Search

Will Artificial Intelligence replace translators in the future?

Updated: May 8, 2019

Will Artificial Intelligence replace our flesh and blood translators in the future? Yes, we are talking about this eternal debate once again.

It’s an interesting topic, but we feel safe since we doubt machines will ever be able to produce certified translation services!

At a time when technology is advancing faster than ever, I decided to launch this question on a platform where users express their opinion. I know that the translation industry is quite skeptical about this because it is convinced that the human factor in the art of translation can never be replaced by a mere machine.

However, most of the people who answered this question came from the field of business and technology, and, as you can read a line below, they do not fully agree with what we, flesh and blood translators, think of the subject.

I share your answers in this first part without further ado:

Gerard (Software Engineer and Investment Analyst)

The New York Times last year published an excellent article on how one of the leading companies in the field of Artificial Intelligence, Google, has managed to transform Google Translate and reinvent itself using the new revolution in the field of Artificial Intelligence: neural networks. It can serve as a reference to know where we are and where we are heading as Artificial Intelligence continues to develop and inhabit spaces that until now only we could occupy.

Neural networks are nothing new, after all the perceptrons, which are the basis of any neural network, have appeared in academic papers since the 1960s. But like many great theories or discoveries, even decades later practical applications were found and the necessary computational resources were available for their proper execution. Over the last decade, significant advances have been made in their optimization and applications. And while there is still a long way to go, advances in the area are becoming more frequent and Automated Learning algorithms are everywhere.

Google itself published a paper explaining the new update in Translate, using what it calls Neural Machine Translation (NMT), and the results are impressive.

The gap between human translation and Google's neural network model is minimal. But Google is not the only company making significant advances in automated translation, big names like Microsoft, Baidu, Yandex... and [insert the name of a Silicon Valley startup] are also doing their thing. So I wouldn't be surprised if by 2020 the gap between humans and AI, in this particular area, has been bridged. Otherwise, just wait a little longer... The one thing machines won’t be able to offer are certified translation services.

Charles

For the utilitarian translation? Without a doubt. A translator may check out a Grisham or Perez Reverte book in 20 languages as they are written. In fact, they will write them for them, so that the translation will come as standard.

Translations of real writers, a Proust or a Joyce? We will be given translations that will allow us to write a soap opera script about the texts, but not the feeling of reading an original provided by the real translator. As for providing certified translation services, no way. Government bodies would never accept them, however good they were.

Edgar, Director of Macroeconomic Strategy

Yes, we can already see it with companies like Skype looking at how to translate calls to sell the service to companies. Since there would be high demand for this type of service, there is an incentive to do so.

Carlos, Degree in Computer Science

The principle of technological uniqueness (when AI equals a person's intelligence) is likely to be reached around 2025. But at the translation level, we are bordering on the limits of the existence of devices capable of performing simultaneous translation at market prices.

Adrian

Yes, of course. All you have to do is look at Google Translate. Although it is true that it has many errors, especially in things like expressions or ready-made phrases, also sarcasms or ironies that we do not understand in our native language, all these things tend to go quite badly with the Google translator.

The translator has quite a few faults but it gets better as the years go by and I believe that within twenty years, we will be able to have almost perfect translations of the main world languages: English, Spanish, French, German, Chinese or Russian.

As we can see form the above, technology is advancing by leaps and bounds and news of new automatic translators that translate in real time are increasingly on the agenda. Now, will the quality become the same as that of the flesh and blood translators of all life?

We continue with these question times to professional experts in technology and robotics, whose opinion interests us very much since we work hand in hand with technology on a daily basis.

Raul, Director of the National Robotics League

No doubt there will still be sworn translators, but in general the idea that you need to know

5 languages will be a thing of the past. In fact, three days ago an there was an article in a major newspaper arguing against the idea that translation could be robotized, and it is surprising to see the columnist patting himself on his back with regards to how important it is that the translation is done by human beings or that "speaking several languages" is “culture”, as if this were 1990. Leaving aside the political uses of language, speaking several languages is what separates us, not what defines us as human beings. As human beings, we are defined by our personality, our physique, our hobbies. And translation? Let machines do it. Teach the kids that they're different from the neighbour's because they speak another language? It’s time to evolve from nationalism to globalisation.

Languages were invented to tell people "you are the enemy of that one, because you speak another language and are accountable to another ruler", it is part of the exclusive nature of the human being, of the "you’re not from around here" prejudice.

Andrew, Apple

With the new technologies of Automatic Learning such as Deep Learning, in recent years important leaps have been made in many areas including Natural Language Processing. More than translations, studying Linguistics (especially the one that uses modern computer techniques) is an excellent option for those who are looking to find a place in today's job market. Some applications of these technologies are help systems such as Apple's Siri, Amazon's Echo and Google Now, but also the automated emotion recognition of Twitter messages, detection of false news, or even many of the problems that day by day Quora engineers are solving (detection of repeated questions, aggressive participants, etc.). Returning to the question, I do believe that automatic translators will continue to improve and statistically (if this has not already happened) will obtain in the near future better performance measurements than most human beings. Now, this doesn't mean that they

will always do it better than people, especially if large groups of translators (crowdsourcing) are used and added in some intelligent way to get a refined final translation. There are many concepts and abstractions in our languages (and way of thinking) that a machine can hardly understand, (no matter how many examples it has been trained with), several examples of this can be found in the excellent documentary "Animated Conversations with Noam Chomsky" by Michel Gondry, which I highly recommend in order to understand a little about the limitations of these systems.

