| Pages in topic: < [1 2 3 4] > | Microsoft developing instant translation tool that 'works like a human brain'
|
|---|
| |
as long as the human brain remains largely unexplored we don't have to worry about that tool I guess.
| | | | | One small detail | Nov 15, 2012 |
John Searle made a great contribution to the whole human brain / computer debate back in the 80's (can't remember book title), putting forward a common-sense approach where, summing it up, if we actually think about it, an AI can never truly mimic human intelligence because, duh!, the human brain is part of a LIVING organism. Might be (surely is) just Microsoft looking for $$, might also be an interesting development, but it can never be what the headline states (is it ever?).
| | | | | @Oliver: proposed change to headline | Nov 15, 2012 |
"that wants to work like human brain"
| | | |
Heh! Could be a good alternative
| | |
|
|
|
Oliver Romero wrote:
John Searle made a great contribution to the whole human brain / computer debate back in the 80's (can't remember book title), putting forward a common-sense approach where, summing it up, if we actually think about it, an AI can never truly mimic human intelligence because, duh!, the human brain is part of a LIVING organism. Might be (surely is) just Microsoft looking for $$, might also be an interesting development, but it can never be what the headline states (is it ever?).
That's not much of a contribution to the debate, to be honest. I don't even find it to be much of an argument. Of course the human brain is part of a living organism but that fact doesn't say anything about whether or not it can be modeled in computer hardware/software. It just means that it's tricky to model, which is hardly a revelation.
When it comes down to it, the brain is just hardware. Neurons, neural connections, neurotransmitters etc. It's unspeakably complex due to the number of neurons and connections, but there is nothing unfathomable about the building blocks or the fundamental principles of its operation. Give it enough time and we will get there, or thereabouts at least.
Now, as I noted above, the hyperbole in the title of the article refers to neural networks, which do mimick the structure and modus operandi of the brain to a certain extent. Neural networks have been around for more than half a century, so it's a bit late to make the "can't be done" argument now.
| | | | | @FarkasAndras | Nov 16, 2012 |
I think it's not about designing breakthrough S2S but about the value of what has been presented.
The problem is that it's rarely happen that the people, who communicate orally with each other in different languages, are emotionally neutral.
Even in teaching, the students tend to be more absent-minded than concentrated, hence high probability of unexpected questions and scenarios.
Here is the point.
Sincerely and with greatest respect,
[Edited at 2012... See more I think it's not about designing breakthrough S2S but about the value of what has been presented.
The problem is that it's rarely happen that the people, who communicate orally with each other in different languages, are emotionally neutral.
Even in teaching, the students tend to be more absent-minded than concentrated, hence high probability of unexpected questions and scenarios.
Here is the point.
Sincerely and with greatest respect,
[Edited at 2012-11-16 13:19 GMT] ▲ Collapse
| | | | Phil Hand 
China
Local time: 02:37
Chinese to English
FarkasAndras wrote:
When it comes down to it, the brain is just hardware. Neurons, neural connections, neurotransmitters etc. It's unspeakably complex due to the number of neurons and connections, but there is nothing unfathomable about the building blocks or the fundamental principles of its operation. Give it enough time and we will get there, or thereabouts at least.
Or thereabouts. This is one of the big errors in AI thinking. In some limited areas, computer technology already outstrips the brain by far. For example, your laptop can concentrate much more processing power on simple problems involving large numbers than your brain can.
By the time we manage to build networks of comparable complexity - and by the time anyone cares enough to reproduce in silicon the odd brain bits that don't do things we think of as useful, like the emotion centres - the technology for number crunching will have advanced many times over again. At that point, if we build AI, it might have some human-like features, but it will also have processing power like a god, so the result will again not be anything remotely like human intelligence.
There will never be a time when our technology is such that it seems relevant to build a thing just like the human brain.
| | | | | Pocket brain :) | Nov 17, 2012 |
When my iPhone will be easily trainable enough to recognize accurately and on the spot every word of the quarrel in my condo accounting office (or similar standard life situations), then I'll believe that it's a breakthrough.
