| Pages in topic: < [1 2 3 4 5 6] > | Poll: Do you think DeepL is something for human translators to worry about? Thread poster: ProZ.com Staff
|
|---|
Dominic D.
Local time: 22:18
Serbian to Russian
+ ...
Thomas Johansson wrote:
Machine translation is coming strong into our industry and an increasing number of translation jobs are being transformed into machine translation post-editing (MTPE) jobs, often carried out on the basis of a machine translation produced by DeepL or a similar technology.
This comes together with price cuttings. People working for agencies are already being forced to accept considerable discounts (30-50%) for MTPE jobs even if no time savings can be seen. ("Forced" in the sense that they'll lose work and income if they don't accept.)
The result of this development will be that traditional translators will be pushed out of the market and increasingly replaced by less competent workers who are willing to accept lower rates.
I would recommend all translators to prepare themselves for this development and start developing alternative sources of revenues.
Damn, you hit me on the head with this just as I was sitting with an assignment late into the night, reflecting upon the miserable future of this industry.
| | | | | but there will only be junk food | Feb 28, 2023 |
Tom in London wrote: Andreas Baranowski wrote:
, machine translation will find its place and ultimately acquire an overwhelming market share. It’s just a question of time. ....leaving the more discriminating part of the market to those of us who also never eat junk food. 
I wish it were like that, but I suspect there is not going to be any such part of the market left.
The tendency is towards a market where everything will be taken over by MTPE, carried out by less competent people and at lower costs. No one will be requesting high-quality translations or even high-quality MTPE any more. More or less everything (except poetry, literature and similar exceptions) will be done through MTPE, and strong discounts will be enforced, leading to low quality output and low incomes for the workers.
This transformation is already almost complete within the EU institutions, which collectively form the single largest employer of translation services in the world and are setting the standard. (Today, at least 95% of all EU translations are MTPE jobs, and generally a discount of 30% is enforced, leading to hasty output by stressed translators, plagued by inconsistencies, incorrect terminology, clumsy formulations, nonsensical references etc.)
The only way to change this course of development is to generate a strong debate and awareness about quality (through studies, research, statistics, public statements from leading industry actors etc.), and make the agencies and all industry actors, incl. end clients, understand that in many cases MTPE jobs will lead to low and unacceptable quality unless it is carried out by highly competent linguists at processing speeds which (in many cases) cannot be compromised.
We need to generate a detailed and refined debate that leads to a sophisticated awareness of exactly when (or in what conditions) MTPE can be used and when it cannot, when it can be performed by a less competent linguist and when it cannot, and when it can be reasonably expected to generate time savings and when it cannot.
We have a battle to fight.
[Edited at 2023-02-28 22:31 GMT]
| | | | | This did not age well | Mar 3, 2023 |
Joëlle Bouille wrote:
As a professional translator, do you think DeepL could help you work quicker and boost your productivity, while providing the same top quality?
Definitely no. Complete waste of my time.
Does a published writer ask a 6-grader to help him, because said 6-grader is making "huge progress"?
Now did it.
| | | | | Depends on state response | May 17, 2023 |
I definitely worry about AI, especially as a relatively young translator.
I'm not sure AI will ever be able to convey the nuance of, say, literature. But I do worry that governments will start to accept documents for visa applications etc via AI translation platforms, on the understanding that AI does a good enough job at conveying meaning and is potentially less open to corruption.
At least in Australia that would make a big job to workflow for translators.
| | |
|
|
|
The problem is that due to vanity, translators closed their eyes and usually fell into a few patterns.
1. This technology sucks; it'll never replace jobs / my job.
2. I'm too good to be replaced.
3. If you fear being replaced, it's because you suck.
---
This allowed the MT industry to gain a lot of ground, which they will obviously not yield. Those translators didn't realize that MT got progressively better, that nobody is safe, and that maybe, ... See more The problem is that due to vanity, translators closed their eyes and usually fell into a few patterns.
