| Pages in topic: [1 2 3] > | DeepL’s CEO: Algorithms won’t completely replace human translators Thread poster: Hans Lenting
|
|---|
Deep in the lake of words
The experts are unanimous: The start-up Deep-L has achieved something that one would hardly have thought possible in view of the dominance of the large Internet groups from the USA. The company has developed an online translator that beats the competing offers of Google, Microsoft and Co. by far. Deep-L is therefore considered one of the hottest start-ups in Europe. The Deep-L translation service did not go online until 2017, but within two years the number... See more Deep in the lake of words
The experts are unanimous: The start-up Deep-L has achieved something that one would hardly have thought possible in view of the dominance of the large Internet groups from the USA. The company has developed an online translator that beats the competing offers of Google, Microsoft and Co. by far. Deep-L is therefore considered one of the hottest start-ups in Europe. The Deep-L translation service did not go online until 2017, but within two years the number of users has risen significantly - to more than 40 million applications per month. In October, the start-up won the honorary award of the first German AI Prize, which is presented by the business magazine Bilanz. The fact that Deep-L can translate better than Google Translate, for example, is attracting attention not only in Germany but also internationally. Just recently, US investor Benchmark bought 13.6 percent of the start-up's shares.
Unlike many tech start-ups, Deep-L is not located in Silicon Valley, but in Cologne's Mediapark. The office in turn corresponds to the start-up cliché: glass walls, beanbags, whiteboards, sweets, a pool table. 60 employees work in the Cologne office, and the number is set to rise to 100 next year. The hierarchies at Deep-L are flat, says managing director Jaroslaw Kutylowksi. Since last summer, he has been the CEO of the start-up company. Gereon Frahling, founder of the company and former CEO, has since then concentrated on research and further development of the translator.
In principle, the Deep-L translator works like its counterparts from Google, Microsoft and others: When the user enters a text into the mask, the data is first sent to the start-up's data centers in Germany, Finland and Iceland. Supercomputers, which can process a million words in less than a second, translate the texts there. They first convert the original text into codes. They then decode these codes and transfer the text into the desired language. Then Deep-L spits out the translated text again and displays it to the user.
Faster than a human brain: The DeepL data centers in Germany, Finland and Iceland are equipped with supercomputers that can process a million words in less than a second. They convert the text into codes and transfer them into the desired language.
Google Translate currently translates into 103 languages. Deep-L only speaks nine languages: English, German, French, Spanish, Portuguese, Italian, Dutch, Polish and Russian. But soon there will be more. Managing Director Kutylowski does not yet want to reveal exactly when. What makes Deep-L different from other machine translators? The Deep-L boss is silent on this as well: "This gives us our competitive edge."
Most translation programs are currently based on so-called neural networks, explains Christian Eisold. He is a consultant at Berns Language Consulting GmbH in Düsseldorf. The artificial neurons of such a network are arranged in layers and linked together in a similar way to the neurons in a human brain. "And this artificial brain independently learns connections based on the data it is fed with," says Eisold. Deep-L says that a new arrangement of the neurons and their connections would better represent natural language than previous translation systems.
Gereon Frahling and Leon Fink founded Deep-L in 2009 under the name Linguee. They developed the first Internet search engine for translations. Frahling and Fink searched the Internet for bilingual texts using web crawlers, i.e. programs that automatically search the Internet. The translations they found were then presented to human translators to improve them. These bilingual texts form the data basis for Linguee. Today, anyone searching for word meanings via the site will find excerpts from the translations in which the word they are looking for appears. The data fed into Linguee served as the training material for the Deep-L translation service.
"Google may have more data, but Deep-L probably has better data," says Samuel Läubli, an expert in computational linguistics at the University of Zurich. He thinks it unlikely that Deep-L will use a completely different approach than its competitors. If the company doesn't want to talk about its technical foundations, this could mean two things: "Either Deep-L has a novel technical approach and doesn't want to waste this competitive advantage. But it could also be that Deep-L's only advantage is its good data base and they want to hide it.
If you use the services of the online translator Deep-L, you do not have to pay anything up to a text length of 5000 characters. In return, Deep-L remembers the texts entered in order to continue learning. "The data is stored exclusively for this purpose", assures managing director Kutylowski. With the paid version Deep-L Pro the texts are deleted directly. The premium version is available for individuals and for companies; it can be used online as well as integrated into translation programs.
In the future, companies will become an even more important target group for Deep-L, says the managing director. His vision: Language on computers should no longer be relevant. "Deep-L should then only run in the background on the computers of companies, so that language barriers no longer play a role in communication.
And yet: "My children are still learning foreign languages", says Kutylowski. Because even he doesn't believe that algorithms will completely replace human translators. Studies have also shown that the translation profession continues to gain in importance, simply because of the increasing globalization. "The market for language services is estimated very differently - between 20 and 50 billion euros," explains computer linguist Läubli. The share of translation technologies in the market for language services is between one and two percent.
"So there is great potential for machine translators such as Deep-L," says Läubli, "at the same time, the market for human translators remains huge. Because machines are often not linguistically correct enough to completely replace a human being. This is particularly problematic in legal documents or in marketing, where customer contact and creativity count. Machine translators also have difficulties with incorrect source texts or the recognition of gender.
