Humans. Machines. Where are we?

Ever since UNESCO first marked International Mother Language Day in 2000 to promote awareness of linguistic and cultural diversity and to promote multilingualism and their importance when faced with the immanent risk of the extinction of many languages around the world.
While this day celebrates the past and the present, in 2022 it is looking to the future with its adopted theme “Using technology for multilingual learning: Challenges and opportunities”

While the discussion was once about “if” machines would replace humans, now it is about the “when” and the “how”. How? Through the development of Neural Machine Translation (NMT) When? Sooner than you think?

While we smile at and share the clumsy translations or memes on social media, the machine keeps on learning and improving. But what are the advantages of machine translation? And what are its disadvantages? Where does machine translation win over human translation? These and other questions arise now that the topic has become more real and worthy of closer attention.

“When an old man dies, it’s as if a library burns down” – African proverb

There have been storytellers and translators since the beginning of time. People needed, and continue to need, people who can share knowledge; who can build bridges between religions, culture, languages, whatever they may be; who can understand an original text and retain sufficient awareness to transmit the meaning and not just the words into a new language – regardless of the form it takes. But this is a slow and painstaking process that needs patience and skill. From this, we started gasping at translations provided at the click of a button.

With the recent emergence of Computer Assisted Translation (CAT), translators have seen their work simplified with the aid of useful features, such as the automatic propagation of repeated sections and the creation of translation memory databases (TMs) that enable the assimilation of translations and their use in other projects.

Now, in the early 21st century, artificial intelligence is helping computers to learn a whole range of languages so they can begin translating. And we return to the “how”: How does NMT work? What are its advantages and disadvantages and what does the future hold for the translation market?

Google Translate and DeepL Translator are just two examples of NMT, of neural networks that are very similar to our central nervous system that is the most advanced method of machine translation currently available that have made remarkable progress in recent years based on AI, Big Data and Deep Learning.

And because it is always evolving, NMT has the ability to learn new languages and refine its translations, influencing and rethinking our working methods. Nevertheless, it still needs a human being to guide and teach it about various sectors and their peculiarities and vocabulary.

But which is better? Time is money, as they say, and with the ever growing demand for translations with ever shorter deadlines, translators are often tempted to turn to machine translations – are they are cheaper, quicker and more efficient? In the end, it is almost like choosing fast food. It is food, but we still feel more satisfied with a home-cooked meal. It takes longer to prepare, but is has something machine translations cannot (yet) replicate: dialect, style, context, creativity, emotion, among other touches that are more human that continue to apply, even when reviewing the final result of a machine translation.

And now? The answer is not so simple, because machine translation is constantly evolving and will continue to be used in various areas, even as there are obstacles to overcome. While this remains the case, t’works will continue towards the future in which machine and human translations coexist, without favouring one over the other. Afterall, they both have skills that are not interchangeable… at least for now.