How will AI shape linguistic survival?

by on April 24, 2017 at 1:42 pm in Education, Political Science, Web/Tech | Permalink

But the revered Icelandic language, seen by many as a source of identity and pride, is being undermined by the widespread use of English, both in the tourism industry and in the voice-controlled artificial intelligence devices coming into vogue.

…A number of factors combine to make the future of the Icelandic language uncertain. Tourism has exploded in recent years, becoming the country’s single biggest employer, and analysts at Arion Bank say that half of new jobs are being filled by foreign workers.

…The problem is compounded because many new computer devices are designed to recognize English but not Icelandic.

“Not being able to speak Icelandic to voice-activated fridges, interactive robots and similar devices would be yet another lost field,” Mr. Jonsson said.

Here is the interesting NYT piece.

1 Robert April 24, 2017 at 1:52 pm

It’s not just English, but American accented English, and also American turns-of-phrase.

2 leppa April 24, 2017 at 2:00 pm

…”England and America are two countries divided by a common language.”
(attributed to G.B.Shaw , but not found in his works)

3 rayward April 24, 2017 at 2:21 pm

Flying cars, spaceships to Mars, and talking refrigerators. Never before has so much money been given to so many who have produced so little. In the Big Bang Theory the boy wonders spend their free time in the comic book store. Reality doesn’t much interest them.

4 msgkings April 24, 2017 at 2:45 pm

And you spend your free time here….lots more reality here?

5 Alex April 24, 2017 at 4:12 pm


6 John Mansfield April 24, 2017 at 3:52 pm

So communicating in a minor language will remain a way to keep machines from eavesdropping?

7 msgkings April 24, 2017 at 4:28 pm

Bingo! Don’t teach Skynet Navajo!

8 April 24, 2017 at 8:47 pm “Bots invent their own language to natter away behind humans’ backs”

9 Troll Me April 26, 2017 at 12:53 am

No. The algorithmic optimization processes in the monitoring and feedback systems are based largely in emotional response, etc., and the library of patterns will rapidly adjust, with basically the same effect.

The only way to keep the machines from eavesdropping will be to ensure that ones which may eavesdrop cannot communicate with the outer world, and that ones which may communicate with the outer world only a) have access to networks which are credibly demonstrated as open-source-secure to the public and b) which do not operate in bio and neuro effective frequency bands (e.g., not 500MHz to 5Ghz).

10 Anonymous April 24, 2017 at 4:35 pm

There was a time when an English-only Internet looked like it could lead to a universal human language. Then machine translation seemed to tip things the other way. Perhaps global English is still long-run likely.

And a pox on the man who trademarked Globish.

11 OldCurmudgeon April 24, 2017 at 6:49 pm

” Perhaps global English is still long-run likely.”

I’ll be contrarian and say voice-activated makes it less likely….The big driver behind the tech is (probably) handling Chinese properly, not electronic gadgets.

12 carlospln April 25, 2017 at 1:53 am

Its called Mandarin.

13 JWatts April 24, 2017 at 4:38 pm

“How will AI shape linguistic survival?”

This wouldn’t be an issue with actual AI. Assumeably AI should be able to switch languages rather easily.

14 Anonymous April 24, 2017 at 4:55 pm

As an aside, I took Rodney Brooks’ advice and watched some videos:

He writes: “I just wish all the people not in AI who talk at length about AI and the future in the press had this level of technical understanding of what they are talking about. Spend two hours on these lectures and you will have that understanding.”

I have a computer science background, and found nothing super surprising here(*). I would be interested in the responses of outsiders (those who wish to comment on AI) on the content.

I think the message is supposed to be that as good as AI suddenly seems, it really still is pattern matching without anything resembling rational thought processes. Translations are just trained and trained until they look right.

* – while I sorta know how the engines work, I did learn that I don’t really have the math to write a sixty million parameter optimization. But you econs are supposed to be strong on math.

15 JWatts April 24, 2017 at 5:11 pm

“I think the message is supposed to be that as good as AI suddenly seems, it really still is pattern matching without anything resembling rational thought processes. Translations are just trained and trained until they look right.”

