https://www.duolingo.com/Ann400400

How will the next revolution in Artificial Intelligence change the language learning landscape?

Hi everyone, it is said that a big revolution in technology is underway.

I was wondering what will happen if some predictions about the next advancements in Artificial Intelligence regarding languages materalize?

Particularly, some people say that chances are we will have a very effective universal translator app in 15 years. How likely is that? And how it would change language learning?

Would websites like Duolingo cease to exist because of the lack of motivation of learning a new language when you can speak it through a device anyways?

Will language learning be so effective and cheap that people will be more keen to learn them? I mean I guess artificial intelligence will make language learning very easy and cheap.

What do you think about the whole thing?

1 year ago

31 Comments


https://www.duolingo.com/Mr_Eyl
Mr_Eyl
  • 13
  • 12
  • 8
  • 7
  • 5

Decent language AI won't stop people from learning languages any more than the microwave stopped them from learning to cook. People with motivation will stay motivated. People without it will succumb to the easy option, as always.

As for a personal opinion on the whole thing, on one hand I'll be extremely happy to have an AI to talk to in any language I'm learning. On the other, I'll probably still be of the opinion that all strong AI needs to be prohibited. :)

1 year ago

https://www.duolingo.com/garpike
garpike
  • 25
  • 24
  • 22
  • 20
  • 19
  • 19
  • 18
  • 17
  • 17
  • 17
  • 17
  • 16
  • 16
  • 15
  • 15
  • 14
  • 14
  • 14
  • 14
  • 13
  • 13
  • 12
  • 11
  • 11
  • 11
  • 10
  • 10
  • 9
  • 8
  • 8
  • 6
  • 6
  • 6
  • 5
  • 5
  • 1167

Brain-computer interfaces will probably make the time and effort spent learning languages considerably less worthwhile, much as learning to cook would be if the alternative were thinking meals onto the table. If all that will be needed to communicate in a foreign language is thinking the thought to be expressed to a near-flawless translation engine, which almost instantly feeds back into the brain in the form of motor signals that allow one to pronounce it perfectly, I doubt many people will want to spend hundreds of hours learning the old-fashioned way.

Of course, this technology is not imminent, but it could quite conceivably happen within a couple of decades.

1 year ago

https://www.duolingo.com/J1ng3rT41ls

lol, my exact thoughts!

1 year ago

https://www.duolingo.com/matfran2001
matfran2001
  • 25
  • 25
  • 25
  • 25
  • 25
  • 23
  • 22
  • 22
  • 21
  • 20
  • 18
  • 14
  • 14
  • 13
  • 11
  • 7
  • 7
  • 7
  • 6
  • 6
  • 4
  • 2
  • 23

I will keep on learning languages because I just enjoy it. And I guess there are many more people like me out there.

With respect to A.I., yes, I really think they will be at an almost native level (with respect to language translation) in the next five years at most.

In the last months there have been some examples of a software passing a "weak-Turing-Test" (that is, around 80% of "normal" people that chatted with it for a long time, thought it was a human), though it is still currently not good enough to fool experts.

But given the rate of improvement (and the rate of "rate of improvement", which is the key) and the huge amount of $$$$ that is being thrown into it, I guess the first AI that passes a "really complete Turing Test" (that is, fooling even experts in AI and computer scientists) will be sooner than 2025 or 2030.

By that time, probably up to 80% of current jobs will be done by machines (AI machines/robots) and humans driving cars, taxis, trucks, ships, planes.... will be probably forbidden.

In most countries they will probably launch a "Basic Unconditional Income" program, which will be the most logical and sensible thing to do at that moment.

I saw (and many others did too) all of this coming, since the late 90s (anyone that was aware of the rate of "rate of progress", that is, the acceleration of improvements, in so many different but related key technologies, could have forseen all this, and even John Von Neumann could see it coming more than sixty years ago).

I didn't know at that time if it would come during the 2010s, the 2020s, the 2030s or later, but now, given the data we now have, I am quite sure that all this is coming very, very soon (during the next 10-15 years).

1 year ago

https://www.duolingo.com/soedori
soedori
  • 21
  • 10
  • 9
  • 8
  • 6
  • 6
  • 4
  • 2

This is key and on point

1 year ago

https://www.duolingo.com/EpicPowerHero

I have my doubts as to whether there will be an AI that can speak a language as fluently as a native speaker, because of all the subtlety of language. Plus I mean, I'm sure that device won't be cheap, at least at first, and it could have errors or break; machines always have errors and break. Inventors should still try though, because who knows what might work.