John, Telecommunications Engineer.

Yeah, it's all a matter of time. Machines are able to learn very quickly based on human translations. There will come a time when all the phrases in the world have been translated into all the languages, and computers will be able to record them all and automatically evolve so that the new translations are perfect.

Google Translate has already improved a lot in the last few months using artificial intelligence, so it's easy to imagine what it will be able to do in 10 or 20 years.

George, Writer, historical and sociological researcher. Sooner or later they will, but they need to learn human reasoning. Once a month, Google sends me a series of phrases in Spanish and their translation into English or French so that I can give my opinion as to whether they are correct or improvable. That's how they teach the algorithm. Unbelievable!

Phillip, Engineer, entrepreneur, information technology evangelist Considering that there are an infinite number of quality translations done by people between language A and language B, and an enormous amount of those translations are really terrible... my answer would be: YES.

Just watch a movie originally made in English and see the subtitles or dubbing into Spanish. They're usually VERY bad. Literal translations are made without any context and without the translator really understanding what the characters are expressing in the original language.

This is really the difference between simply (so-called) knowing a foreign language or being "bi-cultural"... it is a HUGE difference and extremely important in a context of translation from one language to another, especially if we are talking about books, films, TV series...

Returning to my answer: based on the above, I believe that YES, automatic translators will be able to translate with excellent quality between certain languages. We could say that in some cases they already do so, perhaps with specific languages.

In addition, translators DO learn, human translators do not necessarily learn "well"...

As in many other things, it is necessary to speak of specific translators, both automated and specific human translators with name, surname and nationality in order to be able to make valid judgments... otherwise, all this is mere speculation and generalization which, of course, is not entirely valid.

Gabriel, works in Communication Agencies

Oh, yes, I do.

But the question is not whether or not, but when...

The new generation of AI allows 'machines' to learn, as long as there is someone (or some programme) that 'feeds' that learning.

If every time a programme translates something wrongly, someone, or another program, corrects it, computers will be able to incorporate that new information, perfecting the accuracy of the translation.

Danny, Freelance translator No way. No way.

With this very, very strong statement, I do not want to jettison all the advances in machine translation in recent years. In fact, MT is a very useful tool that translators have today.

But on the other hand, language is constantly evolving. On top of that, with social networks, new terms appear that define new activities (tweet, selfie, to give a couple of examples). In addition, there is the issue that in each country there are different variants, which at the same time incorporate these changes, but in a different way. Nor can the context of information be left aside; translating a scientific text may contain standardized terms, but an advertisement targeting different countries must use the language of the local public.

I have no doubt that MT can deliver quality, localized translations for each country. But what the machine lacks is the brain, and it is this that allows you to realize the trends, which words are dealt with today and which are out of fashion, false friends, irony, sarcasm, puns and intentionality of the author.

Therefore, I do not believe that the same quality will be achieved. A similar one, yes, but the human brain will always be needed to make these texts closer and more understandable to the target audience.

In addition, machines will not be able to carry out certified translation services, which is what many people need.

(future translator robot, if you're translating this answer, be nice to me).

Christian, 10 years in the Digital Production Industry, Web, Software Natural language processing (NLP) studies precisely how to do this, and they are already very advanced, and there are many examples of this.

Carlos, Electronics and computing technology: electrometers, ionization chambers

It could be that the translator is learning different ways of interpreting, which in the end is the goal of the translation, that it is understood well and in a similar way. On the other hand, there are so many idiosyncrasies that generate their own words and expressions that are difficult to reconcile. The best example is what we see in Latin America. It's the same language as in Spain, but you move from one country to another and it's a totally different thing, as far as the most commonly used things are concerned. When you travel there, the first thing is people don't understand you, or you don't understand them, and you’re talking about the most common things. Don't you believe me? Ask for bread, in any country of Latin America, as you would ask for it in yours, and you will see. Or ask to eat a dish in a restaurant, you will be surprised.

It will be very difficult for automatic translators to appear. I think that the most sensible thing would be to form a committee by nationalities and place a series of options to choose and validate them. At some point you will reach a model that resembles something understandable and not like the current translators that can give you nonsense.

And so far the answers received. As we can see, these are very different opinions that the reader does not have to share. I'd love to know yours! Get I touch with any thoughts or if you need our high-quality, lowest price certified translation service. As ever, thanks for reading.

0 views

© 2018 Freelance Translation Group

Birth certificate translation? Marriage certificate translation? Divorce certificate translation?
Whatever your needs for standard or official translation, the freelancers at Certified Translation UK can help, for the best price, guaranteed.