The Shallowness of Google Translate

A relative of mine, who is aware of the time that I put into Duolingo, asked me why I bothered since programs such as Google Translate will soon make the need for learning multiple languages obsolete. In response I referred him to the following excellent article by Douglas Hofstadter, a professor of cognitive science and comparative literature:

I thought all of you who work so hard to learn multiple languages might appreciate it too.

July 3, 2018


People ask me why I study languages when everyone speaks English, but I've lived in several countries and experienced cultures first hand because I'm multilingual. People ask me why I paint when you can just take a picture, but they haven't experience the joy of seeing something with your heart and not just your eyes. People ask me why I grow my own vegetables when you can buy them cheaper at the store, but they haven't had a really fresh tomato or corn right off the stalk. People ask me why I sew when you can buy a shirt for so cheap, but they don't get to choose their fabric or make it fit just so.

Go do stuff. Learn languages, paint, sing, play music, grow a garden, build a boat or your own car. That's why we're here.

July 3, 2018

Beautifully written! This inspires me!

July 4, 2018

Very interesting. I value machine translation engines for help with understanding basic sentences although good quality dicionaries and sites like reverso are more useful if you know enough to understand what phrases and words to focus on.

If all you want to do is interact with staff in shops and museums as a tourist then I'm sure machine translation is good enough, although having your phone say "how much is that" or "what time do you close" doesn't appeal to me as much as actually saying those phrases yourself.

The article ably pointed out that machine translation is not much use for more subtle sentences. I liked the example where the point was to contrast 'his' version with 'her' version and I learned some French. :)

July 3, 2018 A different perspective. Google translate is still in its early stages. Translation technology is a wonderful tool that like all early technologies is improving. It is short sighted to judge it by its early stages (remember the internet in its beginnings). Translation technology is allowing people to communicate across language barriers. I do not understand why people feel the need to knock down this amazing human achievement of scientific and social progress, one of the most amazing ones of our time, just to feel better about their own abilities.

July 3, 2018

I do not understand why people feel the need to knock down this amazing human achievement of scientific and social progress, one of the most amazing ones of our time, just to feel better about their own abilities.

Do you believe that Hofstadter is one of those people?

July 3, 2018

I am only familiar with Hofstader from this one essay, so I cannot make such a statement, however that was my impression of the general tone of the article.

"From my point of view, there is no fundamental reason that machines could not, in principle, someday think, be creative, funny, nostalgic, excited, frightened, ecstatic, resigned, hopeful, and, as a corollary, able to translate admirably between languages. There’s no fundamental reason that machines might not someday succeed smashingly in translating jokes, puns, screenplays, novels, poems, and, of course, essays like this one. But all that will come about only when machines are as filled with ideas, emotions, and experiences as human beings are. And that’s not around the corner. Indeed, I believe it is still extremely far away. At least that is what this lifelong admirer of the human mind’s profundity fervently hopes.". Hofstader.

I think that paragraph illustrates my point. I think he gave a fairly fair overview of the current abilities and weaknesses of google translate. However, in my opinion, his view of the future is a bit short sighted.. perhaps I just feel this technology is moving ahead at a far faster pace than the author foresees. (and I am no tech expert, this is based on the articles I have read on A.I and also the rapid changes in technology I have seen in my own life) However, as he candidly admits, he is hoping that the perfection of A.I is far in the future. I get the impression he feels it would somehow undermine the art and literature humans have produced if machines can do the same. However, I feel if anything it would be the opposite. (I am nervous about A.I. too but for other reasons). If a machine could produce great works one day... would that not be a great human artistic acheivement in itself? If a computer composes beautiful music, that does not make the music of Mozart any less so. I think, the article takes a 'shallow' view of both the current and future uses of translation technology and is short sighted. I feel part of this is fear, that the author themself confesses, however, if we are afraid of A.I., of losing our place as creative beings, we should be facing this fear and asking ourselves the tough questions of what it will mean to be human when we are no longer the most intelligent beings on this planet, perhaps talking some cautions in how this technology is developed, rather than the hubris of assuming this will not happen. Anyways... that was my interpretation of the article.

And as to the current benefits of google translate, here is a short article about how translation technology is helping give people all over the world access to information, including important fields such as health care. This technology is making the world more connected, making information far more accessible is ways that go beyond ordering coffee on vacation.

July 3, 2018

I think that paragraph illustrates my point.

In my view, that paragraph does not show him “knocking down” Google Translate to feel better about his own abilities. He is old enough to be aware of both the extents and the limitations of his own abilities (e.g. his “pi-lingualism”), and astute enough to recognize that it is unlikely that there will be an excellent translation of, say, Cervantes’ Don Quixote produced by AI in his remaining years. As he’d noted in a different paragraph,

I wouldn’t want to leave readers with the impression that I believe intelligence and understanding to be forever inaccessible to computers. If in this essay I seem to come across sounding that way, it’s because the technology I’ve been discussing makes no attempt to reproduce human intelligence. Quite the contrary: It attempts to make an end run around human intelligence, and the output passages exhibited above clearly reveal its giant lacunas.

Translating some types of information (e.g. instructions on how to diagnose certain diseases) is distinct from translating other types of information (e.g. double entendres); the first type is much more amenable to automated translation than the second type is, because fewer contexts need to be considered in the first type to produce an acceptable translation. Machines might not need to have their own ideas, emotions, and experiences to eventually be able to translate the second type, but at minimum they would need to recognize others’ ideas, emotions, and experiences, to bring such contexts into consideration in their translations.

July 3, 2018

Speaking as someone who has looked in depth at the technical issues, having just completed a year of academic study for a master's in computational linguistics, the statement you quote is more fact than opinion. The part about fervently hoping that the machines do not soon come to dominate the field, that I grant you is opinion. The rest is a very fair assessment of the state of the art. The current methodology in the field of NLP (natural language processing) is largely based on statistical patterns being extracted from huge quantities of data, rather than from any attempt to build a model that incorporates the semantic and pragmatic understanding, the cultural context, the subtext of the discourse, and the authorial intent behind a particular passage. We have begun, in a fashion, to model the semantics and set/superset relationships of certain words (nouns more readily than verbs) but this process has taken years and the model is not viable across all languages. We have barely mapped the syntax of English, for which billions of new words of data appear every week. This is because there are two schools of thought for tackling this incredibly complex problem of human communication: one comes from mathematicians, and the other comes from linguists. Right now the mathematicians are making some progress on certain kinds of analytic tasks, such as search engines and document classification, that are susceptible to their preferred method; but despite a trillion-word corpus and the fastest computers available, the mathematical method is stymied by even simple examples of translation as given in the article, translation that any bilingual human being can perform easily with our comparatively limited processing capacity. I agree that we will have to build a computer model, as he suggests, that actually comprehends the sentence on some level, before the translation will be of human quality.

July 3, 2018

Not that it matters significantly to the article, but... "Even to a native English speaker, the missing article on the leopard is the only real giveaway that No. 2 was the output of an automaton." No. It just isn't. I knew which was which from the first sentence. Admittedly, I was paying closer attention than I would normally, but even so after the second sentence of the machine translation I'd be thinking that the writer wasn't a native speaker of English.

July 3, 2018

A few weeks ago I took my children (ages 5 and 6) to the library. We ran into one of my son's classmates and his family. I tried to introduce myself and saw a look of panic and terror on the mother's face. I quickly remembered they had just arrived from Colombia two months earlier. I switched to Spanish and she relaxed immediately. Her English is B1 and so is my Spanish. While the kids played, we talked, switching between the two languages as each of us stumbled on the other's language.

There would have been no emotional connection with Google Translate. She wouldn't have felt welcome in her new home. That's why I learn languages, to make a personal connection with another human being

(I'm not knocking Google Translate or its usefulness. I use it all the time to double check words and phrases that I'm unsure of. It has accelerated my ability to learn languages, and therefore, to connect with other people. But at the end of the day, it's a tool, like a hammer. Hammers don't make many friends on their own.)