1 year ago

https://www.duolingo.com/J1ng3rT41ls

The devise has been out for several years already (from what i know) , and it apparently does very well, and is becoming cheaper and into an app you can download on your phone.

1 year ago

https://www.duolingo.com/elizadeux
elizadeux
  • 18
  • 16
  • 15
  • 9
  • 6
  • 4

Machine translation is at this point fairly rough and unreliable. We shall see what will happen when an app or device capable of natural language translation becomes available.

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

Have you tried the most recent versions of google translate for Korean? It often gives near perfect translations. They haven't rolled it out for all languages though, so you may not have encountered it yet.

1 year ago

https://www.duolingo.com/soedori
soedori
  • 21
  • 10
  • 9
  • 8
  • 6
  • 6
  • 4
  • 2

I was really surprised, though I did manage to exploit a couple of blatant errors

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

For sure. I mostly use it when I get some official Korean emails and I don't have time to translate it by hand. It does a surprising good job at translating official stuff, as it focuses on longer patterns now rather than words only. But I wouldn't trust it yet to submit a grant application written in English, translated into Korean.

1 year ago

https://www.duolingo.com/Mobetty
Mobetty
  • 25
  • 25
  • 25
  • 23
  • 21
  • 12
  • 8
  • 7
  • 5
  • 4

AI might someday get good at translation. But humans who compose original music, artists who draw to express their interior world, speakers who have thoughts and words that have never been said or written will always seek connection with others.

Human languages are a barrier for our race (that's the human race, just in case) to understand one another. AI won't get us all the way back to the Tower of Babbel, but effective language learning, AI chatbots, and accurate language translation tools will bridge some, but not all, of our communication needs with our fellow beings.

As for the would-be computational linguist, man is NOT here just to fill the Corpus (the written, transcribed, and tagged sample of human-generated language)... we created our languages to speak and understand one another. Corpora only serve for us to analyze what humans mean and say in our verbal transactions. At least that's my worldview.

1 year ago

https://www.duolingo.com/DragonPolyglot
DragonPolyglot
  • 25
  • 19
  • 17
  • 14
  • 13
  • 12
  • 11
  • 10
  • 10
  • 10
  • 9
  • 9
  • 9
  • 9
  • 9
  • 8
  • 8
  • 7
  • 6
  • 6
  • 6
  • 5
  • 2
  • 2
  • 82

My dad works with computers and he told me that machine learning is getting exponentially faster and better, which includes language learning. He's even told me that because of this I shouldn't need to learn a language. But then I told him "but what about the corpus that the computers use to get the information? Who will contribute to those?"

Also it's fun to know how words work and not have to type it into a machine! I think people will attempt to learn new languages even if computers get as good as the best translators for many reasons.

1 year ago

https://www.duolingo.com/Ann400400

Thanks for the answer. It's very interesting.

By the way what are you referring to with the word "corpus" I didn't understand that part.

1 year ago

https://www.duolingo.com/DragonPolyglot
DragonPolyglot
  • 25
  • 19
  • 17
  • 14
  • 13
  • 12
  • 11
  • 10
  • 10
  • 10
  • 9
  • 9
  • 9
  • 9
  • 9
  • 8
  • 8
  • 7
  • 6
  • 6
  • 6
  • 5
  • 2
  • 2
  • 82

Corpus linguistics is a field of study that analyzes documents to study words. A corpus contains many documents linguists use to make databases and research material.

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

No linguistics required for Google translate. Just machine learning and data mining. A.I. has mostly moved away from hard coding rules and logical statements.

1 year ago

https://www.duolingo.com/DragonPolyglot
DragonPolyglot
  • 25
  • 19
  • 17
  • 14
  • 13
  • 12
  • 11
  • 10
  • 10
  • 10
  • 9
  • 9
  • 9
  • 9
  • 9
  • 8
  • 8
  • 7
  • 6
  • 6
  • 6
  • 5
  • 2
  • 2
  • 82

It's not used in many translators. It's used to make things like dictionaries and graphs, and it aids in research for specific topics like adjectives.

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

Ok, I see what you mean now :)

1 year ago

https://www.duolingo.com/soedori
soedori
  • 21
  • 10
  • 9
  • 8
  • 6
  • 6
  • 4
  • 2

but what about the corpus that the computers use to get the information? Who will contribute to those

Big data+neural networks+deep learning

Boom.

