No, there are insurmountable barriers to human-level computer translation. To properly understand human language it is necessary to live in a world of experience.
What do you think about the future of the translators?
Very, very difficult barriers. Insurmountable? I wouldn’t go that far. Language is complex, but it is extremely finite as well.
I would give translation less than 10 years before a computer can do it nearly as well as a professional. Better before 50 years.
As recording devices are made to interpret the inflections of speech and cameras can be used to interpret body language, an AI. could do the job with more accuracy and without prejudice. Of course such complete honesty can cause huge issues. Sometimes we are better off ignoring.
At this stage of time i would give it a better than 80-20 chance that this state of the art progression will happen. If things proceed with this type of predicitve certainty, the following may happen, not necessarily in the realm of transcribing, or translating: There will be a conflict between man and machine. A war of an intra world type. Namely, there will become an awareness of lurking danger by both, man and machine, and all segments of life will be, and i boldly state/ EVEN NOW, will be effected. Political, social, ecenomic, psychological and technical crisis will be reached at some point, producing a matrix of predicting most probbel results, particlarily focusing on the limits of societal awareness of limits of endurance toward a preception AND an actual sustanance od humanity’s will to live. Thereafter the course will be programmed, and if the thesis, that a minority segment of the population will wish to deny the results of such study will not result in a synthesis; the result of an obvious antithetical force emerging, those very same 1-20% will seethe contra production of continuing that course, since they themselves will place themselves into jepardy. We touched upon this previously, but i don’t see any meaningful resolution, since we are still debating it without the core question, by skirting it with peripheral technical feasibility issues such as translation-transcription.
Here’s an interesting, in-depth exploration of the topic.
math.ku.dk/~m01mwm/The%20Lim … 009%29.pdf
Your wife might enjoy it.
Here are the last two paragraphs:
As I showed in the introduction and in chapter 2, the history of machine translation has a
tendency to repeat itself. The fact that we always seem to have come three quarters of the way
again and again incites the hope that this time, we might be able to cover that last quarter. But
instead of inciting hope, this situation should perhaps rather remind us of the fact that drinking a
fourth beer always seems like a better idea after the first three.
And it would be appropriate to curb our enthusiasm, even though machine translation certainly
is an exciting challenge. Unlike the researchers looking ahead at a promising future from half a
century ago, we can no longer claim that we do not know the cost or the chances. Yet another
half a century of attempts to make “machines that think, that learn, and that create” (Simon and
Newell, 1958) would indeed be resources badly spent.
The idea that we haven’t advanced in algorithm design and computing and therefore we are bound to fail in the same ways seems pretty patently false to me. Yes, people failed at making machines do accurate translations before. Yes, they failed multiple times. If humanity just gave up every time it failed at something a couple of times, we wouldn’t have come very far at all.
Language is the competence to form infinte linguistic terms with a finite inventory of linguistic forms. It has much to do with thoughts, mentality, conceptions, beliefs, imaginations, conventions, experiences, awareness, knowledge, information, communication … and so on. It is such a complex system that one could say that machines could never reach this high competence that humans have. But it is merely a question of time whether machines will be able to use language like humans do. So when?
But that isn’t my idea, or the idea in the paper I linked to. Computer translation doesn’t fail because the algorithms aren’t good enough, it fails because to be a good translator you have to know what it is like to live in our complex world of experience. There is an immense background of understanding about how the world works which we develop through experience.
I liked this example from the paper I linked to:
(a) The city councilmen refused the demonstrators a permit because they feared violence.
(b) The city councilmen refused the demonstrators a permit because they advocated violence.
“They” refers to the councilmen in (a) but to the demonstrators in (b).
You can’t write an algorithm to deal with this. The knowledge that is required to understand who “they” refers to is part of the background of understanding of how the world works, which we gain through experience.
I translate all kinds of texts, technical, legal, commercial. Very often in all these cases the writer is trying to make the reader feel good about some things and bad about others. A technical writer might want you to feel bad about a machine component that fails. Lawyers want to make you feel good about their clients and bad about their opponents, companies want you to feel good about their products and services.
In order to translate something like that, you need to know what it is to feel good or bad, what kind of thing makes people generally feel good or bad.
So that’s the insurmountable problem for computer translation. I’ve just given two examples, but the kind of problems for machine translation which they illustrate occur in almost every text.
English is not a very good example when it comes to understand any of all kinds of the linguistic reference. There are languages with a grammar that shows clearly all kinds of reference between the linguistic forms because the linguistic deep structure is more noticeable / distinguishable in that languages than (for example) in the English language. The linguistic deep structure can be learned by machines as well as knowledge and experience.
It is not the insurmountable problem for computer translation. In the future machines will translate more effectively than humans.
The real truth of it is, many roles cannot be performed by machines.
Not yet, but in the future machines will be able to do it.
There is absolutely nothing that a human can do that a machine cannot be designed to do better.
I can’t wait to see the cyborgs and androids pleading, “Oh poor abused me.” Then following it with a righteous indignation attack, wiping out all of mankind.
I am very sorry, James, but i would really rather die a peaceful death in my bed, then to be wiped out by machines. If You can’t wait to see that happen, You will be among ALL of mankind to be wiped out. There is no guarantee that it will be a very kind death with the human touch.
Twilight Zone, the original, had an episode concerning the displacement of humans by machines. Although Youtube has almost the entire TV series, they don’t display that episode, but has the radio version if you have the time to listen to radio (can be more fun than you might think).
[youtube]http://www.youtube.com/watch?v=O_xpnjsw-9k[/youtube]
I remember enjoying that series as a little kid.
Most humans already are “machines” and brain dead zombies. Do you actually believe the average person of average intelligence “thinks” about existence or lives a “worthy” life???
The difference between “human”, machine, and zombie, is that people falsely believe a human is “higher, above, superior” than the others. But how is presumption justified?
People take humanity for granted. People convince themselves that they’re much smarter than they are; because humanity is represented by something higher than humanity.
Humanity is a “right, entitlement, privilege” conferred by something above humanity, that turns homo sapien from merely animal into “something higher, pristine, godly, and divine”.
Machines have already replaced human beings. Because human beings were machines (slaves) from the beginning.
You’d be amazed as to how much of those original series applies to today’s society. But of course they remade the series in the 90’s, leaving out sensitive issues. There were other robot vs man issues in that series. Rod Serling was every bit as good as Orson Wells.
This is his sales pitch to his prospective advertisers;
[youtube]http://www.youtube.com/watch?v=6de1C2LWN1M[/youtube]
Oh I know, 
they had reruns even waaay back then… I probably saw each one a few times at least. They were very spot on in some cases. I tried the remakes, hated them. I don’t have the thirty minutes to watch the vid but, knowing that character, it would be worth the time if I had it.
Incidentally Serling was one of the first cardiac bypass patients, and he died on the operating table during surgery. I guess technology has not quite caught up to the facing challenges. The earliest ventures always come with risk.