AI translation challenges the value of learning languages

AI translates in real time and shakes up the Chinese business

Does it make sense to pay thousands of euros and spend four years studying so your daughter learns Chinese? That is the question hovering over the statement attributed to NVIDIA's CEO—"the end of languages, AI will translate everything"—and technology is already answering it in practice. Real-time machine translation has moved from brochure promises to daily tools: some work half a year in Japan using Bluetooth earbud translators, while others buy clothes in foreign markets relying only on their phones. The debate is no longer whether AI translates, but what remains of the economic value of learning a language.

From promise to daily use: translators that actually work

The discussion starts with skepticism toward grandiose announcements. For years, a future that never arrived was sold: flying cars, lunar colonies, the Madrid-Barcelona Hyperloop in twenty minutes. Nuclear fusion, they say, is always forty years away. With AI, however, the leap isn't a prediction: it's a reality used daily. One participant reports communicating in Japan with an instant translator in earbuds, noting its popularity among locals. Another sent a screenshot of a Japanese video game to a language model and received an image translation.

The qualitative leap compared to translators from a decade ago is evident. Languages that previously stumped machines, like Persian, are cited, with claims that current models grasp context with minimal explanation. A merchant summarizes the practical consequence: automatic translation opens a truly connected European market, where buying cheap in one country and selling high in another no longer requires linguistic intermediaries. It is, he says, Charlemagne's dream fulfilled by algorithm.

How long does it take to amortize four years of Chinese?

The most repeated case is the mother paying for her daughter's Chinese classes. The calculation handled is stark: reaching a professional level in Chinese takes about four years, and by then the opportunities justifying the effort may have vanished, unless the profile fits the top tier sought by companies. Automatic translation doesn't eliminate the language, but it erodes its salary premium.

Not everyone buys the argument. Some maintain that machine-mediated conversation generates distrust: nuances are lost, misunderstandings arise, and negotiations become slower. Others defend that studying languages isn't just communication, but a way to structure thought: those who master two languages approach problems from more angles. The circulating analogy is the calculator: its existence doesn't miccionan nobody needs to learn addition.

The dot-com precedent and AGI hype

Distrust points not at the technology, but at the financial narrative surrounding it. The parallel with the dot-com bubble repeats: back then, accumulating users sufficed to receive investment without revenue; today, promising general intelligence always around the corner suffices. The scheme, it is argued, is sustained by future promises rather than what technology offers now. AI is here to stay, but the impact will be smaller than advertised.

Against this, the response is that neither Windows nor Apple guaranteed anything in the eighties, yet today everyone uses a computer or smartphone. The fundamental question—whether general intelligence will ever arrive—remains unanswered, and the article stops there: nobody can prove it impossible, nobody can prove it close. What is discussed with data is the physical chain: without rare earth extraction, lubricant refining, and maintained power grids, no robot translates anything.

Translator as a profession in decline

The most tangible blow hits employment. A translator with over twenty-five years of experience seeking alternatives because work has collapsed is mentioned, along with an English philologist who only wants to translate, not teach. Automatic translation hasn't eliminated the profession but has hollowed out its middle rung. The dubbing sector isn't spared either: video channels show synthetic voices imitating original announcers with mixed results, anticipating that voice and accent cloning will be indistinguishable in a decade.

The counterargument is that the machine makes structural errors: it generates responses through statistical inference, not reasoning, failing at trivial tasks like counting letters or drawing fingers. Those failures have decreased but remain intrinsic to technology trained on massive data volumes. The magic lies in the data, not comprehension.

---

The most honest conclusion is unsigned: if AI translates everything, the value of a language plummets; if AI also cannot replace human judgment, the language retains value. Both things at once. And while this is decided, someone continues paying thousands of euros for Chinese that might no longer be needed.