This will be most useful in China

by on October 5, 2017 at 7:53 am in Education, Science, Web/Tech | Permalink

Google is the first major tech company to build the Babel fish.

The search company, which is now making a slew of its own hardware products, announced the Google Pixel Buds at a San Francisco event today (Oct. 4). The earbuds connect wirelessly with Google’s latest smartphones, but more importantly, they’re able to access Google Assistant, the company’s virtual personal concierge, which launched exactly a year ago. Through this software, Google claims the earbuds can translate 40 spoken languages nearly in real time—or at least, fast enough to hold a conversation.

Here is the story at Quartz.  It’s funny how economists used to come up with theories that platform monopolies would stifle innovation…

1 Beliavsky October 5, 2017 at 8:17 am

Should American students study foreign languages or just rely on Babel fish and Google Translate when they encounter something in a foreign language? I studied French in high school and have never used it.

2 dan1111 October 5, 2017 at 8:35 am

I agree that we are reaching the point where we should start asking that question.

I would not be surprised if language barriers were eliminated by technology for most practical purposes in five years.

3 prior_test3 October 5, 2017 at 9:36 am

‘I would not be surprised if language barriers were eliminated by technology’

And yet before, you seemed very aware of the problems of translation regarding a book that is read by millions of people.

Of course, it may depend on what you mean by language barriers – if one cannot be bothered to even learn a few short phrases in a foreign language, but instead relies on a computer between two speakers, it might just be a surprise to realize that many people actually prefer human interaction, at least to the extent of being greeted in their own language.

4 dan1111 October 5, 2017 at 10:04 am

This is why I said “for most practical purposes”. Nuanced, accurate translation is hard, and for translating written documents the acceptable error rate is extremely low (since there is no feedback process to correct it). I don’t expect computers to eclipse humans in translating the Bible (for example) any time soon.

What I mean is that very soon, computers will enable easy spoken communication between two humans who don’t share the same language. It won’t be perfect, but it will be very good, certainly better than the communicative ability one acquires by studying a language in school (even for a number of years). This will transform a lot of things.

5 Todd Kreider October 5, 2017 at 10:12 am

That was essentially my prediction for Japanese back in 1998 when studying the language. I’d be 50 in 2020 and said statistical translation, which will improve exponentailly, will end human translation (written) apart from literature by 2015 and that even spoken Japanese would no longer be useful for a Japan related job by 2020 since the technology would apply to ear devices as well. I ended up predicting this would happen sometime between 2018 and 2023 and closer to 2023 before Japanese is mostly no longer considered a skill for a non Japanese working in business, etc in Japan.

The problem with predicting this in 1998 is that everyone studying languages thought I was absolutely nuts for saying this.

I added that knowing Japanese very well will still be useful for talking to Japanese friends and in social situations, which is part of business, but it will no longer be important for reading reports and attending meetings.

6 Ben October 5, 2017 at 9:06 am

I don’t think enough Americans travel to use other languages. I live in the UK and it is almost unheard of unless you are on a low-income to not go abroad at least once a year whereas when you ask Americans, it is essentially the norm. Although America is clearly much larger and a great country to travel in, it’s certainly not great enough to never leave it.

Instead of not learning languages, Americans should travel enough to use them.

7 dan1111 October 5, 2017 at 9:43 am

As an American who has lived in the UK for awhile–sure, it’s partly cultural. But it’s also opportunity.

From major British cities, there are dozens of foreign destinations that you can fly to cheaply, often for less than £100 for a return ticket. You can also drive to Europe or take the train.

Americans have far more limited destinations that can be reached inexpensively: Canada, Mexico, and a few Carribean islands. Even flying to these destinations typically costs a lot more than taking EasyJet or RyanAir somewhere. Going anywhere beyond this will usually cost at least $1000 per ticket. Plenty of Americans do travel to Europe and elsewhere, but they tend to be “once in a lifetime” type trips rather than a typical summer vacation, due to cost.

And of course, there are many more domestic destinations, with a lot more variety, than Britain. Plenty of British travel is to all-inclusive beach resorts in various countries (hardly a cultural experience), and Americans don’t have to leave the country to do this. Likewise for skiing holidays and various other things.

And Americans get hardly any vacation time!

I agree that the cultural experience of going to other countries is great, and more people should do it (including more Americans). But it’s not simply a case of Americans being too parochial.

8 prior_test3 October 5, 2017 at 9:47 am

‘I don’t think enough Americans travel to use other languages.’

