thread, I think ten years out is about right for a mobile app widely available at reasonable cost that will displace human interpreters for most use cases that now require human interpreters. As I further commented on the the earlier thread 37 days ago, Right now a lot of interpreters in the United States make a lot of part-time income from gigs that involve suddenly getting telephone calls and joining in to interpret a telephone conversation in two languages. This is often necessary, for example, for physician interviews of patients in emergency rooms or pharmacist consultations with patients buying prescribed drugs (where I last saw a posted notice on how to access such an interpretation service). The IBM Watson project is already targeted at becoming an expert system for medical diagnosis, and patient care markets will surely provide a lot of income for further development of software interpretation between human languages. It's still good for human beings to spend the time and effort to learn another human language (as so many HN participants have by learning English as a second language). That's a broadening experience and an intellectual delight. But just as riding horses is more a form of recreation these days than a basis for being employed, so too speaking another language will be a declining factor in seeking employment in the next decade. If there is economic demand for interpreting between a given language pair, there will be great economic incentive to develop software-as-a-service to do interpreting for that language pair. I would expect Japanese-English and Spanish-English and Chinese-English interpreting all to decline as a human paying occupation by ten years from now, even though international communication will surely increase in the same time frame.

SiVal 14y ago

Sure, people with a pocket translator will be "fluent" in every language just like people who don't know the difference between "divide by one-half" and "divide in half" but have a calculator app on their phone are now brilliant at math.