The evolution of the translation machine

[Authary]

Intelligence IS computing power.
A monkey has significantly less computing power than a human.
The humans 5000 years ago had the same computing power as us, but they didn't know how to use it properly.

Oh shit, I guess I wasted the last few years studying artificial intelligence. Better tell my colleagues so we can just shut down the whole laboratory and go on with our lives.

Now let's say your frozen for 10000 years and wake up when we have machines capable of computing 100x faster than a human brain. Now make it count from 1 to... anything. That's intelligent, uh ? No. It's all about what you compute, not about how fast you do it. The human brain is relatively fast but it sucks when it comes to anything not related to language, spatial perception or stuff like that..

And your last sentence contradicts your first.

[haxartus]

Now let's say your frozen for 10000 years and wake up when we have machines capable of computing 100x faster than a human brain. Now make it count from 1 to... anything. That's intelligent, uh ? No. It's all about what you compute, not about how fast you do it. The human brain is relatively fast but it sucks when it comes to anything not related to language, spatial perception or stuff like that..

Imagine that you had the right algorithms on a today's PC.
You can ask a simple question and wait 10 minutes for the answer.
Is that intelligence ?

[Authary]

Imagine that you had the right algorithms on a today's PC.
You can ask a simple question and wait 10 minutes for the answer.
Is that intelligence ?

What kind of simple question requires 10 minutes of modern computing ? My (very) poorly programmed algorithm learns 1600 relationships between conjunctions of concepts from a set of examples in less than 2 minutes. I couldn't do it myself in less than a day.

[haxartus]

What kind of simple question requires 10 minutes of modern computing ? My (very) poorly programmed algorithm learns 1600 relationships between conjunctions of concepts from a set of examples in less than 2 minutes. I couldn't do it myself in less than a day.

"How are you feeling today ?"
But without automated answers like "I'm fine". That's not AI.
I want complete understanding of the concept of feelings, complete understanding of the concept of time, and the difference between biological organisms and machines. I want an intelligent answer, not a pre-programmed one.

[Deleted]

The humans 5000 years ago had the same computing power as us, but they didn't know how to use it properly.

Yes they did... but there community did not allow to many specialist, some of them become master artisan in like say painting and in holding religious ceremony but the rest was force to be master artisan in Hunter-gatherers.

[SanshaT]

Yes they did... but there community did not allow to many specialist, some of them become master artisan in like say painting and in holding religious ceremony but the rest was force to be master artisan in Hunter-gatherers.

One of the problems with most people I believe is that we do not focus. We are raised in a society that trys to force us to learn all subjects and we wind up being mediocre in all of them. Then we go into college where we TRY to focus on a certain subject but are often forced to learn even more completely unrelated ones as part of the curriculum which in turn takes away from the one we are trying to learn.

If a translation machine is designed JUST to translate and nothing more I do not think it would require near the power of the human brain to be adequate at its job.

[haxartus]

Properly translating a book is something infinitely more difficult than anything a translation machine is capable of, or would be capable of in the next 15-20 years.
As for simple, basic translations... Google Translate already does the job fairly well.

[SanshaT]

Properly translating a book is something infinitely more difficult than anything a translation machine is capable of, or would be capable of in the next 15-20 years.
As for simple, basic translations... Google Translate already does the job fairly well.

I think we will probably arrive at that point within the next 10 years or so given the rate of progression we have now.

[haxartus]

I think we will probably arrive at that point within the next 10 years or so given the rate of progression we have now.

And why do you think that ?

[SanshaT]

And why do you think that ?

Going by the rate of improvement from translation machines 10 years ago and the ones today, if you double that progression mathematically you have close to perfection.

Moores law

[haxartus]

Going by the rate of improvement from translation machines 10 years ago and the ones today, if you double that progression mathematically you have close to perfection.