[Edited at 2012-11-17 12:26 GMT]
One more, free input into breakthrough engineering: we teach not by striking the dedicated spots on the student's head (if we do, just to increase the concentration), but by our own quiet and familiar voice, without pressing the ... See more When my iPhone will be easily trainable enough to recognize accurately and on the spot every word of the quarrel in my condo accounting office (or similar standard life situations), then I'll believe that it's a breakthrough.
[Edited at 2012-11-17 12:26 GMT]
One more, free input into breakthrough engineering: we teach not by striking the dedicated spots on the student's head (if we do, just to increase the concentration), but by our own quiet and familiar voice, without pressing the keyboard keys.
[Edited at 2012-11-17 12:38 GMT]
[Edited at 2012-11-17 12:40 GMT] ▲ Collapse
| | |
|
|
|
| you guys are in for a surprise | Nov 18, 2012 |
I'm amazed that the majority of translators on this forum (with the noteworthy exception of Farkas), who are by definition reasonably intelligent and well educated, should be so unaware of what's happening in technology.
Microsoft are not simply proposing to create a machine that emulates the human mind, but rather one that out-competes the BEST human minds. Machine translation may still suck, but it was non existent only 30 years ago and already out-performs most humans. Which of us co... See more I'm amazed that the majority of translators on this forum (with the noteworthy exception of Farkas), who are by definition reasonably intelligent and well educated, should be so unaware of what's happening in technology.
Microsoft are not simply proposing to create a machine that emulates the human mind, but rather one that out-competes the BEST human minds. Machine translation may still suck, but it was non existent only 30 years ago and already out-performs most humans. Which of us could possibly translate instantly across 50 languages? Most people have difficulty mastering their own tongue (...). Has no-one on this forum seen Watson play Jeopardy?
Ray Kurzweil has just published a book (http://howtocreateamind.com/) that sets out a roadmap for generalized artificial intelligence, which should be with us within 20 or 30 years. At that point nobody can tell what will happen. What is sure is that human translators will not be needed, although "post-editors" with excellent language skills will never be without work.
PS. To get an idea of the exponential pace of technological development, see Kurzweil's book "The Singularity is Near". ▲ Collapse
| | | | Phil Hand
China
Local time: 02:37
Chinese to English
| Don't get the logic here | Nov 18, 2012 |
david young wrote:
Microsoft are not simply proposing to create a machine that emulates the human mind, but rather one that out-competes the BEST human minds...although "post-editors" with excellent language skills will never be without work.
If these machines are going to be so clever, why will post-editing be necessary?
Honestly, what you've said there makes no sense. Either computers will be smart enough to understand (and produce) human language or they won't. If they're not, then they can't translate. Post-editing is never going to make sense as a process, because if you need human involvement, you might as well have top-quality human involvement, i.e. translation. Otherwise you're just slowing down the computer for no significant gain.
Now, if computers do become smart enough to understand language, we'll be out of a job, but that'll be cool, we can just buy a robot with our savings and it look after us till the end of our days. But I don't think that day is coming anytime soon.
| | | | | No surprises I'm afraid | Nov 18, 2012 |
david young wrote: I'm amazed that the majority of translators on this forum (with the noteworthy exception of Farkas), who are by definition reasonably intelligent and well educated, should be so unaware of what's happening in technology. Microsoft are not simply proposing to create a machine that emulates the human mind, but rather one that out-competes the BEST human minds. Machine translation may still suck, but it was non existent only 30 years ago and already out-performs most humans. Which of us could possibly translate instantly across 50 languages? Most people have difficulty mastering their own tongue (...). Has no-one on this forum seen Watson play Jeopardy? Ray Kurzweil has just published a book ( http://howtocreateamind.com/) that sets out a roadmap for generalized artificial intelligence, which should be with us within 20 or 30 years. At that point nobody can tell what will happen. What is sure is that human translators will not be needed, although "post-editors" with excellent language skills will never be without work. PS. To get an idea of the exponential pace of technological development, see Kurzweil's book "The Singularity is Near".
