Real time translation: mediation technology or human augmentation?

Ten years after leaving Germany, I still ask friends and family to bring German newspapers when they visit. I enjoy the challenge of deciphering them, with the typical German super-long sentences that only reveal their meaning at the end. This process feels more fulfilling with pen and paper. Over the years, my friends have learned this habit, and now I don’t even have to ask - they just bring Lesematerial to me.

Recently, a friend brought me a Die Welt edition with an article by Yves Bellinghausen: Muss ich noch Sprache lernen? (Do I still need to learn languages?). This kind of reflection has become prominent in our times, not only regarding language learning. It’s the same dilemma with coding: Should someone still learn how to code properly, or wait until AI can really create everything with just a prompt?

The article touched on the history of translation technology, from the first “Babel Fish” used at the Nuremberg trials to today’s advancements. What particularly intrigued me was how languages naturally go through cycles of unification and diversification. Take Latin, for instance - it once acted much like English does today. But after its widespread use and consolidation, it diverged into vernaculars, eventually giving rise to Italian, French, Spanish, and others. Even now, English’s global influence often leads to local adaptations and new forms. In Portuguese, for example, “notebook” refers to a laptop, while “face” colloquially means “Facebook.” It’s a new language being formed.

Language is created by the needs of expression. Thus, it must be a function of context. When two people interact, they create new language - negotiating terms that become available for their future conversations. Think of “inside jokes” or “jargon,” unique linguistic branches within a specific group. Anti-colonial psychologist Geni Nú?ez characterized relationships as “grammars” in her book, a metaphor that aligns closely with my lived experiences.

Unifying all languages into one? It’s not just impossible - it’s pointless. Context will always demand diversity in expression, and divergence would eventually occur anyway.

Real-time translation? For sporadic interactions - like going to a hairdresser in Poland, as described by Yves Bellinghausen - it’s incredibly useful. For consistent, longer-term interactions, like negotiating, working, or living abroad, it’s less about translating words and more about understanding cultural nuances and ways of thinking. Real-time translation without inference of meaning wouldn’t substitute our language skills, because language skills = all the “mechanical” aspects of language (grammar, pronunciation, syntax, etc.) + the culture they represent. I do think, though, that such tools could change how we learn new languages.

Duolingo? It might lose its purpose, as it largely focuses on teaching languages for those sporadic interactions. However, its exceptional user engagement could redefine language learning as a hobby, even without significant practical applications. Their investments in AI and in new subjects such as music are likely a realization of such market changes.

The challenge with real-time translation isn’t the technology - Google Translate is now remarkably good. The problem lies in design. Not in the obvious burden of getting your phone and opening the app, but in how it currently disrupts the dynamic of a dyad. Current tools act as a “third party” in conversations, mediating them. For real-time translation to succeed in scale, I believe it needs to shift from being thought of as a mediator to an extension of individual skill - a seamless companion that augments us, instead of something that stands between us.