Language as a "Side Effect" of Learning: What's Left When AI Translates Everything
I imagine the year 2030: I'm sitting in a Tokyo restaurant, wearing Meta Ray-Bans with a camera and microphone, and the waiter is speaking Japanese. My glasses translate in real time—the menu suddenly "becomes" Polish.
And then the question I keep hearing today pops up again: why invest 1,000 hours in learning a language if technology will handle translation for me? The answer is simple: learning a language was never just about the language itself.
When Translation is Instantaneous, Real Communication Begins
A translator in your glasses can handle words. It cannot handle what most often determines success in real-life situations: cultural context.
In Japan, simply "understanding the sentences" doesn't solve the dilemmas that appear in an instant:
- how to behave at the table and what is appropriate,
- whether to eat sushi with chopsticks or hands,
- which conversation topics are neutral and which are out of place,
- what not to say in business, so as not to hit someone's historical or political sensitivities.
This is especially evident in conservative environments (like in Great Britain). You can "have a translation" and still fail in the first five minutes, simply because you don't read the norms and codes of behavior.
Technology is Great… Until It Stops Working
The second element is plain reliability. Everyone knows those moments when:
- there's no internet on the train,
- there’s no network at a conference,
- your device glitches at the least convenient moment.
If your whole communication plan relies on a device, a single failure can throw you into an "alien world" without a map.
There's also privacy. In a business conversation, a very practical question arises: is it really worth handing over nuances and sensitive content to an external system that "listens," processes, and transmits data further?
Most Communication Is Nonverbal Anyway
There's something else that gets lost at the very start in "textual" translation: nonverbal communication. Most meaning is conveyed not in words but in dynamics, gestures, pauses, and tone.
When I learn a language with people (e.g., with native speakers), I also learn the "movements" of a culture:
- Southerners (Italians, Spaniards) communicate differently—more expressively,
- Nordic and Germanic peoples are often more reserved,
- In British business, gesture and form control often matter, but there’s also a second pole: spontaneity and linguistic "roughness".
That’s why it makes more sense to talk not about "languages," but about broadly understood communication.
Online Language Learning Works—As Proved by a Real Case in Education
The University of Linguistics (WSL) in Częstochowa switched to online teaching over 20 years ago—not out of fashion, but out of student needs.
Those were very concrete stories:
- illness,
- a broken leg,
- wheelchair use and architectural barriers,
- regular dialysis,
- work trips,
- no way to commute from small towns to a bigger city.
Online was not a “gadget”. It was a way for someone to be able to study at all and genuinely change their life.
AI as a Safe Testing Ground: Less Shame, More Attempts
Scientific materials in Nature (2025) highlight one of the biggest barriers in speaking: fear of judgment. People are afraid of making mistakes, rolled eyes, or sighs.
AI doesn’t do that. Thanks to it, you can:
- repeat endlessly,
- practice pronunciation,
- train dialogues without social stress.
But this same mechanism brings a trap: it’s harder to make an effort for someone who doesn't judge. This is where the teacher’s role comes back—not as a “checker,” but as someone who can spark motivation and sense the person on the other end.
An experienced lecturer can spot needs instantly—sometimes after just one 45-minute session:
- this person has something to say but is blocked,
- this one understands but stays silent,
- that one needs a push, while another needs space so as not to get discouraged.
AI is task-focused. People see context.
If AI Handles 50% of the "Boring Work," the Teacher Gains 50% Real Work
AI is great at checking grammar, pronunciation, and spelling. There are also findings that such tools can save teachers up to 50% of their working time—especially with repetitive tasks.
That’s a good direction, as education has plenty of “ironing”:
- clicking in systems like Moodle to extract data,
- manually creating exercise variants,
- monitoring progress that can be measured automatically.
When technology lifts administrative burdens from people, teachers can more quickly:
- create smart materials,
- diagnose who needs support,
- strengthen those who are thriving,
- and… leave those who do great alone (which is also a skill).
Personalization Isn't a Luxury: Numbers Show Lack of Fit Can Set You Back
In experiments on listening comprehension using ICOL systems, an interesting contrast emerges:
- the AI-using group improved their results by 11 percentage points,
- in the traditional control group, as many as 30% of students scored lower on the final test than on the initial one.
That's a practical argument that the "one level for everyone" model is not only ineffective but may even be harmful for some people.
In real-life teaching, personalization already happens—it just takes up teachers' time because, for example, you have to prepare four sets of exercises. AI can generate those sets faster, but you still need a person who knows who to give which set to—and why.
An Essay Written by AI Doesn't End Education—It Just Changes the Rules
The classic "write an essay" task can now truly be "checked off" by a generator. But that doesn’t have to mean the end of homework.
Lecturers can build assignments so that:
- the model "spits out" a poor or overly general answer,
- the student needs to ask further questions, refine prompts, make corrections,
- and ultimately use their own reasoning, or else the text will be easy to spot as formulaic.
This leads to a healthier approach: AI as process support rather than a thinking crutch. Even if someone struggles to write, they can analyze generated sentences and learn constructions from concrete examples.
The Skill AI Won’t Replace for a Long Time: Holistic Thinking and "Love for Culture"
The hardest skills to replace are holistic competencies:
- understanding full context,
- working with nuances,
- interpreting culture, history, literature,
- ability to adjust the message to the person and the situation.
A good language teacher at levels A1–A2 still must know the language much better—if only to knowingly not use constructions that would hinder a student's understanding. That’s knowledge “beyond the tool”.
And the beginning of effective language learning is often something AI doesn’t have: emotion. People come to Spanish through soap operas, music, food, and dance. They come to English through the culture and literature of Great Britain. That energy is best conveyed by a human.
Finally, there’s another practical point: most models work best in English. That’s why a sensible strategy is: learn English and learn AI—that combination will be marketable for a long time.
Conclusions (Specifically)
- Real-time translation does not replace communication: norms, gestures, tone, taboos, and context.
- Technology can fail, and in business there’s also the issue of conversation privacy.
- AI is great at repetitive tasks and can free up even ~50% of a teacher’s time for actual work.
- Personalization is essential: in ICOL data, AI brought +11 p.p., while in the traditional group 30% regressed.
- Online learning can work if it's live, with controlled processes, and stationary exams.
- The most future-proof are holistic competencies and motivation based on emotion and culture.