The Future of Translation: A Matter of Patterns or Delicate Craft?

As the famous Chilean translator, Marina Orellana, once said, “Translation is not transliteration, that is, the transcription of words from one language to another. (…) What really matters is to grasp the ideas and express them successfully.”

If this is, indeed, the essence of translation, then the craft of the linguist would seem to resemble that of the poet, whose delicate craft consists in conveying images and ideas that transcend the formality of the material text to produce something more elusive and ethereal: meaning itself. Especially in literature, but also present in almost any form of artistic expression, meaning is an ever-changing and evolving entity. How could machines, incapable of making any kind of creative association, keep up with the wordsmith’s irreplicable spark of creation?

Expecting an application to make a heartfelt interpretation and translation of a text is, to me, as pointless as expecting a toaster to compose an original jazz piece.

Yes, machines, as we now know them, can be programmed to establish and associate all kinds of intricate patterns at incredible speeds, gradually minimizing the chance for mistakes. But, breaking patterns in a chaotic and erratic way, capable of creating new possible associations is something only humans can do. For now…

This year companies like Google have started implementing the use of AI in the way their engines process translations, improving the way in which such platforms associate entire segments (instead of single words), potentially having the entire world wide web as a constantly evolving language database.

This is where the clash between man and machine starts to get interesting: Artificial Intelligence and Neural Networks have started to replicate the process of “understanding” to a point computers have developed the ability to “reason” and “learn” from their own mistakes, allowing their capabilities to evolve within a certain pool of knowledge.

It would seem we are gradually arriving to what many Sci-Fi movies warned us so much about: the day humans become obsolete and replaced by machines.

I can’t deny that mentioning such human characteristics while referring to electronic devices gives me an odd sensation to say the least. The gleam of HAL’s red eye slowly singing “Daisy” in 2001 Space Odyssey sheds an ominous glow of caution on what seems to lie ahead.

But do not fear fellow linguist! The threat of Skynet in the guise of any sort of MT knocking on your door, like the Terminator, ready to blow your head off is still far down the road. Perhaps, it may not even come to pass.

The key to survival, it might seem, lies in our imperfect humanity, for the way we chose to communicate is as imperfect as our own human nature, something self-perfecting algorithms might fail to understand.

It’s true that technology is a force to be reckoned with that is constantly on the move, and it’s likely that someday these programs get substantial improvements to what they can offer us today. They will probably produce acceptable translations, but there’s no doubt that they will need a certain degree of human intervention to make them into something more than just empty lifeless shells. For it doesn’t matter how fast a microchip is or how much information it can process in a second.

Computers will never understand the meaning of a text as a human being do because they lack the ability to create them in the first place. So far, no computer has been able to replicate or program that mysterious spark of comprehension that takes place inside the human brain.

Automatized translation, even the most sophisticated, will remain a useful and convenient tool. At its best, translators could use these programs to pre-translate repetitive and tedious technical texts, such as user manuals, weather reports, minutes (anything with predictable patterns), which they can post-edit and improve later on. But more often than not, the best and quickest solution will be to translate everything from scratch, using the aid of a CAT tool.

The idea that machine translation software can replace translators is, in a way, almost utopian. For brain activity can be mimicked, yes, but the true soul of a translator remains unique and un-replicable.

In fact, if such day ever comes, computers will not only replace translators, but the entire human race.

Skynet, it seems, has bigger fish to fry.