1. This technology sucks; it'll never replace jobs / my job.
2. I'm too good to be replaced.
3. If you fear being replaced, it's because you suck.
---
This allowed the MT industry to gain a lot of ground, which they will obviously not yield. Those translators didn't realize that MT got progressively better, that nobody is safe, and that maybe, some clients don't care so much about quality.
It doesn't help that MT progress is uneven; it's very good in some language pairs, but still bad / unacceptable in rare language pairs. This may give a false perception to some translators that the translators who ARE being affected are bad professionals.
The result is a world where they are replaced by tools that are "good enough"; sure, there may be jobs, but the discounts may be so huge that eventually, translators may not be able to pay their bills from translation alone, pushing humans out of the industry. ▲ Collapse
| | | | jyuan_us 
United States
Local time: 16:18
Member (2005)
English to Chinese
+ ...
It has been almost 5.5 years since the initial post, but MT in general hasn't been much improved. It probably won't be much improved in another 5.5 years.
[Edited at 2023-07-23 10:39 GMT]
| | | | Dominic D.
Local time: 22:18
Serbian to Russian
+ ...
| Wait and see | Jul 23, 2023 |
jyuan_us wrote:
It has been almost 5.5 years since the initial post, but MT in general hasn't been much improved. It probably won't be much improved in another 5.5 years.
[Edited at 2023-07-23 10:39 GMT]
Well, let's wait and see until it gets a solid boost from AI, with AI scanning and correcting the pre-translated text for inconsistencies, logical errors and more... And then AI will also do super-accurate OCR. And then we're all bound to pick a different career path.
| | | | Dan Lucas
United Kingdom
Local time: 21:18
Member (2014)
Japanese to English
| Life goes on | Jul 23, 2023 |
Dominic D. wrote:
Well, let's wait and see
How long should we wait, beyond the 5 years that have already passed? Some of us have been making decent money over that time. Maybe the next 5 years will be different, but jyuan_us makes a reasonable point: the sky is always falling in translator land. It was falling twenty years ago for some people (for proof, just go back and read early forum posts about plummeting rates).
I do think we'll be seeing some interesting changes over the half-decade to come.
Regards,
Dan
| | |
|
|
|
| We need to rethink our marketing. | Jul 24, 2023 |
jyuan_us wrote:
It has been almost 5.5 years since the initial post, but MT in general hasn't been much improved. It probably won't be much improved in another 5.5 years.
[Edited at 2023-07-23 10:39 GMT]
MT has become better. For example, many machine translation systems would have trouble with relative pronouns in multiple sentences. Now, many systems are context-aware, and pretending that everything is all right creates a "head-in-the-sand" mentality where we all pretend nothing is wrong, but that prevents translators from uniting and improving work conditions. Is this what we want for us as a category?
Like I mentioned in another topic, the fact that many of our colleagues have trouble raising rates should ring an alarm and our solidarity. If our profession weren't pressured by MT and AI, we wouldn't have posts of colleagues considering leaving for other professions, questioning the future of translation as a means of living (at least a full-time one, as opposed to something to supplement income), and we wouldn't see the professionals in the subtitling world questioning their job. Why would you quit a well-paying profession with prospects?
Of course, I personally still think that while MT has improved, humans won't completely disappear, for a few reasons that can never be solved:
- If a human translator screws up, in theory a company can go after them. Google Translate / DeepL offer the service as-is, and aren't interested in solving that issue, as it would require them to be more accountable.
- If you rely completely on Google / Microsoft / DeepL / AI for translation, any of these companies being gone could hinder your workflow. Suppose that you have a company, and you ditched all human translators for a workflow that is 100% specialized and reliant on the Google Translate API. What if Google decides to discontinue the API, or change it in a way that requires you to drastically change your workflow? Your company will be in a bad spot.
- A human translator offers convenience, and will make sure to follow whatever workflow is needed to get the problem solved.
