By Danny K. Rajee
I remember one of my friends stated that she thinks in English. It was 2007, we were presenting our papers in New Delhi. Later I realised that even machines are created to think in English and Machine Learning’s genesis is the English language. I kept telling myself, “The Machine thinks in English!”
That moment has stayed with me because it made me ask a question I haven’t been able to shake since: what would any of this have looked like if English hadn’t become the world’s shared tongue?
The answer, I suspect, is that it would have looked a great deal slower, messier, and far less accessible to most of humanity.
We are living through one of the most consequential technological shifts in human history. Artificial intelligence is no longer a science fiction premise. It writes code, drafts legal documents, assists surgeons, tutors schoolchildren in rural villages, and translates bureaucratic forms for refugees who speak six words of the host country’s language. It does all of this with a fluency that would have seemed miraculous to anyone born before the internet age. And almost all of it runs on the backbone of English.
This isn’t an accident. English is the dominant language of scientific publishing — well over half of the world’s peer-reviewed research is written in it. It is the language of global commerce, of aviation, of diplomacy in rooms where thirty nationalities are trying to agree on something. When the engineers who built the foundational large language models went looking for data to train their systems, they found it overwhelmingly in English. Wikipedia, Common Crawl, Reddit, news archives, academic papers, legal filings, novels, manuals — the sheer tonnage of high-quality English text gave these systems something no other language could match: depth.
That head start matters enormously. The models that emerged from that training understand nuance, context, sarcasm, ambiguity, and metaphor in English in ways they simply cannot yet replicate in Swahili, Bengali, or even Mandarin, despite Mandarin being spoken by more people. Language models are, at their core, statistical pattern-recognition engines. The more text you feed them in a language, the better they get at that language. English had centuries of a printing press, a century of mass literacy, and decades of internet dominance. It arrived at the AI era with an enormous structural advantage.
For those of us who grew up speaking English as a first language, this can be easy to take for granted. But travel to a village in Assam or a township outside Nairobi, and the stakes become immediately clear. A farmer in Meghalaya who can type a question in English to an AI assistant gets a detailed, nuanced, accurate response. The same farmer asking in Khasi — his mother tongue, the language he dreams in — might get something garbled, incomplete, or simply wrong. The machine, for all its sophistication, is still far more comfortable in the English tongue.
We must keep this in mind: AI models are predominantly trained on English-centric data. While top AI models are multilingual, their proficiency varies significantly, leading to lower quality, less creative, and less accurate outputs in many other languages, especially in technical or specialized fields.
This is a profound and uncomfortable irony that journalists, educators, and policymakers need to sit with honestly. English did not become the world’s lingua franca through merit alone. It arrived on the back of empire, of trade routes enforced by warships, of missionaries and administrators who told entire civilisations that their languages were inadequate. And now, in the twenty-first century, the technology that promises to democratise knowledge and opportunity is in many ways replicating that old hierarchy — rewarding fluency in English and penalising those who never had the chance to acquire it.
At the same time, it would be dishonest to ignore what English-language AI has genuinely enabled for hundreds of millions of people. Students in West Africa who use AI chatbots to get feedback on their essays because their schools have no qualified English teachers. I have met small business owners in Southeast Asia who use AI to write professional emails to foreign clients, cracking open markets that were previously inaccessible to them. For these people, English and AI together are not instruments of exclusion — they are ladders.
“The English language is nobody’s special property. It is the property of the imagination: it is the property of the language itself.”Derek Walcott
The English language, whatever the complicated history attached to it, has become the operating system of global knowledge exchange. And in the age of AI, that role has intensified rather than diminished. To understand this technology — to use it effectively, to critique it meaningfully, to regulate it sensibly — you need to engage with it in English. The most important debates about AI safety, ethics, bias, and governance are happening in English-language journals, conferences, and policy documents. Those who cannot participate in that conversation are being shaped by decisions they had no hand in making.
What is the writer’s responsibility in all of this? I think it is the same as it has always been: to name the thing clearly, without flattery and without panic. English is enormously powerful in the AI era. That power brings real benefits and real injustices, often simultaneously, often to the same person. English is a working global language.
The machines learned to think in English. The question now is whether the rest of the world will be given a fair chance to be heard by them — and whether those of us who already speak the language will use our advantage to push for that, or simply enjoy the tailwind and say nothing.
I have spent years speaking and writing about the world as it is. I would like, before I am done, to write a few stories about the world as it ought to be. A writer from the Hills of Shillong!
