| Pagina's in het onderwerp: [1 2] > | Poll: In your view, how long is it before computers translate by themselves? De persoon die dit onderwerp heeft geplaatst: ProZ.com Staff
|
|---|
This forum topic is for the discussion of the poll question "In your view, how long is it before computers translate by themselves?".
View the poll results »
| | | | Lieven Malaise
België
Local time: 07:37
Lid 2020
Frans naar Nederlands
+ ...
Someone should offer a training course in asking clear questions that aren't as vague as a cloud, instead of asking questions that only raise more questions.
| | | | Lingua 5B 
Bosnië en Herzegovina
Local time: 07:37
Lid 2009
Engels naar Kroatisch
+ ...
Lieven Malaise wrote:
Someone should offer a training course in asking clear questions that aren't as vague as a cloud, instead of asking questions that only raise more questions.
I thought it was only computers. Now humans do that too?
| | | | Thomas Pfann
Verenigd Koninkrijk
Local time: 06:37
Lid 2006
Engels naar Duits
+ ...
The computer says:
The question is fairly clear, but it could be slightly refined for clarity and precision. Here's a revised version:
"In your opinion, how long will it take for computers to achieve fully autonomous translation?"
This version emphasizes the idea of fully autonomous translation and maintains the conversational tone suitable for a forum.
[Edited at 2024-11-21 09:33 GMT]
| | |
|
|
|
Lingua 5B
Bosnië en Herzegovina
Local time: 07:37
Lid 2009
Engels naar Kroatisch
+ ...
They already provide fully autonomous translation. The quality and errors are a whole different matter.
| | | | Denis Fesik
Local time: 09:37
Engels naar Russisch
+ ...
Banning most of the world's languages while leaving just a select few. Banning most things within those select few that make them special like cultural codes and nuances, historical backgrounds, spiritual quests, dialects, scientific traditions, folklore, jokes, puns, and pretty much anything that will take too much computing power to express in another language. Unifying and sterlilizing existing language corpora (besides which, no language will be allowed to exist) and creating sets of transla... See more Banning most of the world's languages while leaving just a select few. Banning most things within those select few that make them special like cultural codes and nuances, historical backgrounds, spiritual quests, dialects, scientific traditions, folklore, jokes, puns, and pretty much anything that will take too much computing power to express in another language. Unifying and sterlilizing existing language corpora (besides which, no language will be allowed to exist) and creating sets of translation rules to be applied on pain of severe sanctions. Delegating all translation jobs to computers whenever they need to be done. If things keep going as they are, all that should be quite doable; it's just that I don't believe they will keep going like that ▲ Collapse
| | | | Lieven Malaise
België
Local time: 07:37
Lid 2020
Frans naar Nederlands
+ ...
Lingua 5B wrote:
They already provide fully autonomous translation. The quality and errors are a whole different matter.
Exactly, hence my remark about clarity. I can only imagine that the asker meant to ask how long it will take before computers are able to translate error-free, but that's a long way from what is being asked now.
| | | |
A strange question on a translation platform that is constantly promoting CAT tools and... AI!
| | |
|
|
|
| It won't take 10 years | Nov 21 |
They are already capable of delivering fit-for-purpose translations for many objectives. Most of this translated content is new (i.e., it was not previously produced by human translators due to the prohibitive time and cost involved). However, with the ongoing rapid developments in the field, I expect human translation to be significantly diminished within a decade.
| | | | JaneD
Zweden
Local time: 07:37
Lid 2009
Zweeds naar Engels
+ ...
Computers aren't human. Only humans can translate the sensations, memories, associations and depths of meaning that a human has written.