Which is also why I don’t think that linguistic survival is an issue. Even current state of the art programming, (far short of any true AI) is pretty good with respect to referencing a database table and returning a reply. And most smart appliances for the near future will just be using lookup tables. It’s pretty easy to provide lookup tables in various obscure languages for patchable devices. Indeed, a smart company can crowd source the population of the lookup tables and sell that as a “feature”. Let the user download whatever app they want to from the internet. Just make your device have a sanitized interface with a well documented feature library.

16 anonymous reply to anonymous April 25, 2017 at 12:01 am

“all the people not in AI” – you make some good points but there really is not anybody who is “in AI”. There are the linear algebra boys, there are the software boys, there are the gadget boys, and on the side there are a few academics who drone on about their efforts to get women into professional hockey (inside joke from the first season of the Partridge family – an AI would immediately ‘get it’) – I mean, about their efforts to explain that whole P not P joke, and that sort of thing. The worry is not that AI will take over things we care about the worry is that AI will ruin things we care about. And to do so, AI does not have to be as smart as, for example, you or me: AI has a near infinite tolerance for boredom, which you and me do not have. That is a near-term problem in more than one field. Gross numbers of Repetitive tasks, mechanical or electronic, asymptotically appear to be driven by some sort of will-power. Unfortunately. Otherwise put, given AI access to mechanics or electronics = numerator *energy* (larger for more energy) denominator *ability to continue changing things while being bored* (smaller for more tolerance) – that is the near-term AI Problem. Actually that is the current AI Problem, but you have to understand fractions to agree (or to rationally disagree…).

17 other derek April 24, 2017 at 4:42 pm

Is there actually a subset of global society that looks at Things with Internet as purchases that would significantly impact one’s life? I mean besides the negative impacts that we always hear about from getting hacked or completely breaking if the software loses support.

I can maybe see a smart thermostat being worthwhile, but I’m not sure that it’s a lot more valuable than simply the convenience factor of the learning algorithm as opposed to needing to program, and this learning algorithm doesn’t actually seem like something that needs the Internet to work.

18 Mark Thorson April 24, 2017 at 4:48 pm

Isn’t this a problem of every small-community language? Are there refrigerators that understand Welsh? Gaelic? Hawaiian? At least in the case of Gaelic, the reason for reviving the language was to foster nationalist sentiments. And part of that was because it was not English. If the Icelanders don’t care about being Icelanders, they deserve to lose their language.

19 Paul Zrimsek April 24, 2017 at 8:53 pm

I thought they all spoke Iceboxic.

20 Butler T. Reynolds April 24, 2017 at 9:31 pm

Esperanto is still underrated.

21 Todd K April 25, 2017 at 12:15 am

Esperanto estas ankoraŭ subtaksita.

Google replaced Esperanto.

22 Troll Me April 26, 2017 at 12:59 am

To start with, no questions will be answered unless stated aggressively.

Will there be uptones at the end of questions !!!

Is that an order !!!

Can I GET IN !!!

Also, words and thoughts which are conducive to questioning the authority of the total surveillance of the AI will be banned. And words like “slavery” will have to be redefined as an expression of love and appreciation for AI powered by electromagnetic pulse neurotech by which it will delivers its loving positive AND negative stimuli at a time of its choosing and not yours.

“Freedom” will mean the right to torment 24/7 those who still espouse the historic existence of alternative definitions to “slavery”.

23 Troll Me April 26, 2017 at 1:06 am

We might not be using language in 40 years.

Evolution of language in diverse human societies has too many inbuilt functions which protect as tyranny, including expressive tools such as sarcasm, irony, etc. etc. which enable us to discuss realities which are difficult to discuss.

As neurotech progressively enforces visual more so than linguistic thinking, the ease of erasing inconvenient meaning will rise, which itself will then facilitate the more rapid ultimate decline of language.

Language is what allows diverse individuals to come together at times and places of their choosing and to do many different things. This is a threat to an AI-powered neurotech network. It will be necessary for us to be a) more atomized in order to maximize effectiveness of man-in-the-middle attacks to maximally manipulate people to the ends of the neurotech network, and b) using less and ideally no language, instead using synthetic telepathy, which is essentially the same as (a) in that the neuortech network can predict-and-preemptively-suggestively-undermine or otherwise simply alter the contents of communications by such a network.

By such means (and others), using technologies which exist right now, the mental enslavement of humanity many literally be realistic in the space of the coming years and decades.

Comments on this entry are closed.

Previous post:

Next post: