How AI Can Help Protect Minority Languages

A piece I wrote in July discussing the impact of generative algorithms on translation, “Has AI Finally Toppled the Tower of Babel?”, caught the attention of my friends at the Center for European Policy Analysis (CEPA). After extensive editing and adding some context about Spain’s recent request for Catalan, Basque and Galician to become official languages in the European Union, it was published on CEPA’s website with the title “Will AI topple the European Tower of Babel” (pdf).

Now, Marcos Sierra, from Spanish online newspaper VozPópuli, has used it in an article in Spanish (pdf) in which he proposes that Spain’s Congress use automated algorithmic translations instead of human interpreters to enable the use of these three languages in its sessions.

Until now, automated translations, despite having improved significantly over the last few years, provided results that, at best, simply gave an idea of the content, without much nuance, and sometimes choosing words that made no sense . This was because these programs initially worked through a text word by word, then learned to do so sentence by sentence and then paragraph by paragraph, while the more sophisticated ones were eventually able to apply some generic context.

With the advent of generative algorithms and massive computing power applied to their training, translation programs have improved by leaps and bounds in recent years, and those that are now being used by the general public literally improve by the day. Many of us have tested the power of generative algorithms such as ChatGPT or Claude to correct and edit our own texts and then translate them into other languages, and were impressed by the quality of the results: no strange words or dubious syntax. Although a good human translator may still be advisable for some texts, much of the time, an algorithm is sufficient.

What’s more, these algorithms can copy our own writing style if fed with texts we’ve authored, to the point that they read as if we had written them, even when translated into another language.

Where do we go from here? First of all, many translators are probably reconsidering their future, and to what extent they can add value to a text. I would imagine that somebody working as a simultaneous translator, such as those employed in parliaments like the European or Spanish parliaments, sees their job as inherently automatable. Other translators who are able to carry out more thoughtful, careful work requiring greater creativity to introduce nuances or even improve the original text so as to make it read as though it been written by a literate native speaker are likely to remain in demand.

However, all this is based on the premise that languages are more than mere communication. For me, they are also repositories of culture, tradition, identity, and filled with subtleties and richness we don’t want to lose by all speaking a few global languages. After all, the only languages in which we can really express ourselves are those we learned as children or have spent many years using constantly — and even then, we may struggle. For almost all of us, the most comfortable and efficient thing experience is speaking in our own tongue.

Of course, language is also political, which is perfectly legitimate: if I want to reinforce my identity, my culture or my desire for self-determination by using the language of my country or region, why shouldn’t I, if the technology allows me to?

The European Parliament’s budget for translators is more than €1 billion (less than 1% of the EU’s budget), and pays the wages of a lot of translators. Little wonder that Brussels is less than enthusiastic about admitting more languages. But to argue that “it makes no sense, because there are common languages already covered by translators” is to deny the cultural background of these languages, effectively condemning them to a slow death. Since more and more people speak English, let’s understand each other in English, right? Well, no. Why English and not another language? English has the advantage of allowing us to understand the importance of the United States or the United Kingdom, but after Brexit, Malta is the only member state where it is now spoken. How about French? Or German, which is more difficult but at least forces you not to interrupt and to listen to what the other person is saying until they have finished their sentence. Good luck with that.

The simple truth is that rather than doing away with other languages, so we end up only speaking one, we now have the option of our own personal translator: algorithms trained by ourselves, based on how we speak and write, that can render what we are saying into any other language. Not “a translator”, but “our personal translator”. This seems the ideal solution, to recognizing collective and individual identity.

Which isn’t to say there won’t still be problems: personalized automated translations can never be completely simultaneous; there will be failures: generative algorithms hallucinate, and will continue to do so. Some people may train their translation algorithm with the speech and texts of others, creating “hybrid identities” between the way a person expresses themselves in one language and in another. That said, anybody who speaks another language will have experienced the strange feeling of taking on a new identity. When I speak in French, English or Galician, even my voice changes!

In conclusion, the use of generative algorithms to preserve the identity of people and languages while allowing us to understand each other in real time deserves to be explored. A minority language should not be seen as an annoyance or a nuisance, it’s a treasure trove of culture and a rich repository of interesting flavors and human specificities. Therefore, there is no point ignoring the reality that different languages can be used in politics and other spheres, and that this tendency will increase as the technology improves. And I suspect that it won’t be long in coming.

—

This post was previously published on Enrique Dans’ blog.

***

You Might Also Like These From The Good Men Project

You Said ‘Race’, but Are You Actually Talking About Race?

Understanding the Nonbinary: Are You Confusing Gender With Sex?

The Difference Between Compassion for Those With Disabilities & Ableism?

‘Masculinity’ Is Having an Identity Crisis