Despite the potential: the creators of Deep-L do not want to sell their start-up company for the time being. "We currently feel quite comfortable being independent," says Managing Director Kutylowski. A move is initially planned for next year: Because the workforce will almost double, the company is moving to new premises outside the city. The offices are to become a little more colorful - matching the design of Deep-L, announces Kutylowski. But the company is also looking forward to the near future in terms of technical changes. "There will be some exciting innovations at the beginning of the year," promises Kutylowski. What these will be is of course still a secret.
Source: SZ ▲ Collapse
| | | | Tom in London
United Kingdom
Local time: 10:04
Member (2008)
Italian to English
| Very interesting but---- | Jan 2, 2020 |
Very interesting but it's far too expensive - unless you use it all the time. I only use it occasionally, and it always needs to be corrected/rewritten.
[Edited at 2020-01-02 09:37 GMT]
| | | |
Since you don't need CAT tool integration and only access to the web translator (which also comes as a Mac/PC application), you can use the Starter plan, for EUR 5.99/month.
https://www.deepl.com/en/pro.html
Data Confidentiality
Your texts are deleted immediately after you’ve received the translation
Use the Web Translator without limits
Translate as ... See more Since you don't need CAT tool integration and only access to the web translator (which also comes as a Mac/PC application), you can use the Starter plan, for EUR 5.99/month.
https://www.deepl.com/en/pro.html
Data Confidentiality
Your texts are deleted immediately after you’ve received the translation
Use the Web Translator without limits
Translate as much text as you like with your single-user license
5 document translations per month
Faster document translation via the Web Translator; fully-editable files ▲ Collapse
| | | | Tom in London
United Kingdom
Local time: 10:04
Member (2008)
Italian to English
|
|
|
Although I agree a renewal reminder would be nice, I personally don't find it a deal breaker, since you can easily manage your subscription (and cancel the monthly or annual renewal) anytime. It's literally two clicks away.
| | | | Hans Lenting
Netherlands
Member (2006)
German to Dutch
TOPIC STARTER | Impact on hour jobs | Jan 2, 2020 |
More important than pricing schemes or (missing) renewal announcements, I think, is the huge impact on hour daily work as translators.
The good news is that even the CEO of a top player in the MT business is fair enough to admit that there will always be need for human translators–though their work will drastically change.
The not so good news is that (potential) clients probably won't pick up this very important nuance (that MT will never ever be error free).
<... See more More important than pricing schemes or (missing) renewal announcements, I think, is the huge impact on hour daily work as translators.
The good news is that even the CEO of a top player in the MT business is fair enough to admit that there will always be need for human translators–though their work will drastically change.
The not so good news is that (potential) clients probably won't pick up this very important nuance (that MT will never ever be error free).
There's important work to do for us to make both the public and clients aware of this shortcoming of MT. ▲ Collapse
| | | | | we can do our bit... | Jan 2, 2020 |
by refusing to accept MTPE jobs... 
Edited for typo
[Edited at 2020-01-02 14:07 GMT]
| | | | | Algorithms won’t completely replace human translators | Jan 2, 2020 |
But I am afraid at the end they will destroy our business as we know it (that is, less jobs, lower prices and accepted lower quality. It is already happening now).
Hans wrote: 'There's important work to do for us to make both the public and clients aware of this shortcoming of MT'. I agree, but how? Let the first one who knows, please stand up, because I have no idea, or better said, nobody will listen.
| | |
|
|
|
Hans Lenting
Netherlands
Member (2006)
German to Dutch
TOPIC STARTER | ATA Position Paper on Machine Translation | Jan 5, 2020 |
If reliable and secure translation is desired, machine translation should not be used without the ongoing involvement of professional translators.
For translators, this means that their tool sets have been expanded by yet another potential resource that will prove valuable and increase productivity if used appropriately.
| | | | | DeepL’s CEO: Algorithms won’t completely replace human translators | Jan 6, 2020 |
Yeah, just like IKEA won't completely replace carpenters.
| | | | | Questions to those who started it and are involved | Jan 8, 2020 |
I am not against technology at all. But what is going on now is totally insane. It is way different than any other previous industrial revolutions. People who are pushing these hypes still do not understand that the final result (whether intentional or unintentional) of these tech-hypes will be the total downgrading of humans (devolution). Those translators who are using these tools will forget to translate by themselves, they won't have the basic knowledge how to translate: beginners won't have... See more I am not against technology at all. But what is going on now is totally insane. It is way different than any other previous industrial revolutions. People who are pushing these hypes still do not understand that the final result (whether intentional or unintentional) of these tech-hypes will be the total downgrading of humans (devolution). Those translators who are using these tools will forget to translate by themselves, they won't have the basic knowledge how to translate: beginners won't have it, experienced translators will forget it. Those lawyers who start to use algorithms to create documents will not have the basic knowledge either to be good lawyers. Same applies for physicians or scientists etc. Plus this will erase the magnificence of individuals in humans. Due to greed, rush for money and rush for popularity these CEOs and inventors and their investors push the whole world into a world where no one wants to live. Not even themselves. A world without human touch and soul.
