I remembered back in 1998 when I attended the first congress on Translation and Interpreting. One of the workshops I attended was related to Machine Translation (MT). The speaker compared a text created by a human translator with another one generated by translation software. The differences were crystal clear: the human translator had produced a well-written text while the machine translation software had delivered a nonsensical word jumble. Undeniably, MT has evolved a lot since those days, so much so that some translators are afraid of disappearing from the translation scene altogether.

Be that as it may, software developers and major companies have being trying unsuccessfully to create the perfect machine translation almost since the introduction of computers. These days, the topic of machine translation is covered in practically every translation congress, seminar, and workshop organized by translators, associations and translation agencies, but MT is still struggling to jump over the same hurdles it encountered in 1998–and in the early 1950s.

Language is a complex system that includes much more than the words we can find in any dictionary. And it goes way beyond knowing two languages. It involves idioms, implied meanings, cultural references, euphemisms, irony, and a long list of other aspects that only professional translators can deal with perfectly. And trying to find these equivalents in another culture is something that only beings with brains can do. As renowned author Anthony Burgess says: “Translation is not a matter of words only: it is a matter of making intelligible a whole culture.”

Translation is an art. To perform it properly requires skills and tools such as creativity, language sensitivity, linguistic knowledge, interpretative skills and cultural background that go far beyond the mere transfer of words from one language into words from another. Under no circumstances can a machine possess all of these attributes. I have seen no convincing evidence to date that, even if a translation program were  fed all the comparative information in the world (something that would surely imply insurmountable technological and economic problems), it would ever be capable of correctly deciding, between two versions of the same sentence, which was the most appropriate one to convey a certain context. So how can we imagine that a machine will ever translate accurately?

Throughout history, technology has been replacing human beings in many fields. However, technological advances face clear-cut limitations whenever their aim is to take over activities, like translation, that involve the need for genuine human intellect and knowledge. In fact, even the most sophisticated machine still needs for a human being to press a button in order for it to start working. Human beings, it seems, simply can’t get rid of human beings, no matter how hard they try.

The translation industry indirectly forces translators to use MT, based on the notion that if they don’t embrace MT, they will disappear from the scene. Under this pretense, many translators have already started accepting post-editing tasks, unconsciously “teaching” the machine “to translate” and trying to adapt to an invented reality that seeks to do us in. From my point of view, despite the many “attractive” labels created for these translators, you simply STOP being a translator as soon as you become a post-editor. Not only because you end up feeding the machine, but also because you start doing an automatic and repetitive task that destroys your true role in the communication cycle, your role as a language and cultural mediator. And this is a doubly self-destructive and self-effacing surrender, considering that the ultimate goal the MT/post-editing mix is, without a doubt, to continue to undermine the translation profession, while filling the pockets of the handful of giant translation wholesaler agencies that are seeking complete control of the translation market.

Luciana Lovatto 
Edited by Dan Newland
Illustrated by Juan Tavella