Now, if we want to be seen as offering more than a commodity, we as an industry need to market ourselves differently. We need to show that translation is not just about translating words, but we need to do so in a way that is painless and convenient to our clients. It's similar to how people can make a pizza at home or buy a premade one from a supermarket, but they still go to pizzerias for the experience of going out and socializing. But it needs to be a collective effort.
[Edited at 2023-07-24 01:53 GMT]
| | | | Dominic D.
Local time: 22:18
Serbian to Russian
+ ...
| I'd say another 3-5 years | Jul 24, 2023 |
3-5 years I'd say, but you'll call me a false prophet, I know.
Naturally, it's not like the industry is going completely down any time soon, but downward pressure on rates and declining demand is likely going to be massive. The market is probably going to do to us what Google has done to its workforce lately.
There's two types of translations: critical (legal, marketing, accounting etc.) ones and not-so-critical (operating manuals for internal use, MSDS, corporate bulls... See more 3-5 years I'd say, but you'll call me a false prophet, I know.
Naturally, it's not like the industry is going completely down any time soon, but downward pressure on rates and declining demand is likely going to be massive. The market is probably going to do to us what Google has done to its workforce lately.
There's two types of translations: critical (legal, marketing, accounting etc.) ones and not-so-critical (operating manuals for internal use, MSDS, corporate bullshit for internal employee reference) ones. Hard to tell which ones are in majority, but the latter group is likely to just quit the market in the coming years.
In the recent years, when I see a translation done by another translator, I can definitely tell it's made by a human, because it contains a certain number of grammar errors and has stylistic flaws.
[Edited at 2023-07-24 09:02 GMT] ▲ Collapse
| | | | Dan Lucas
United Kingdom
Local time: 21:18
Member (2014)
Japanese to English
| The same old chestnut | Jul 24, 2023 |
Carlos A R de Souza wrote:
pretending that everything is all right creates a "head-in-the-sand" mentality where we all pretend nothing is wrong, but that prevents translators from uniting and improving work conditions.
I don't think anybody here is pretending that "everything is all right". Some people are doing fine, some are not doing so well, and some have left the profession. The business - like many others, not coincidentally - has changed over the past 20 years and will continue to do so.
Even those who are currently comfortable recognise, I think, the potential (remember we are discussing forecasts, expectations, and assertions, not facts) for disruption down the road. Responses vary from "It'll be different but fine" to "It's the end of the industry", which is as one might expect given the enormous diversity of linguistic services and those engaged in providing them.
But the uniting and collective effort stuff has, I am afraid, no grounding in economic reality. Translation is a global, deregulated industry, with no requirement for a physical presence at the location to which the service is provided. Regulation at the country level tends to be rare and presumably would in any case have very little effect. If somebody wants a manual translated from German to English, no regulations in, say, Germany, would apply to a translator in the US or Japan.
Given the previously tremendous diversity in the backgrounds and conditions of freelancers around the world, how are you going to get translators to agree on anything, let alone a complex initiative such as organising labour for higher rates and so on?
Let me phrase the question differently: can you find a successful precedent for such a global undertaking in any other industry, let alone a deregulated one like translation?
This has been discussed umpteen times before, and no persuasive arguments have been put forward by the "let's organise!" side. It's a pipe dream.
Regards,
Dan
[Edited at 2023-07-24 11:00 GMT]
| | | | Anne Brackenborough (X)
Germany
Local time: 22:18
German to English
| We're not all in the same boat | Jul 24, 2023 |
Sorry to state the glaringly obvious, but we are not all experiencing the same changes.
My German clients already thought the English they learned for years in school, on the Internet and on holiday was good. Now they have a tool that makes it better. All they have to do, in their opinion, is read the machine translation and "check it's translated right" - and it costs them nothing, compared with a professional translation, which costs them something. Great!
Reasonably... See more Sorry to state the glaringly obvious, but we are not all experiencing the same changes.