Yep – around 40% of Americans have a passport (with maybe a tenth of that total being passport cards, which are pretty much only good for visiting Canada and Mexico) https://travel.state.gov/content/passports/en/passports/statistics.html

One of the larger reasons Americans don’t travel is simple lack of free time to spend a month in another country. With the odd exception of those connected to universities, actually. It was easy enough to arrange a 4 week vacation period at GMU as a normal Commonwealth of Virginia employee back in the 80s (of course, that also pretty much represents 2 years of saved vacation at one go).

9 JMCSF October 5, 2017 at 11:30 am

Have you ever summered in Spain? It’s underrated and more people should definitely try it.

10 Alistair October 5, 2017 at 11:44 am

To be fair to Americans, the US is so large that many varieties of holiday can be accommodated domestically and there are few foreign alternatives at short distances. For a European, the situation is almost reversed.

It would be interesting to see the distribution of holidays by distance from home from place visited. I’d guess the US distances compared well with the European ones. It would be a more meaningful comparison than the whole “passport percentage” thing.

11 Ben October 5, 2017 at 1:10 pm

One of my main turn-offs from living in the US a few years back was the lack of guaranteed vacation time.

Totally crazy, it’s an example of a country so wrapped up in what’s best for GDP while failing to appreciate a citizen’s non-economic quality of life.

12 dan1111 October 6, 2017 at 7:54 am

As an American moving to Britain, at first I had no idea what to do with >30 days vacation per year.

I figured it out, and it would be hard to go back.

13 Slocum October 5, 2017 at 1:03 pm

My take — skip the classroom instruction. It sucks. In my experience, if you want basic functionality (in a European language, anyway), 10 hours of Michel Thomas Method is worth more than a couple of years of high-school or university foreign language classes. Also — go to France (or Montreal, at least).

14 Kevin P October 5, 2017 at 8:20 am

Most useful in China? Depends if they come with a built in VPN.

15 ChrisA October 5, 2017 at 8:21 am

I wonder if this will increase the dominance of English? A near perfect audio translation will never be as good as perfect fluency, you will always struggle to follow subtle details from such a translation. English speakers are rarely going to be in the position where this is important (because most business people can speak english) so lessening their incentive to learn a different language. However non-English speakers will now be even more likely to be missing these subtle details and so will be forced to learn english as they struggle with the not quite perfect online translation.

16 wiki October 5, 2017 at 9:56 am

It is already happening abroad. One of the things that globalization is bringing to the non-English speaking world is a two tiered class structure where the elites are fluent in English, are likely to have the human capital to work in the “international” labor market abroad, and have full access to international media (or at least a big slice of it) and social apps through their fluency and comfort with English. The best schools even in their home country will have far more English and be more English based than the norm.

The rest — whether poor or comfortable — will be linguistic hoi poloi heavily confined to the sphere of the domestic economy and neighboring regions that speak similar languages. Whether this is a major or minor headache depends on one’s wealth and the relative importance of one’s home country native language and cultural influence. For countries with very linguistically distinct languages the social penalty in terms of a glass ceiling for the non-fluent will be quite severe.

17 JWatts October 5, 2017 at 10:23 am

“is a two tiered class structure where the elites are fluent in English … he best schools even in their home country will have far more English and be more English based than the norm. …The rest — whether poor or comfortable — will be linguistic hoi poloi heavily confined to the sphere of the domestic economy and neighboring regions that speak similar languages. ”

I was recently talking to a couple of younger French Canadians and asked them about learning English. They both said they became fluent in English from playing video games. For online team play, everyone has to understand each other and be able to comprehend and react quickly. It’s quite possible that video games are increasing English fluency among the non-elites to a far greater degree than we’ve ever seen in the past.

18 wiki October 5, 2017 at 10:43 am

Those are trivial numbers in a rich country surrounded by English speaking nationals.

Consider how little the average Japanese, Taiwanese, or Korean speaks English despite the emphasis on English in public school for decades and their stated desire to learn English. English outside of Moscow and Petersburg is very limited in Russia. Ditto for China once past the developed East. Even India for that matter is divided by fluency in English and familiarity with international standards. Filipinos can enjoy American movies without subtitles, but fluency still separates the elite and well-traveled from the run of the mill.

19 JWatts October 5, 2017 at 11:13 am

“Those are trivial numbers in a rich country surrounded by English speaking nationals.”

I think you underestimate the number of young adults playing video games. The numbers aren’t trivial. Among American children 91% play video games.
https://www.digitaltrends.com/computing/91-percent-of-kids-play-video-games-says-study/

” surrounded by English speaking nationals.””

