إذا كان الذكاء الاصطناعي يستطيع الترجمة فورياً، فلماذا نتعلم لغة أخرى؟

From live speech translation in video calls to auto-dubbing on TikTok, the technology to dissolve language barriers has arrived. Real-time translation powered by artificial intelligence (AI) is now embedded in everyday life.

Tools from OpenAI, Meta, Google and many others now offer near-instant translation across dozens of languages, and they keep improving.

All this raises a vital question. If machines can do this faster and more accurately than humans, is investing years in learning another language still worth it?

The logic is appealing. Humans have always offloaded cognitive work onto tools. Writing reduced demands on our memory. Calculators removed the burden of mental arithmetic. AI sits within this long tradition. Used well, it can support learning and expand access in ways that matter enormously.

But there’s a difference between using a tool to extend your capabilities and using it to avoid doing something altogether. That distinction becomes important when you are not just replacing a skill, but a form of cognitive and cultural engagement.

The effort is the point

Effort plays a central role in how we acquire knowledge.

Psychologists use the phrase “desirable difficulties” to describe challenges that may feel inefficient, but produce stronger long-term retention and understanding.

Struggling with grammar, searching for the right word, or constructing meaning across multiple languages engages brain networks that support memory, attention and cognitive flexibility. Over time, they consolidate knowledge far more deeply than passive exposure.

Sustained mental engagement contributes to what researchers call cognitive resilience – the brain’s capacity to maintain function as we age. Managing multiple languages is one form of this engagement. It requires the brain to resolve competition, monitor context and adapt dynamically.

These are not trivial demands. And they’re difficult to achieve if you just use translation tools passively, such as resolving the meaning of a foreign phrase with the click of a button.

What multilingualism research actually shows

The evidence on multilingualism is often presented as a simple “bilingual advantage”, a shorthand that obscures a more complicated picture. Some studies report benefits for attention or working memory, while others find no differences. The truth appears to be more selective.

Our recent study examined cognitive performance in 94 adults aged 18 to 83, using both visuospatial and auditory tasks across working memory, attention and inhibition. Put simply, we looked at how people process and respond to information they see or mentally map out in space (visuospatial) and information they hear (auditory) . Examples include remembering sounds, focusing on visual patterns, or ignoring distractions.

Our study measured multilingualism as a spectrum, not a category. This allowed us to capture diverse language backgrounds and experiences. Multilingual participants spoke a range of languages with varying levels of proficiency and daily use, reflecting the linguistic diversity common within multicultural communities.

Across most tasks, multilinguals and monolinguals performed similarly. However, one pattern was striking. Individuals with richer, more diverse multilingual experience showed markedly better performance in visuospatial working memory. These effects were most pronounced in older people.

This suggests that multilingual experience doesn’t broadly enhance cognition, like some headlines claim. Instead, it may help preserve specific functions over time.

Separate population-level research has also linked multilingualism to later onset of Alzheimer’s disease and better overall ageing outcomes, though the mechanisms continue to be debated.

Overall, however, it appears that sustained use of multiple languages represents a form of mental activity with effects that accumulate across a lifetime.

What AI translation can’t replicate

AI translation excels at speed and accessibility. For many practical purposes, it works remarkably well. But it operates through pattern recognition, not lived understanding. It can struggle with cultural context, humour, register and emotionally embedded meaning, especially for languages with less representation in training data.

At best, AI captures literal dimensions of language while missing social ones. Consider the scene in the 2003 film Love Actually where Jamie, played by Colin Firth, delivers an awkward but sincere proposal to Aurelia in broken Portuguese.

It is moving because of the effort, vulnerability and intent his imperfect words carry. Resort to real-time translation software and what remains is information, not expression.

This is the deeper distinction: translation is not the same as participation. Learning a language involves understanding how people think, their values, and how meaning is shaped by context and history. This cultural literacy develops through interaction and experience. We can’t fully outsource that to systems that translate on demand.

The multilingual participants in our research spoke to this directly:

I definitely think in Telugu, but I remember numbers and count using English. Afrikaans is the language of my heart and best used to express intense emotion. English is the language of business and used mostly in everyday life.

These are not descriptions of switching between translation modes. They are descriptions of inhabiting different selves.

AI will continue to change how we engage with language learning. It can personalise instruction, minimise barriers and provide feedback at scale. What it can’t do is replace the cognitive and cultural work that comes from learning a language. This work leads to a deeper relationship with how other people see the world, and with how you express yourself. And that difference still matters.

Olivia Maurice completed her PhD at the MARCS Institute, Western Sydney University.

Mark Antoniou receives funding from the Australian Research Council.