The Eternal Value of Learning Languages

Michaël Oustinoff

Professor in Translation Studies

Michaël Oustinoff is Honorary Professor of Translation Studies (University of Côte d’Azur, UCA, France) and the author of La Traduction (PUF, Paris, 2022 (last ed.); translated into Arabic, Korean, Japanese, Persian and Portuguese) among other publications. He is a member of the editorial committee of the journal Hermès (Cognition, Communication, Politics), CNRS, Paris, and has a command of more than 10 languages, his main ones being English, French, German, Portuguese and Russian.

AI & EDUCATION — INTRODUCTION

In a position paper by the British Academy entitled “Language Matters More and More”, languages are considered “vehicles of new modes of thought”: “Learning a foreign language has the effect of denaturalizing the native language and culture […]. Reading, more slowly, writing in foreign languages enables a different kind of thinking from that which goes on when students read in their own language.” For someone like me, who started learning languages back in the 1960s—when bilingualism was considered detrimental for a child’s development, whereas today the “bilingual brain” and its cognitive benefits are extolled all over the place—such statements still resonate with an odd back-to-the-future ring. “One day, you’ll be able to watch movies in more than 50 languages on Netflix,” I would tell that wide-eyed child. Add the Internet, and now the AI revolution, and we have every reason to believe the present and future of languages is shining bright. Or do we?

There is a looming cloud in the sky, though. Nicholas Ostler clearly anticipated it before the advent of DeepL and the like: “Language technology will take care of interpreting and translation, and foreign-language learning will become an unnecessary chore, except for specialists and enthusiasts. Active communication with speakers of other languages will no more require a special skill than is actually needed to read a foreign text in translation, or to follow the subtitles of a language video.” This could, of course, be welcomed as a blessing—as Microsoft argued back in 2016: “The personal universal translator has long been a dream of science fiction, but today that dream becomes a reality.” With over 7,000 languages spoken today, and no one in their right mind willing to learn them all, seamless machine translation does sound like good news.

A comforting though generally little-known fact is that language learning is not the preserve of the young, nor of those lucky enough to grow up in multilingual families. Kató Lomb (1909–2003) is the perfect illustration of this: a late starter who became a simultaneous interpreter in nine or ten languages, she kept learning languages throughout her life, and in ‘Polyglot: How I Learn Languages‘ she explained how she eventually came to know 16 languages at several levels of ability. “We should learn languages because language is the only thing worth knowing even poorly,” she wrote.¹ She used the tools available at her time—newspapers, radio, undubbed films, textbooks. No doubt she would have embraced today’s tools too—and placed AI literacy at the top of the list.

Bearing this in mind is the best way to take the current AI revolution in our stride. 

But blessings come with curses in disguise, and this is the heart of the matter: “There are three main risks due to overreliance on that technology: “deskilling,” or “loss of previously acquired skills”; “mis-skilling,” or “reinforcement of incorrect behavior due to AI errors or bias,” and “never-skilling,” or “failure to develop essential competencies” (ibid.). On the upside, AI represents “unprecedented opportunities” for the development of “expert practice” (ibid.), so how can we separate the wheat from the chaff? The answer put forward by the authors is fostering “AI literacy,” which only boils down to what Karl Jaspers already soberingly expressed in 1949: “This much is clear: technique is merely a means, in itself neither good nor evil. […] Technique is independent of what can be done with it, as an autonomous entity an empty power, ultimately a crippling triumph of the means over the end.”²

The issue is not to do away with AI because of the very real risks of deskilling, mis-skilling and never-skilling—though these risks are well-documented. The real question is whether AI enhances human language learning or quietly replaces it. This is precisely the challenge now facing Europe. The European Union’s AI Act addresses part of this by requiring, in Article 4, that providers and deployers of AI systems ensure “a sufficient level of AI literacy”³ across the board. With 67% of UK teenagers routinely using AI tools,⁴ literacy in how these systems work is no longer optional.

We should learn languages because language is the only thing worth knowing even poorly.

Yet AI literacy alone is not enough. One of the main hurdles is navigating today’s overwhelming abundance of linguistic resources—a striking reversal from the scarcity of the 1960s, when languages like Albanian, Basque, Hungarian, or Turkish were barely accessible outside their borders. Today, DeepL or Google Translate will switch you to Swahili, Xhosa, or Zulu in seconds. But even for the major languages of the world, which Nicolas Ostler (op. cit.) called the “Big Beasts,” it has become excruciatingly hard to make heads or tails of the profusion of educational tools—AI-powered or otherwise—at your disposal in the classroom or on your own. In such a maze, knowing how to learn with AI, rather than simply outsourcing learning to AI, becomes essential.⁵

And there is a deeper danger still: the belief that languages are interchangeable, and that a lingua franca—whether today’s English or some future Universal Translator—will simply do the trick. This is precisely what Mandela understood when he said: “If you talk to a man in a language he understands, that goes to his head. If you talk to him in his language, that goes to his heart.” In an era of powerful machine translation, that distinction matters more than ever—and it should be the fundamental tenet of any serious approach to AI literacy in language education, in Europe and beyond.

References

1. Lomb, K. (2008), Polyglot. How I Learn Languages (translated from the Hungarian by Ádám Szegi and Kornelia DeKorne), Berkeley, Kyoto, TESL-EJ, p. 37 (originally published in 1995). Kató Lomb ended up knowing 28 languages.
2. Jaspers, K. (1949), Vom Ursprung und Ziel der Geschichte, Zürich, Artemis, p. 161: “Das jedenfalls ist offenbar: Technik ist nur Mittel, an sich weder gut noch böse. […] Die Technik ist unabhängig von dem, was mit ihr zu machen ist, als selbstständiges Wesen eine leere Macht, ein schließlich lähmender Triumph des Mittels über den Zweck.” (our translation). 
3. European Union (2024), Artificial Intelligence Act, Article 4. https://artificialintelligenceact.eu/article/4/ 4. Unicef (2025), Guidance on AI and Children. Updated guidance for governments and businesses to create AI policies and systems that uphold children’s rights. 
5. Abdulnour, R.-E.E., Gin, B., Boscardin, C.K (2025), “Educational Strategies for Clinical Supervision of Artificial Intelligence Use”, Maltham, MA (USA), New England Journal of Medicine, Nov. 12, 2025, 393:786-797.