Will AI replace Humans in translations ?
Will translators be
replaced by AI ?
by Guillaume
As the
globalisation process progresses and the world feels smaller every day, the
need for translation has been greatly increasing since then. The invention of
new means of transport allowed us to travel anywhere in the world in less than
24hrs. This change has led translation demands to skyrocket. Translation
has enabled culture exchanges that changed our interactions with each other and
helped us learn about the world. It is nearly impossible to imagine the
translation sector collapsing in that context.
However,
since the invention of the computer, the web and other technologies that have
drastically simplified our lives, some people think that machines will take
over the industry and completely replace humans. Technologies have greatly
improved in such a short amount of time, that progress seems to be limitless.
Recently,
Google have invented a prototype of AI, that is able to have a conversation
with humans. The machine’s speech is natural to a point which the person it
called didn’t notice she was talking with an AI. Its capacity to react to its
interlocutor’s sayings was perfect.
IBM’s
new prototype would also be an outstanding proof of AI’s potential. This
machine is able to have a debate and make arguments based on its interlocutor’s
sayings. Most of the people spectating the debate were impressed, by the AI’s
ability to formulate relevant arguments, based on articles and references it
found on the web. The prototype’s language was also smooth and natural.
Hearing
that, you may be wondering “Why am I undertaking a translation master‘s degree
if machines will replace us soon ?“ Easy Tiger, you don’t have to worry about
your future or the decisions you’ve made so far. There are several logical
reasons why machines will not be able to replace humans in the translation
sector.
The IBM
prototype I presented above, display an issue that outline the whole problem
that machines come across in translation. They are programmed in a very
specific way, to have appropriate responses regarding the domain in which they
are evolving. It means that they’re not versatile. 700 journalists were
spectating the debate to judge it and in the end IBM’s AI didn’t win. Most of
the AI arguments were found on internet because it cannot develop arguments by
itself.
There
are also concerns about the cost of developing such complex AIs. It takes
billions of euros just to create the program of one of these imperfect
prototypes. The technology is not ready for sale thus casual customers wouldn’t
be able to afford such an expensive tool. It wouldn’t even be useful for them
in their everyday life. Casual people need translations when they travel
abroad, for casual and simple things in predetermined contexts. AI may be of
use for this kind of interactions as it is a predetermined domain and thus
simple to translate. On top of that, nobody hires a translator for a weekend in
Venice just to ask for a pizza in a restaurant.
Nevertheless,
most of the demand in translations concentrates in specific domains that are
complex. Even professional translators struggle to perform translations in
these domains, as the vocabulary used is unusual. Most of the time translators
must conduct research in the subject in order to choose the appropriate terms.
Precision is the basic requirement for these tasks.
Being
precise and accurate in such a wide variety of subjects is a very difficult
task for computers. Investing such high amounts of money to create imperfect
machines to replace translators wouldn’t be profitable for companies.
AIs are
good at dealing with matters that are set in stone such as math. On the other
hand, they struggle to manage subjective and variable matters such as
languages. Some contexts do not allow mistakes, for instance, the
interpretation of a diplomatic debate. A misunderstanding may cause serious
consequences on diplomatic relations between both countries. There are software
made to simplify the task of professional translators. However, it is well
known that an overuse of these tools implies consequences on the translation’s
quality. It is thus most unlikely that machines would replace humans in the
translation process overall. Though, it could take a bigger part in certain
domains, such as simple “travel translations” which doesn’t require high and
diverse skills.
Guillaume Vancampen
This is a really interesting article. I just have one question for you: when you mentioned the Google AI project, were you talking about Google Duplex? Because I saw a demonstration on YouTube - Google Assistant called a hair salon and managed to take an appointment as a human would have done. I was quite impressed, so that I did some research about it and I found out that Google was working on that project for several years and that Google Duplex is already available on iPhones and Android smartphones running with version 5.0 or newer since March 2019. For now, this AI project only deals with online booking systems, but according to me technology is evolving quicker than we think. I am not pessimistic enough to say that AI will replace humans in translation, but I do think that it will radically change the way translators are working. One day, a text may be translated entirely by machines and humans will only have to review it. ‘Translators’ will then be ‘post-editors’… Who knows?
RépondreSupprimerFloriane F.
Yes I am, indeed talking about google duplex. Thank you very much for pointing that out.
SupprimerYour insight is very interesting but I do not share it, maybe we will have the chance to talk about it someday.
Guillaume.V
I found your article really interesting and I agree with you on the fact that AI will never be able to manage complex situations. The field of democracy is probably one of the best examples because professional translators will always be needed in this domain. How about the medical field? I think the use of AI in this domain could also lead to disastrous consequences. In a recent study by BMJ (the British Medical Journal), the neural machine translation system of Google Translate was put to the test in the medical field. 10 medical statements were translated by the tool into 26 different languages. Overall, only 57.7% of the translations were deemed correct by native speaking translators (they focused only on the sense; grammar mistakes were allowed). In some situations, there were serious mistakes; for example, “Your child is fitting” translated in Swahili to “Your child is dead.” In Polish, “Your husband has the opportunity to donate his organs” translated to “Your husband can donate his tools. I think this is another example of how AI translation can be dangerous and could even directly endanger people's lives.
RépondreSupprimerClara M.
In this modern world many think that the age of AI will take over all human activities. But in the first place it is the humans that have created the AI. At present too, using translation tools give you a result which may send a very different message from what we are trying to convey. So its better to rely on humans fro translation services.
RépondreSupprimerTranslation Services In Pune | Certified Translation Services In Pune | Translation Company In Hyderabad | Translation Services In Hyderabad