Here you have a point. Clearly French Canadians are exposed to far more English than your average non-English speaker. However, the root point stands, global social media communications (video games being a highly underrated form) are exposing a lot of non-English speakers to the language in a far greater manner than the past.

“speaks English despite the emphasis on English in public school for decades ”

Most people learning a foreign language in school, never use it. The social media landscape is a positive feed back loop for foreign language fluency.

20 Slocum October 5, 2017 at 8:21 am

I’m skeptical about how well this will work in real-world, noisy, man-on-the-street situations. As a linked article points out, Google’s translation struggles with Chinese written text:

https://qz.com/795446/quartz-has-tested-googles-goog-new-ai-powered-translator-and-it-needs-to-learn-more-chinese/

Add in the extra errors from speech recognition before translation, and the potential for comedy seems high. But perhaps if the speakers are careful to use short, simple sentences and speak slowly and clearly, this will work well enough to be useful. And that aside, there’s no technical reason you’d need $159 wireless earbuds to do this vs $3 wired earbuds connected to your smart phone.

21 dan1111 October 5, 2017 at 8:31 am

I don’t know how much you use voice recognition, but it has come a long way in the last few years. The voice recognition on my (Android) phone works very well. It even performs surprisingly well in noisy situations like driving my car with music on.

I also find Google’s translation and text-to-speech capabilities to be quite impressive. (I’m not using it to translate Chinese, though according to your link it performed great on casual conversation, but had trouble with academic writing and poetry) This is really just connecting the different services up.

22 Todd Kreider October 5, 2017 at 12:24 pm

There was a major jump in Japanese to English interpretation in 2015. Westen Japanese speakers may laugh at the quality, (Japanese is much further from English grammatically than Chinese) but it wasn’t close to that level in 2013. Then last year Google added a neural net system and there was a clear jump in Google Translate despite errors. But the interpretaion (audio) is just using GT, and I’d expect another jump in a year or two.

There was a small army of translators and interpreters who voluteered long hours at the 1998 Winter Olympics in Nagano but those numbers will be heavily reduced for the 2020 Summer Olympics in Tokyo.

23 Petar October 6, 2017 at 6:13 am

+1

The earlier Google Translate results were how I realized that Japanese is more distant to Indo-European languages than Chinese. And then came the huge improvement with the neural net for Japanese, so much so that I started using GT for Facebook and Twitter translation from Japanese to English, since the GT quality became so much better than whatever those platforms used as native translation services.

24 Slocum October 5, 2017 at 12:45 pm

“This is really just connecting the different services up.”

Yes. But any errors in voice recognition will have the potential to throw a wrench into the translation — unlike humans, I doubt the translation algorithms are going to be able to handle guessing what the speaker probably meant to say when it encounters a word that doesn’t make sense.

25 dan1111 October 6, 2017 at 7:58 am

Yes, using the two systems together does compound errors. But as the error rates get low enough, it will matter less. Also, conversation allows the back-and-forth to work around this.

Also, I’m not sure how the system works, but it would be logical to show the voice recognition result on your phone, so that you could check if the text was right (and correct it) in the event that the hearer is confused.

26 rob October 5, 2017 at 8:25 am

I doubt it (the title)
A- All google services are blocked by the great firewall in China
B- While machine translations work pretty well for similar languages, Chinese is light years different and almost all translations coming out of google/baidu/bing are either laughable or misleading. I agree with the general point about platform monopolies though but only because you have several competing in parallel.

27 rayward October 5, 2017 at 8:29 am

A different perspective on Google gadgets: “But one nagging question lingered for Google, which makes nearly all of its money from selling online advertisements: Is it finally serious about making devices?” https://www.nytimes.com/2017/10/04/technology/google-pixel-hardware.html I say follow the money, and there’s the answer.

28 Stuart October 5, 2017 at 12:11 pm

Google is still driven by founders, not managers. That means, like Facebook, it matters more what founders want to do with their billions than any need to make the money machine run harder.

29 Ben October 5, 2017 at 8:51 am

Don’t use singular events/products to disprove or prove certain macroeconomic trends. I expect a highly respected economist to know that. If you do that you can prove absolutely anything about the economy.

Anyway, Google doesn’t have anything close to a monopoly on hardware, it’s an extremely competitive industry.