No, not joking. How do you think AlphaGo learned Go? Do you think that she suddenly knew how to play it? No.

She was fed game data and numerous simulations to make her "learn it". Like so, deep learning could be used in order to "teach" the machine a language. I'm personally going into the field, and it's extremely exciting where we're headed.

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

If I may ask, are you a grad student in Korea? Our university here (UOS) is applying for major funding to develop a machine learning program. Seems like Korea is definitely hopping on to that wagon...

1 year ago

https://www.duolingo.com/soedori
soedori
  • 21
  • 10
  • 9
  • 8
  • 6
  • 6
  • 4
  • 2

Alpha Go sort of invigorated us (by sort of, it was big). I'm not a grad student (yet), but it's great hearing you go to UoS! My father is an alumni there :). I'm actually a student in the United States. Btw if you would like, please apply for our course ;)

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

Yes, Alpha Go was a pretty big thing. I remember following one of the games on a stream while in the US. I'm a postdoc at UoS now. I plan on applying for the course when it comes out. If you're talking about alpha testing, I did submit one or two applications after some of the announcements. But I guessed there were too many applications... Should I send a new one or do you still have them in your records?

1 year ago

https://www.duolingo.com/soedori
soedori
  • 21
  • 10
  • 9
  • 8
  • 6
  • 6
  • 4
  • 2

Could you send a full application (contributor, not alphatester)? I just want to see where you fall on right now :)

After all, the course is technically finished, but alternate answers need to be added!

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

Sorry, had to go out for a second.

Maybe some confusion there. I'm not a fluent Korean speaker at all. I am in the process of learning Korean. I finished the reverse tree. Hopefully passing TOPIK 3 soon. You can see my level through one of my writings on my blog https://myjourneytokoreanfluency.wordpress.com/2017/08/09/059-%EC%93%B0%EA%B8%B0-%EC%98%B7%EC%9D%B4-%EB%82%A0%EA%B0%9C%EB%8B%A4/

After that, If you still think it would be useful, I'd be happy to apply tonight after work. No hard feelings if you'd reject me based on my application :)

I could provide alternate answers in English fairly easily, not so in Korean.

1 year ago

https://www.duolingo.com/soedori
soedori
  • 21
  • 10
  • 9
  • 8
  • 6
  • 6
  • 4
  • 2

Wow, not too bad! Work on it a bit and I am sure you can get near TOPIK 5 easily ;)

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

Thanks. So should I still submit an application?

1 year ago

https://www.duolingo.com/J1ng3rT41ls

:O

1 year ago

https://www.duolingo.com/andypinguwi
andypinguwi
  • 18
  • 16
  • 10
  • 9
  • 4

This question seriously keeps me up at night! I run around in circles thinking about this but the only thing I can agree on (with myself lol) is that I do think it will halt the learning of "less popular" languages. Especially with the increasingly common use of English. But AI will always be an intermediator, unless it could be installed into our brains somehow, so I think people with a real enjoyment (or people that really need to use a language) will continue to learn.

1 year ago

https://www.duolingo.com/Garm31

AI, no matter how advanced, cannot translate. Language is a function of the soul, and no machine can comprehend a soul.

So no, you will have to learn languages to get reasonable translations.

1 year ago

https://www.duolingo.com/nleconte
nleconte
  • 16
  • 14
  • 12
  • 10
  • 7
  • 2

Depends on what you understand by reasonable translations.

Evidence shows that for functional translations (for daily functional use), machines are already very close to doing a very good job. Near-human like.

If you're thinking about poetry, as you say, to touch the soul, then, I agree, AI might still have a long way to go.

Then again, writing a good poem is something very subjective. It is sometimes a very simple succession of words where the reader needs to find the deeper meaning. Some specifically tuned machines might be able to trick you into believing their limerick has been written by a human... by combining the millions of existing poems into new ones.

Not that I necessarily would be happy with such evolution, from an artistic point of view.

1 year ago

https://www.duolingo.com/J1ng3rT41ls

YIKES!! I have no idea, but I hope Duo will not "cease to exist t because of the lack of motivation of learning a new language when you can speak it through a device anyways." I know about that, but what about those that do not have that device or can not afford it?? I hope it will always stay cheep. PLEASE, NO MATTER WHAT HAPPENS, KEEP DUOLINGO UP!!! :) ;)

1 year ago
Learn a language in just 5 minutes a day. For free.

Learn a language in just 5 minutes a day. For free.