July 21, 2018

My experience with google translate tells me that it's usefulness is very limited, it can give you the incorrect translation even in some simple cases (at least in case of my mother-tongue), for example, if you type in name of one middle aged musician (not famous world wide though) the translation you get in English is Rihanna. That's one example, there are plenty of others.

July 3, 2018

Without benefit of reading the article or doing research, my comment relies on my personal experience with the translate app on the phone. It seems to have vastly improved over the past year and continually improves. On occasion I used it here when rapid fire sentences seem like one big word and words I know are there but don't seem to be spoken. The app has nailed it exactly as spoken. I also use it with native speakers for new words for me and so far that has worked out. However, I would not trade it for fluency.

July 3, 2018

[deactivated user]

    I think there is a sense of pride we can feel when we learn a language through hard work rather than taking the easy way out. Maybe it's just me.

    July 4, 2018

    I use Google translate to supplement Duolingo.

    July 3, 2018

    I want to read literature in other languages. Google translate isn't going to get meanings that would be lost in translation even if a person did the translating across. No machine will ever be able to fully translate the depth of language.

    Besides that, when you meet a native speaker of another language, they appreciate hearing you speak it even if they "know English". A computer can't replace that.

    Honestly, I feel like this is something people who are monolingual and don't want to bother learning another language say so they don't have to feel guilty. I think a lot of people just see language learning as boring and wouldn't do it unless they had to.

    Learning a language is good for my mind just as exercise is good for my body. Do people ask "Why do you exercise? Exercise is a waste of time because modern medicine can fix any health problems I get from not exercising."

    There is nothing that relaxes me the way language learning does.

    July 3, 2018

    Have a lingot for sharing this really interesting article! I have to say I agree with Hofstadter about the current state of Google Translate. I assume that language and computer scientists are working on translating not only the words but the ideas as well. How long it will take until they have some measure of success is beyond me.

    Still, even if or when AI translators get better at getting the point across, I will still continue to learn foreign languages, because it is simply something I very much like to do ( :

    July 4, 2018

    Thank you so much for this post! I want to share this with some monolinguals I know. Really, the amount of misunderstanding there is...they say, "oh, I speak (some) Spanish!" and I ask how they learned, and they say "Google Translate!" Then I speak with them and feel bound to tell them that their articles and adjective endings and other trivial pieces that mean so much are...totally and completely distorted. Or, when I talk about how my dream is to learn Russian into fluency and travel to Russia, and they say, "that is way too hard. Can't you just learn a few phrases and limp along with GT?" Okay...maybe I could. But why would I? Do I want to just be that chick in the restaurant who wants that everyone speak English because she didn't take any time to learn these things? Do I want to knowingly or innocently frustrate people? No! Maybe it is my pride...I'm not sure. But I also realize the difference between the openness of someone speaking in their language and someone having to use a foreign one with you...I would way rather get to know the "real them" if I could. It's so valuable and the conversations are so much deeper. And also, as I say to GT zealots all the time: "Okay, think of this. You get captured in another country. They are questioning you out in a shed somewhere. Where is your phone?" ;))))

    A funny story was when my ex decided that he would help me practice Russian by using GT to write me and having me write my own original answers back. I thought, sure, this isn't optimal. But it's funny...Well, we were doing pretty well, and then I asked in Russian if I should write something else in Russian. He wrote back, "Yes! Try me!" (in the most American way possible). The translation was shielding an entirely different motive, if you know what I mean. It was also formed as the way a lady would say it. Then he asked me to write something, I responded in Russian: "Google Translate is evil. Stop now." Needless to say, that was the end of Google Translate practices and the end of my already-damaged hope in it. :)))

    I hope it gets better! I hope that one day newscasters from anywhere in the world can speak! But as for me....I'll stay lone ranger as long as possible, because I want the skill.

    July 4, 2018

    Darn right, man! All my experiences with Google translate have been just laughable. Once you get beyond one word it just can’t get it. You can’t translate context and intent... the things that really make up the meaning of a sentence. Google may “know everything” but they will never UNDERSTAND anything !

    July 7, 2018
    Learn a language in just 5 minutes a day. For free.