Perhaps the question should be rephrased: How long will it be before we all become so numbed to the slop produced by machine translation that we lose our ability to appreciate and prefer human language?
| | | | | Other: They already do | Nov 21 |
We have neural MT and large language models that translate by themselves. I believe the main question is quality. Is the computer's product suitable for a specific purpose in terms of quality (accuracy, style, tone, etc.)? Sometimes yes, sometimes no. With technological development, non-human-edited translation will become suitable for more purposes, but I don't think it'll become suitable for all purposes, whatever the technological advancements.
| | | | | They are already doing it | Nov 21 |
AI already doing it, and this is not a new thing, maybe now is more accurate than before. First we had Ace Translator in the 90's, than Google bough it and make it Google Translate, than the MT apps entered the translation field. All these MTs are just simple "find & replace" function like in Word or Excel, with more complex algorithms with regex and other functions. I think that we all here see the idea behind AI, and it's to get rid of the translators and the costs for translating, and eventua... See more AI already doing it, and this is not a new thing, maybe now is more accurate than before. First we had Ace Translator in the 90's, than Google bough it and make it Google Translate, than the MT apps entered the translation field. All these MTs are just simple "find & replace" function like in Word or Excel, with more complex algorithms with regex and other functions. I think that we all here see the idea behind AI, and it's to get rid of the translators and the costs for translating, and eventually all translators to become reviewers. The irony in this is that we as translators can translate very good without MT apps, but anyway we need to buy a MT app, because the client says so. So we, translators, don't need AI, the client needs. We can simply translate the file in Notepad (free Windows text editor) and send it to the client, but the client's conditions are "you must have MT app". So if you ask me, we are doing more than reviewing translations, we are also doing data entry in client's AI matrix. And after all this screwing with us translators, we have scammers who are trying to get free service and wasting our time. So, my advice is to raise your rates up for reviewing MT, mentioning the time consuming data entry process.
[Edited at 2024-11-21 16:29 GMT]
[Edited at 2024-11-21 19:53 GMT] ▲ Collapse
| | |
|
|
|
| They're almost there | Nov 21 |
I've seen translators on here insisting, for years, that they will never be able to translate like a human.
Maybe.
But that's not the point.
The question is whether they will be good enough to translate most of what humans used to, with quality that is sufficient to meet clients' needs.
And that is already happening, or is imminent.
At this point, rejecting CAT tools and AI is like the horse and buggy driver who laughed at cars in the early 1900s... rat... See more I've seen translators on here insisting, for years, that they will never be able to translate like a human.
Maybe.
But that's not the point.
The question is whether they will be good enough to translate most of what humans used to, with quality that is sufficient to meet clients' needs.
And that is already happening, or is imminent.
At this point, rejecting CAT tools and AI is like the horse and buggy driver who laughed at cars in the early 1900s... rather than learning how to drive one. ▲ Collapse
| | | |
Justin Peterson wrote:
I've seen translators on here insisting, for years, that they will never be able to translate like a human.
Maybe.
But that's not the point.
The question is whether they will be good enough to translate most of what humans used to, with quality that is sufficient to meet clients' needs.
And that is already happening, or is imminent.
At this point, rejecting CAT tools and AI is like the horse and buggy driver who laughed at cars in the early 1900s... rather than learning how to drive one.
Nobody is saying to reject CAT tools, all I'm saying is that the translators need to teach the AI model to translate by data entry and while they are doing it that in CAT tool which they must buy it if they want to work, they should be aware that eventually he or she will be rejected for further jobs by the same company which uses and sell them the CAT, AI or whatever you call it. In some other post on this forum I saw that someone said that AI don't let him to edit some strings, because they've already were considered as true. So less words count for the reviewer, right? That's the whole point of the AI. In the end it will be just crumbs for chickens. The big slice never gonna be for the workers.
This is happening around the world in all IT related fields, video and audio productions, in automotive industry and pretty much in everything. We can't do anything about it, and we should accept it that we are already screwed. And if those people who are now trying to shut down the FBI, because that would be best in theirs interest, internet will become paradise for the scammers. If I show you the unsuccessful tries to log in into my emails you will be shocked that the IP's are from the same countries as the clients where I submitted application for a freelance job. And to whom should I complain? To the ProZ staff?
| | | | Denis Fesik
Local time: 09:37
Engels naar Russisch
+ ...