I read somewhere that a company's CEO who started to use bartender robots is paying money to those bartenders who lost their jobs due to his investment. This is the problem: these people still do not understand that there are things that money can't buy. If somebody goes to a bar or club, she/he wants to see people. And yes even the bartender should be a human whom they can talk to, who understands their human questions or problems... who has a human mind and soul.
So
Where is the soul and mind of these CEOs and investors and inventors?
What is going on with these people?
Why are they not able to see the future consequences of their greedy investments?
and (!!!) an important question:
Why those people who are able to see the consequences do not dare to speak their mind and resist?
Where are those people who also asked these questions for themselves do not speak their mind in these forums?
[Edited at 2020-01-08 10:05 GMT]
[Edited at 2020-01-08 10:07 GMT]
[Edited at 2020-01-08 18:58 GMT] ▲ Collapse
| | | | Tom in London
United Kingdom
Local time: 10:04
Member (2008)
Italian to English
| It's the system | Jan 8, 2020 |
We live in a system in which the only ultimate aim is to generate profit. Anything - no matter how necessary it may be - that doesn't generate a profit doesn't get done.
The largest cost item in any activity is people. People are expensive, unreliable, inconsistent, and they get tired, emotional, and uppity. So we need to do away with people. That's why everything is being robotised.
The propaganda part of the system makes sure that anyone who resists this "technologica... See more We live in a system in which the only ultimate aim is to generate profit. Anything - no matter how necessary it may be - that doesn't generate a profit doesn't get done.
The largest cost item in any activity is people. People are expensive, unreliable, inconsistent, and they get tired, emotional, and uppity. So we need to do away with people. That's why everything is being robotised.
The propaganda part of the system makes sure that anyone who resists this "technological progress" is laughed at and called a Luddite.
[Edited at 2020-01-08 10:18 GMT] ▲ Collapse
| | |
|
|
|
Hans Lenting
Netherlands
Member (2006)
German to Dutch
TOPIC STARTER | Ways to use MT | Jan 8, 2020 |
Katalin Szilárd wrote:
Those translators who are using these tools will forget to translate by themselves, they won't have the basic knowledge how to translate: beginners won't have it, experienced translators will forget it.
I think that this all depends on how translators (have to) use MT. Do they have to edit projects that have been completely translated by MT (PEMT–for me my worst nightmare)? Or do they start from scratch and can create their translations themselves, with possible assistance of MT when desired? (E.g. to get typos identified, save keystrokes or even get terminological hints.) In this scenario, MT is just one additional resource. MT suggestions at the segment level can be useful–but they don't have to. It's the professional responsibility of the translator to ignore them or to use useful snippets. If he wants.
So much for the 'technical' aspect at the translator's side. The price decrease because of assumed usefulness of MT via PEMT or even worse, without PEMT, is the real problem here.
Perhaps we can convince Mr. Kutylowski to place a big disclaimer above the DeepL input boxes: DeepL's results can only give a hint of the meaning of the source text. For exact translations contact a professional translator.
But I'm not sure whether this isn't conflicting with his business model.
| | | |
Tom in London wrote:
We live in a system in which the only ultimate aim is to generate profit. Anything - no matter how necessary it may be - that doesn't generate a profit doesn't get done.
The largest cost item in any activity is people. People are expensive, unreliable, inconsistent, and they get tired, emotional, and uppity. So we need to do away with people. That's why everything is being robotised.
The propaganda part of the system makes sure that anyone who resists this "technological progress" is laughed at and called a Luddite.
[Edited at 2020-01-08 10:18 GMT]
You speak as if those who are involved in this (CEOs, investors and inventors) wouldn't be humans.
Those who are involved must be also people. They probably have children..
They are basically taking away the future of younger generations.
I'm wondering what they will say in this dystopic future to their children, grandchildren or even totally unknown younger people? How will they explain their decisions and acts? Will they say?: I did it for you? I didn't see (or did not want to see) the consequences...?
And why those people who see the consequences do not speak their mind? Cowardice? Do they think if they do not speak, the system will at least let them live and survive? Isn't this familiar from the history?
[Edited at 2020-01-08 10:36 GMT]
[Edited at 2020-01-08 11:06 GMT]
[Edited at 2020-01-08 18:59 GMT]
| | | | | I see a brighter future | Jan 8, 2020 |
Katalin Szilárd wrote:
I'm wondering what they will say in this dystopic future to their children, grandchildren or even totally unknown younger people?
[Edited at 2020-01-08 10:36 GMT]
[Edited at 2020-01-08 11:06 GMT]
Children of the future will be grateful, just as we are grateful for the technology progress in the last two centuries that enable us to lead a more enriched life, even if a lot of professions were sacrificed along the way.
| | | | | Pages in topic: [1 2 3] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » DeepL’s CEO: Algorithms won’t completely replace human translators | 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 » |
| | CafeTran Espresso | You've never met a CAT tool this clever!
Translate faster & easier, using a sophisticated CAT tool built by a translator / developer.
Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools.
Download and start using CafeTran Espresso -- for free
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