My German clients already thought the English they learned for years in school, on the Internet and on holiday was good. Now they have a tool that makes it better. All they have to do, in their opinion, is read the machine translation and "check it's translated right" - and it costs them nothing, compared with a professional translation, which costs them something. Great!
Reasonably, they don't have the same confidence when it comes to Mandarin, Japanese, Serbian or other less familiar languages.
The quality of the machine translations also still differs from language to language. So clients using machine translation from Mandarin, Japanese, Serbian etc. into German can see that it's not perfect yet. But the translation from English reads like decent German.
We're all at different stages in a process that will go different ways at different speeds depending on our language combination and and specialities. It makes no sense to argue as if it's all down to our translating skills and no more. ▲ Collapse
| | |
|
|
|
Dominic D.
Local time: 22:18
Serbian to Russian
+ ...
Anne Brackenborough wrote:
Sorry to state the glaringly obvious, but we are not all experiencing the same changes.
My German clients already thought the English they learned for years in school, on the Internet and on holiday was good. Now they have a tool that makes it better. All they have to do, in their opinion, is read the machine translation and "check it's translated right" - and it costs them nothing, compared with a professional translation, which costs them something. Great!
Reasonably, they don't have the same confidence when it comes to Mandarin, Japanese, Serbian or other less familiar languages.
The quality of the machine translations also still differs from language to language. So clients using machine translation from Mandarin, Japanese, Serbian etc. into German can see that it's not perfect yet. But the translation from English reads like decent German.
We're all at different stages in a process that will go different ways at different speeds depending on our language combination and and specialities. It makes no sense to argue as if it's all down to our translating skills and no more.
Indeed. I can tell Polish -> English is near-excellent with DeepL, and the other way it's very good to poor, depending on the field and quality (clarity) of the source. The main issue remains consistency of terminology and phraseology, but these issues will probably be resolved with the advent of AI.
| | | | | Thank you for your comments! | Jul 24, 2023 |
Dan Lucas wrote:
I don't think anybody here is pretending that "everything is all right". Some people are doing fine, some are not doing so well, and some have left the profession. The business - like many others, not coincidentally - has changed over the past 20 years and will continue to do so.
"MT translation won't seriously affect translators / won't improve" is another way to pretend things are all right – because if it's never a threat, why should you care?
Dan Lucas wrote:
But the uniting and collective effort stuff has, I am afraid, no grounding in economic reality. Translation is a global, deregulated industry, with no requirement for a physical presence at the location to which the service is provided. Regulation at the country level tends to be rare and presumably would in any case have very little effect. If somebody wants a manual translated from German to English, no regulations in, say, Germany, would apply to a translator in the US or Japan.
There's only so much we can do, but does that mean we shouldn't try to organize at all?
Anne Brackenborough wrote:
Sorry to state the glaringly obvious, but we are not all experiencing the same changes.
(...)
We're all at different stages in a process that will go different ways at different speeds depending on our language combination and and specialities. It makes no sense to argue as if it's all down to our translating skills and no more.
Exactly, Anne. You said everything. Usually, the less samples a language has, the less machine translation can be trained, which impacts negatively on the quality.
Romance languages, for instance, are heavily used by the UN and WHO, which train machine translations with their materials.
Japanese and Chinese do have many texts that can be used to train those systems as well, but what complicates matters is their ideogram writing system – Japanese, for instance, tends to leave a lot implied by the context. Language models fare better in these contexts, but "traditional" machine translation systems do not.
[Edited at 2023-07-24 13:18 GMT]
| | | |
jyuan_us wrote:
It has been almost 5.5 years since the initial post, but MT in general hasn't been much improved. It probably won't be much improved in another 5.5 years.
[Edited at 2023-07-23 10:39 GMT]
I thought it was improving exponentially?
| | | | | Pages in topic: < [1 2 3 4 5 6] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Poll: Do you think DeepL is something for human translators to worry about? | 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 » |
| | 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 