It's mIt's much more complex than you think. I've said it like 10 times by now.
When you do a word by word translation, much of what the original author was trying to convey is lost, which destroys the feeling of book.
A good translator knows that and attempts to recreate the structure of the text in a way that it's closer to the original meaning, which means that a big number of changes are made to the structure of the sentences and the text.

[SanshaT]

It's much more complex than you think. I've said it like 10 times by now.
When you do a word by word translation, much of what the original author was trying to convey is lost, which destroys the feeling of book.
A good translator knows that and attempts to recreate the structure of the text in a way that it's closer to the original meaning, which means that a big number of changes are made to the structure of the sentences and the text.

Perhaps not for a novel, but I could see it working fine for instruction manuals, tv menus, news websites and things of that nature.

[Nino]

voice translators are in phones now....

[haxartus]

Perhaps not for a novel, but I could see it working fine for instruction manuals, tv menus, news websites and things of that nature.

One of the problems is that each language has to be studied by the machine differently, and it's barely working with English right now. English is one of the most basic languages for studying, so it's not really hard for a machine to create grammatically correct sentences and stuff like that.
But try translating a news site from English to Japanese...

[v2prwsmb45yhuq3wj23vpjk]

Perhaps not for a novel, but I could see it working fine for instruction manuals, tv menus, news websites and things of that nature.

I would say even for a novel. I think he grossly underestimates what computers are capable of even now, let alone in 10 years. It will all come down to how much development time is put into it.

[haxartus]

I would say even for a novel. I think he grossly underestimates what computers are capable of even now, let alone in 10 years. It will all come down to how much development time is put into it.

I think that you have never translated anything serious in your life. I doubt that you even know a second language on a level to translate something like a novel.

[SanshaT]

One of the problems is that each language has to be studied by the machine differently, and it's barely working with English right now. English is one of the most basic languages for studying, so it's not really hard for a machine to create grammatically correct sentences and stuff like that.
But try translating a news site from English to Japanese...

To my knowledge most translation machines originate from the USA and set their base language as English. Because of this they tend to do better going into English than any other language.

If other countries want ones that are optimized going into their languages they would probably have to start developing their own. I know there are some Japanese made translation machines that can do better going into Japanese than the ones we have made here in the United States.

[Fojos]

Intelligence IS computing power.
A monkey has significantly less computing power than a human.
The humans 5000 years ago had the same computing power as us, but they didn't know how to use it properly.

Didn't know how to use it properly? What, in your mind, is using it properly? All our knowledge is based on the studies made by people several thousand years ago, and I'm pretty certain they knew how to "use" their brain. They probably knew how to use their brains better than most people alive today, because they had creativity, something we severly lack today - where many study for money, and not to know.

[v2prwsmb45yhuq3wj23vpjk]

I think that you have never translated anything serious in your life. I doubt that you even know a second language on a level to translate something like a novel.

And what are your credentials?

Actually, I suppose it doesn't really matter what either of us say because it could just be a lie anyway. The basis for my opinion comes from the observation of the growth of the capability of technology. What could happen and what will happen are naturally very different. As I said before, it will come down to the amount development time put into working on this technology. I believe it could get to a very advanced point in the next ten years, but whether that will happen or not is another story entirely.

[haxartus]

And what are your credentials?

Actually, I suppose it doesn't really matter what either of us say because it could just be a lie anyway. The basis for my opinion comes from the observation of the growth of the capability of technology. What could happen and what will happen are naturally very different. As I said before, it will come down to the amount development time put into working on this technology. I believe it could get to a very advanced point in the next ten years, but whether that will happen or not is another story entirely.

I don't think that you understand how translation works.
Machines translate words. Humans don’t translate words. They translate meanings.
When you are proficient in a given language, you don’t translate anything in your head anymore. You think in the language and when you read something, it's meaning appears in your head. Not the words, not the grammar, the meaning. And then you think of the best way to express the meaning in the other language. This is the hardest and most important part.
Machines have no concept of meaning. None. This part of the translation process is completely lost to them. It's not about computation power, or advancement in the algorithms. They simply can't do it.