| Ever smaller leaps | Nov 21 |
Since the beginning of the industrial revolution, technology has been developing by leaps, but those who studied them have noticed that the leaps kept becoming smaller and smaller. I'd argue that we can't talk about the recent AI craze as a breakthrough comparable to those that began to happen early into the tech race. I'm quite immune to all sorts of hype, so what I'm seeing as a translator doesn't fascinate me: AI's translation performance is not much higher (if at all) than what I saw quite a... See more Since the beginning of the industrial revolution, technology has been developing by leaps, but those who studied them have noticed that the leaps kept becoming smaller and smaller. I'd argue that we can't talk about the recent AI craze as a breakthrough comparable to those that began to happen early into the tech race. I'm quite immune to all sorts of hype, so what I'm seeing as a translator doesn't fascinate me: AI's translation performance is not much higher (if at all) than what I saw quite a few years ago when MT engines first got neural. If they can read your TM, the better, but here's the deal: whenever you tread on virgin ground with your translation, no AI outputs will help you. As translators who always think about translations, we may well be afraid to imagine that one day they really will spend all the energy it takes to translate 99% of the texts that need to be translated using AI tools that can deliver acceptable results. The rest of the energy will go into other tasks (won't energy be available galore, after all?).
I recently watched a video about some research involving a tokamak. This tech dates back to 1950s, and today they were first able to contain plasma for five seconds. Thermonuclear power is right around the corner, isn't it? Oh, yes, there's hydrogen (again, a very old idea cherished by green energy enthusiasts). I'm talking about green hydrogen produced by electrolysis of water rather than one obtained from natural gas. This process takes more energy than you can get by burning hydrogen in air, requires metals that can't be produced in green ways, and then there's the embrittlement problem. It can be avoided by burning hydrogen in pure oxygen, but how do you deal with temperatures around 2000°C? You spend more energy to build a wind turbine than it will produce over its entire life cycle. Solar panels and batteries can't be recycled once scrapped without a heavy toll on the environment (our customer is building recycling plants so I've traslated a lot of related materials). The only player in the global nuclear energy market that can recycle nuclear fuel is ROSATOM (and who in the first world is crazy enough to deal with ROSATOM?). But progress is still exponential, so somehow all these problems will be solved in no time.
Here's a translation problem. You're doing dimensional inspection on hex bolts, and the inspection records are in Turkish English (likely produced by a machine and maybe with some Russian involved—or maybe not, I'm not sure). There's a dimension called "diagonal," and as a translation pro you know that normal English calls it "width across corners" (and normal Russian calls it a circumcircle diameter). Then there's a "key dimension"—and this time it's harder to guess that this should in fact be rendered as "width across flats" (key = wrench). There's "%BREAKDOWN," which takes my level of experience to decypher as percentage elongation at break. And then there's "Master control go / no go." This is quite obscure to me, I have no good ideas (while knowing that control = inspection), and translating into Turkish and back doesn't help. This, mind you, is a simple, dictionary-level problem, it's just that I don't speak Turkish. But I wonder if someone who can't speak Russian can still prompt AI enough to deliver correct Russian translations for things like "compressive capacity," "flexural capacity," "elastic capacity," "plastic capacity," "bearing capacity," etc., all as used in structural engineering in the context of limit-state design? I've checked this one on machinetranslation.com; none of the engines were able to produce anything close to sensible results, but what if it's possible with clever prompt engineering? I'm not going to KudoZ with this because I know how to translate each expression. It's not a problem for someone who knows how the industry talks about things. I gave more examples in my other posts—and can give many, many more while entertaining no doubt that eulogies for translation AI will not be gone from this forum anytime soon
[Редактировалось 2024-11-22 06:05 GMT] ▲ Collapse
| | | | | Pagina's in het onderwerp: [1 2] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Poll: In your view, how long is it before computers translate by themselves? | Trados Business Manager Lite | Create customer quotes and invoices from within Trados Studio
Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.
More info » |
| | Trados Studio 2022 Freelance | The leading translation software used by over 270,000 translators.
Designed with your feedback in mind, Trados Studio 2022 delivers an unrivalled, powerful desktop
and cloud solution, empowering you to work in the most efficient and cost-effective way.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