30 Careless October 5, 2017 at 4:16 pm

No one is claiming google has a monopoly on earbuds

31 Dots October 5, 2017 at 9:12 am

Peter thiel quips that they R&d in part to confuse questions about monopoly in their profit centers, so antitrust can b credited with matchmaking geniuses with tech problems

32 A B October 5, 2017 at 1:49 pm

Not just a quip. Companies such as Google, Facebook, and Amazon try to create a technological desert around it. For example, it will now be impossible for a startup to make enough money from auto-translation to have the funds to venture into, say, search.

33 JCC October 5, 2017 at 9:15 am

It’s funny how economists used to come up with theories that platform monopolies would stifle innovation

Keynes once said: When facts change, I change my mind

34 prior_test3 October 5, 2017 at 9:52 am

It is also funny to think how those economists just might have grown up with AT&T as their one and only phone company, who just happened to be the company that decided which equipment you were allowed to attach to AT&T’s network.

35 Haakon Birkeland October 5, 2017 at 12:24 pm

There are plenty of competitors in the mobile phone industry. Let’s see what happens if Apple buys Samsung before we start talking about platform monopolies and innovation.

36 Mark Thorson October 5, 2017 at 9:43 am

It wasn’t enough that they track you and read your mail. Now they will also eavesdrop on your conversations.

37 prior_test3 October 5, 2017 at 9:50 am

But at least they aren’t stifling innovation, right?

38 Rob Terrin October 5, 2017 at 10:38 am

I do not see this a net positive development. Language learning is one of the best ways to improve your mind, stave off dementia, connect with people, broaden your perspective, develop empathy, increase creative capabilities, etc.

There is a mountain of evidence that language learning (at any age) has individual benefits and the externalities are obvious. Here’s one overview paper:

“Bilingualism has been associated with improved metalinguistic awareness (the ability to recognize language as a system that can be manipulated and explored), as well as with better memory, visual-spatial skills, and even creativity”
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3583091/

39 Haakon Birkeland October 5, 2017 at 1:02 pm

The general trajectory of Silicon Valley is to destroy any love of wisdom and knowledge that might still be present in humanity.

There’s still a very good chance that the internet in general is an epistemic net negative. Pizzagate and the anti-vaccine movement show that access to information is useless without an education in critical analysis. Perhaps one-on-one instruction where the wise and experienced directly instruct the next generation will always be better than a screen covered with advertisements.

Just look what information technology has done to science. Reproducibility has been replaced with regression analysis, “Nullius in verba” replaced with bashing someone’s brains in with a colorful chart.

40 dan1111 October 6, 2017 at 8:12 am

-1

There are some bad human tendencies, and technology hasn’t made them go away. And sometimes people use technology for bad things.

However, please give some evidence that technology is causing any of this. There was loads of ignorance before the internet existed. And also, it was a lot harder for those who wanted to information to get it.

41 Haakon Birkeland October 6, 2017 at 2:33 pm

Evidence? What kind of evidence? A link to an article filled with charts and spreadsheets from someone with the proper credentials? Is that what you consider evidence?

This argument does not depend on evidence. Furthermore, the argument you are trying to have has nothing at all to do with what I said.

I never said that there was no ignorance before the internet. What I said was that the internet, and especially the commercial and ad-subsidized internet, is not some cure-all for the plight of ignorance.

We have successfully waged war on ignorance in the past, hence my quote from the Royal Society. You speak of wanting evidence but you don’t really know what that word means. You speak of information but in the same manner as Eric Schmidt, “that more information is better, even if it’s wrong.” What is information without the ability to critically access the source?

Care to take up the argument on his behalf or would you concede that drowning in a sea of bad information might be worse than no information at all? Is it not as simple as Mr. Schmidt and yourself seem to be making the issue and is it perhaps more complex and nuanced than either of you might want to admit?

You’ll also notice that I am asking questions, not speaking in definitive proclamations. I do not claim to have the answers. I claim that neither you nor Eric Schmidt have the answers.

42 Todd Kreider October 5, 2017 at 1:34 pm

Papers on bilingualism apply to children who were using both languages at very young ages – under 2 or 3 years old. These bilinguals supposedly have both languages bound in one area of the brain. Those English speakers who start to study German at 14 and become very fluent in their 20s are still not bilingual in the sense many studies use the term. They have language centers where English and German are mostly separate but with *some* overlap instead of almost complete overlap.

(Someone correct me if I’m wrong but Readers Digest hasn’t let me down in the past.)

43 dan1111 October 6, 2017 at 8:06 am

-1

The lack of ability to communicate with people who speak other languages is a huge barrier with huge costs. Even if you are bilingual or trilingual, there are billions of people in the world that you can’t communicate with. Overcoming this with technology would be a massive step forward for humanity.

If being bilingual really is hugely beneficial, people can still choose to do that. I’m skeptical of that claim, though. The research I’m aware of (including the one you link to) shows specific cognitive benefits but doesn’t consider systemic benefit. It’s quite likely that bilingual people develop certain skills, abilities, and knowledge more, but at the expense of other skills, abilities, and knowledge.

44 Anonymous October 5, 2017 at 11:26 am

Monopoly? If we are using the “phone” perspective it is a duopoly, where Apple has most of the profit and Google has most of the market share. That has driven intense competition.

45 Engineer October 5, 2017 at 12:07 pm

Will this facilitate the demise of bilingual education in the US?

Or encourage linguistic Balkanization?

There may be some financial savings from eliminating the requirement for human translators in legal and medical environments, if the quality is deemed adequate. Acceptance in those environments may be one benchmark to evaluate the functionality achieved.

46 Scott Mauldin October 5, 2017 at 12:48 pm

This would facilitate the growth of intermediate-size languages like Swahili, Polish and Urdu that are not global languages but are large enough markets to get service on these devices, as speakers of those languages would not have to learn any others; it will further the extinction of tribal and local languages and dialects that have too small of a market to receive translation service on these “Babelfish” devices in the near future.

47 Mark Thorson October 5, 2017 at 12:55 pm

It’s too soon to say they won’t attempt to support every language for which speakers are available. If the process is automated, there might not be a cost to that, so why not?

48 mpowell October 5, 2017 at 2:45 pm

It’s all based on machine learning. You need huge data inputs for training. Most likely, the data does not exist for smaller languages.

49 Scott Mauldin October 5, 2017 at 4:18 pm

There are many small tribal languages and particularly dialects that don’t have standard written forms, let alone bodies of literature to work from. Plus what about non-phonetic writing systems like Sinic languages, in which dozens of spoken languages – some linguistically unrelated – share the same corpus scriptorum? You’d have to do lots and lots of recordings before this became viable for them.

50 Careless October 5, 2017 at 9:37 pm

Well, sure, but while that sucks for them, 99% of the world is still happy

51 Scott Mauldin October 5, 2017 at 12:53 pm

“legal and medical environments” – these are at the opposite end of the spectrum. Medical translation deals in scientific terms and measurements with far less nuance than Legal translation, in which a slightly ambiguous shade of meaning can completely alter the acceptability of treaties or trade deals.

52 Engineer October 5, 2017 at 2:03 pm

I was thinking ER type “where does it hurt” md–patient dialog, rather than scientific discussions.

53 Scott Mauldin October 5, 2017 at 2:53 pm

I was thinking of that as well, but where is the nuance in that? Are there ambiguous lingual variations in terms like “here” and “kidney”?

54 Engineer October 5, 2017 at 2:05 pm

And for legal I was thinking police , court, public defender, etc type work.

55 apoptosis October 5, 2017 at 12:10 pm

re: “economists used to come up with theories that platform monopolies would stifle innovation…”

AT&T perhaps being a prime example?

56 Mark Bahner October 5, 2017 at 1:09 pm

“This will be most useful in China.”

It will be most useful in Catalonia. 🙂

57 Larry Summers the Great October 5, 2017 at 1:36 pm

I *told* you all that studying foreign languages in school wasn’t a good use of time!

58 Taeyoung October 5, 2017 at 1:51 pm

I assume “this will be most useful in China” because it will enable the the Party to spy on conversations with foreigners in realtime without having to go to the trouble of placing bugs and agents — the phone will be the bug and Google will be their agent.

That seems pretty useful, yes.

59 Arnold Layne October 6, 2017 at 5:55 am

Everyone is assuming these will work as intended. They forget that Google’s history in non-search related consumer ventures, especially when it comes to hardware, has been one big failure after another. The buggy Chromecast is exhibit A.

Prediction: Their translator buds, like their driverless cars, will work in select, super specific and easy situations, and fail miserably when challenged by real-world scenarios (e.g., background noise, poor wifi, slang, etc.)

60 dan1111 October 6, 2017 at 8:58 am

You are ignoring that Google has demonstrable success in the technologies underpinning this specific product, particularly translation.

And their self-driving cars, still in development, have yet to be proven a failure.

61 Otis Pyushiram Gulati October 6, 2017 at 3:52 pm

China? I’d love to have this for every time I step into an NYC cab

Comments on this entry are closed.

Previous post:

Next post: