[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/jp/ - Otaku Culture

View post   
  [?]In ghost mode, only threads with non-archived posts will be shown
File: 767 KB, 440x370, 1355976264851.gif [View same] [iqdb] [saucenao] [google]
10238424 No.10238424[DELETED]  [Reply] [Original]

When machine translators advance to the point of perfection will people stop learning other languages?

I know I will.

>> No.10238434

Linguist here.

The answer has been "next decade" for many decades.

Honestly I think it's impossible.

>> No.10238435

Most people will. But I don't think it's going to happen in our lifetime, so you should have said "I know I would".

>> No.10238437

100% accuracy in machine translation is impossible. Even expecting 50% is very naïve of you.

>> No.10238438

>>10238434
>>10238437
It may be possible in the future, when machines will be capable of human-like thoughts.

>> No.10238439

>machine translators advance to the point of perfection

Won't happen any soon, if ever. You'd pretty much need a strong AI for that.

>> No.10238450

Perfect output may be possible, but I'm not sure if people could make a perfect interpeter without inventing an AI with humanlike intelligence.
This is because language heavily relies on how our minds work. Things like context, exaggerations and comparisons, as well as puns, humor and other emotions behind words. These things don't make sense without an understanding of the world and knowledge of how it works, sounds and looks.

>> No.10238453

>>10238438
It never will, the environment plays a good deal in the evolution of language. You wouldn't get a better human, you'd just get a waste of a good computer.

Just fucking learn Japanese you lazy bum.

>> No.10238468

>When machine translators advance to the point of perfection will people stop learning other languages?
Yes, because we will all be wiped out by Skynet.

>> No.10238475
File: 315 KB, 1024x740, IBM_PC_5150.jpg [View same] [iqdb] [saucenao] [google]
10238475

>>10238434
>>10238435
>>10238437
>>10238439
>>10238450
>>10238453
>640K ought to be enough for anyone.

>> No.10238480

>>10238475
You're an idiot.

>> No.10238489

50 years+- seems plausible.

>> No.10238529
File: 63 KB, 692x530, janeway.jpg [View same] [iqdb] [saucenao] [google]
10238529

>>10238424
No, even with universal translators, people would learn new languages, it will show how much you care about the culture of your fellow alien neighbors

>> No.10238530

A hundred years or so at this pace.

Unless, I dunno, the current patent system gets abolished or something else happens so the rate of innovation in software accelerates again.

>> No.10238571

I think the waifu androids will come first.

>> No.10238582

You may not be aware of this, being a shutin neet and all, but generally people converse with each other face to face using words and stuff. You don't really have the option of using machine translation on the spoken word.

>> No.10238578

>>10238530
Yeah, because not being able to make a living out of your code will sure make people want to take on big projects like this as opposed to fart iPhone apps. Getting rid of the patent system would result in the most horrendous overdose of carbon-copy shit ever for a good fifteen years at least. Bootlegworld would suck.

>> No.10238580

>>10238571
Agree.
So yes, I will stop learning the shit and play with my waifu instead.

>> No.10238591

>>10238434
Sounds like you're not so unbiased.

>> No.10238607

>>10238582

Google Glasses and other AM form factors are right around the corner. You're an idiot akin to the people who screamed: 'NO YOU CAN'T RIDE A TRAIN, YOUR BLOOD WILL BOIL'

>> No.10238610

>>10238607
Yeah and then we're gonna stick fish in our ears to translate alien languages automatically

>> No.10238617

>>10238610

Great comeback. It'll never happen because you're too dumb to understand it. Gotcha, d00d.

>> No.10238620

>>10238610
Douglas Adams references are fun but don't really help making any point whatsoever
This is going to be the future at one point. If we're going to live to see this is a whole other story, though.

>> No.10238634

we already have calculators and computers that do all the math for us, but we still have to learn mathematics, the same will be applied to translators, and if you still think that is impossible to do a 100% translation you are an idiot

>> No.10238637

>>10238591
Not that guy, but it's not necessarily a "bias". There is such a thing as "knowledge", you know? If you tell a physicist you're going to throw a paper ball in America with your bare hand and it might hit someone in Japan with 2 tons of pure power, well, they're going to say "bullshit". And it would only be a bias if you were Superman.

>> No.10238641

>>10238617
>too dumb to understand it

It's funny because I'm a grad student in AI and can pretty much guarantee that I know more about these things than you.

>> No.10238646

>>10238641

WWOAH, A GRAD STUDENT IN AI

S-SUGEEEEEEH (i am making fun of you btw)

>> No.10238652

>>10238641

I'm Michael Abrash, cumstain, and I say you're wrong. Get out of my sight and don't let me catch you back here again, or I'll free this world of you.

>> No.10238662

>>10238646
>(i am making fun of you btw)
your words pierce into my soul, I am destroyed

I have been made fun of, I can never go on. To be made fun of by a shithead on /jp/. My life is over.

>> No.10238667

>>10238617
It will never happen because language is ever-changing, the capacity of AI will always be limited by their programmers and even if it could truly learn with a similar mindset, the environment in which it would learn would be different than a person's and so the experience would not be the same, vastly inferior in any form.

If you disagree, consider studying more about linguistics. There is a difference between "scientific spirit" and just saying a pile of mud can sing and anyone trying to tell you that is wrong is biased.

>> No.10238670
File: 65 KB, 534x584, ntmon.jpg [View same] [iqdb] [saucenao] [google]
10238670

>>10238662
Fascinating.

>> No.10238672

>>10238662

What the fuck did I just tell you?

I'm going to split your little nerd ass in half if you're not outside this board on the count of five. I am not playing with you and you are in real danger right now.

>> No.10238679

>>10238667

>the capacity of AI will always be limited by their programmers

lol what

you are retarded. yes let me just jot this down as a new law of the universe, right under 'every action has an equal and opposite reaction' - 'silicon computation can never exceed the dumb arbitrary barrier of a human mind'

>> No.10238680

>>10238672
>I'm going to split your little nerd ass in half
a-ahh, please be gentle~

>> No.10238690

>>10238641
If you really study AI, you should know that spoken word recognition is the smallest of problems here, one that is actually going to be solved in the predictable future.

>> No.10238689

>>10238634
We only have to learn maths because we can't actually use a calculator without knowing it

>> No.10238691

>>10238679
Except you're taking it out of context, moron. We're talking about teaching a machine to learn the same way we do, so of course it's going to be limited as fuck.

>> No.10238702

>>10238691
>We're talking about teaching a machine to learn the same way we do

It appears to me you were talking about teaching a machine to learn what we learn. That's a crucial difference.

>> No.10238699

>>10238691

uh no it won't

you are dumb and need to pay more attention in your AI classes kiddo

it doesn't matter how inefficient the 'brain algorithms' are if you can pile heaps and heaps of figurative gray matter onto them

and they're not inefficient given we have the equivalent of several supercomputers packed into a space the cubic size of 2-3 grapefruits

>> No.10238708

>>10238699
Oh. My. God.

You are seriously this retarded. I thought you were joking.

The issue is not with the processing speed but the process per se: it will be limited and flawed by design.

>> No.10238713

>>10238708

>The issue is not with the processing speed but the process per se: it will be limited and flawed by design.

this is literally gibberish

i don't think you actually know what you're talking about at all d00d

>> No.10238711

just buy a fuggen gook and make him translate everything for you

>> No.10238712

I don't care. I want to understand Japanese raw. If machine translation ever becomes 'perfect' and accessable for the masses, well then I'll be like a snob elitist hipster showing off my skills without machines.

>> No.10238714

>>10238702
Not only what, how. Which is both necessary for its "understanding" of language (a cornerstone of true translation as opposed to just picking words from a database is to understand context) and the ultimate limit to impose on a machine that cannot experience the world in the same way as we do. You'll be telling a blind kid about color: it's the "Mary the super-scientist" argument all over again.

>> No.10238745

>>10238424
We'll have loli sex bots way before that, so who the fuck cares?
I'd learn languages even more vehemently so I could look down on people.

>> No.10238754

>>10238714
No, it's just what. It's always just what. It doesn't matter how it works, it doesn't matter if it "understands" what it does, only results matter. You're never trying to emulate human intelligence, you're trying to make a machine perform a task. If pulling things from database works, nothing else matters.

Of course it's possible that in some specific cases pulling things from database will not work, but this can only be determined experimentally, not a priori. And we're nowhere near the point of even beginning to develop and test methods of translation sophisticated enough for it to be determined.

>> No.10238765

>>10238754
>You're never trying to emulate human intelligence

>if we throw enough figurative gray matter at it it'll work

Make up your mind, idiot.

>> No.10238768 [DELETED] 

>>10238689
The same could be said about language, in a way. Human language is much more chaotic than mathematics, so a digital computer would be able to handle math better than any language that any human speaks.

>> No.10238771

>>10238765
Whom are you quoting?

>> No.10238773

>>10238754
If it cannot understand it the same way we do, the translation will not be 100% accurate. Which is exactly what we've been discussing all along, the possibility of flawless translation.

Checkmate.

>> No.10238774

>>10238771

>>10238754
and
>>10238699

If you're going to be a bitch about greentexting, at least do it when it's appropriate.

>> No.10238780

>>10238771
pretty sure she was quoting >>10238699

>> No.10238786

>>10238774
It is perfectly appropriate and in fact necessary. It turns out you're quoting two different people. Two different people, you see, can't be expected to reconcile their position. And anyone would be confused if you put some other person's words in their mouths.

>> No.10238792

>>10238768
I wasn't trying to start a debate or anything
was just saying

>> No.10238803

>>10238773
>If it cannot understand it the same way we do, the translation will not be 100% accurate.

But that's wrong. There are already many tasks that the machines perform accurately without "understanding" things the way humans do. (Not 100% of course, you hardly ever can achieve 100% in the real world, and it's certainly impossible for humans. Not to mention things like translation of natural language can't even have an evaluation function that would allow proving said 100% accuracy.)

>> No.10238817

>>10238803
Except what was being discussed here was the possibility of perfect translation that would make learning other languages unnecessary. You did the equivalent of entering a bar where people are talking about old school punk rock, get up on the counter and scream like a monkey until you had everyone's attention and then say you like Green Day.

>> No.10238836

>>10238817
All he said was it won't be exact 100% because that's impossible for humans and machines. It can still make language learning unnecessary.

>> No.10238854

>>10238836
Seek professional help. I'm outta here.

>> No.10238855

>>10238680
you think I'm joking you FUCKING retard I'll BREAK your fucking JAW

>> No.10238860

>>10238817
>perfect translation that would make learning other languages unnecessary

The notion of the possibility of a "perfect" translation was already dismissed at the beginning of the thread. That doesn't preclude a translation being good enough to make learning other languages unnecessary.

>> No.10238865

>>10238854
Don't be stupid. If humans can understand 99% of the time and machines can understand 98% of the time, would you spend years learning a language?

>> No.10238909

Dumb.

>> No.10238911

>>10238865
Also, if even people using the same language often misunderstand each other, it's quite unreasonable to think you'll ever achieve 100% understanding of foreign language even if you learn it, or to expect the machine translation to be of 100% clarity to you.

>> No.10239620

Learning a language is fun.

>>
Reply to thread [?]Don't expect anything heroic. Your post will not be uploaded to the original board.
Name
E-mail
Subject
Comment
Action