Actually, the first public demonstration of MT took place in 1954 (http://en.wikipedia.org/wiki/Georgetown-IBM_experiment), and research started long before that. And although that experiment was unsuccessful, ever since then, MT was 'just around the corner'. Just as it is now. We will probably not see fully automated high-quality machine translation in our lifetimes, and I wouldn't be surprised if no one ever does.
And AI is a such a broad term that people often have misconceptions about it. Both Watson, Deep Blue and many others are examples of Weak AI, while MT requires Strong AI (http://en.wikipedia.org/wiki/AI-complete). And we're still far away from that.
The bottom line is, I'd be more than glad to print out this post and eat it if Microsoft really does it, but most likely it will work only on sentences carefully selected for the demonstration.
| | | | | IBM simulates 530 billon neurons, 100 trillion synapses on world’s fastest supercomputer | Nov 19, 2012 |
http://www.kurzweilai.net/ibm-simulates-530-billon-neurons-100-trillion-synapses-on-worlds-fastest-supercomputer
compare this to only 60 years ago: http://gajitz.com/compute-this-worlds-oldest-working-computer-on-display/
Technology advances at an exponential rate, but our human brains are designed and trained to reason in linear terms. This is why most people fail to see what's just round the corner.
The reason I think post-editors will be necessary for the foreseeable future has to do with style, which calls on emotions, feelings about what "sounds nice/right". In addition, how many humans (including translators) are able to write grammatically and syntactically perfect prose? Thus, while computers will soon be able to produce near-perfect translations, most of the source texts will be imperfect, and a translator-linguist with the ability to read the author's mind, and thus to resolve ambiguities and imprecision, will always ("always" being a few decades...) be necessary.
What translators, and all other workers/consumers should be most concerned about is how the profits generated through technological advances will be distributed (see the minimal corporate taxes paid by Google, Facebook, etc., and their executives' multimillion-dollar bonuses and stock options). If a computer takes over our jobs, will it mean we get to lounge in the sun, or will we be reduced to doing unskilled work just to survive? But that's another kettle of fish... ▲ Collapse
| | |
|
|
|
| Don't believe the press release | Nov 19, 2012 |
The part that works vaguely like a human brain (or a fruit fly's brain for that matter) in this system is the speech recognition, not machine translation. Reading the original article quickly, I got the impression that the actual machine translation component is probably just Bing. Geoffrey Hinton is actually doing a web course right now on Coursera about the kind of deep neural nets used in this speech recognition system. They are ingenious, but I'm not sure that technology transfers well to ma... See more The part that works vaguely like a human brain (or a fruit fly's brain for that matter) in this system is the speech recognition, not machine translation. Reading the original article quickly, I got the impression that the actual machine translation component is probably just Bing. Geoffrey Hinton is actually doing a web course right now on Coursera about the kind of deep neural nets used in this speech recognition system. They are ingenious, but I'm not sure that technology transfers well to machine translation (it would probably be too computation-intensive, speech recognition is an easy task in comparison). ▲ Collapse
| | | | Phil Hand
China
Local time: 02:37
Chinese to English
david young wrote:
Thus, while computers will soon be able to produce near-perfect translations, most of the source texts will be imperfect, and a translator-linguist with the ability to read the author's mind, and thus to resolve ambiguities and imprecision, will always ("always" being a few decades...) be necessary.
In other words, computers won't actually be able to understand human texts.
So much for that argument.
| | | | | Statistical MT | Nov 19, 2012 |
We're talking here about computers understanding texts, and yet the very best machine translation systems today (GT, Bing...) use a method called 'statistical machine translation', which requires no understanding of the text whatsoever.
| | | | | Pages in topic: < [1 2 3 4] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Microsoft developing instant translation tool that 'works like a human brain' | Anycount & Translation Office 3000 | Translation Office 3000
Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.
More info » |
| | Protemos translation business management system | Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!
The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
Translation news
