AI and inequality: a reply to Arundhati Roy

When I started reading Arundhati Roy’s essay, “Artificial Intelligence Is Capitalism’s Final Destination“, I read the first paragraph, then the second, and then the third. I understood that she was saying something important. But I could not fully understand what she was saying. The English was difficult, the expressions unfamiliar, and the arguments needed more time than I had on my own. So I did what many people like me would probably do. I took a screenshot of the first paragraph, opened ChatGPT, and asked it to explain it in simple language. And then I continued reading. That small experience changed how I think about artificial intelligence (AI) and inequality.

I understand Roy’s concerns about AI. Technology can concentrate wealth, exploit human labour, reproduce existing inequalities and give more power to those who already have it. I am not denying these dangers. But I think another side of this conversation deserves attention: what happens when people who have historically lacked access to knowledge and guidance begin using AI?

AI and inequality: opportunity does not start equal

AI does not benefit everyone equally, partly because people do not begin with equal opportunities.

Someone who speaks English fluently, has studied at a well-resourced university and has access to mentors and professional networks may use AI to work faster. Someone from a rural area, a poorly resourced school, or a marginalised community may use the same tool to understand an unfamiliar concept, practise English, or discover an opportunity they didn’t know how to pursue.

I come from the latter world.

I use AI to understand complicated articles, discuss research ideas, explore scholarships abroad, refine motivation statements and improve my English. Sometimes I ask questions I might feel embarrassed to ask another person. Sometimes I want to explore an idea when nobody around me is available for that conversation.

AI cannot replace a good teacher, mentor or friend. Its answers can be inaccurate and biased, and I have to question what it tells me. But for someone with limited access to guidance, the opportunity to ask questions without feeling judged can be meaningful.

This does not erase the digital divide. It changes the question to ask about it: can a technology that reproduces some inequalities also help people navigate others?

The time to create

I am a performer and filmmaker, but I also have an administrative job that takes up about 12 hours of my day. Emails, reports and routine paperwork consume time I would rather spend developing a film, writing a script or thinking through an artistic idea.

AI helps me complete some of these tasks more efficiently. In a way, I am buying back my own free time using AI.

That time matters. Not everyone can afford an assistant or arrange their working life around creative pursuits. For people who must earn a living through demanding jobs, reducing the time spent on routine tasks can create room for work that would otherwise remain unfinished.

When AI is discussed as a threat to human creativity, it is worth asking who has the time and resources to create in the first place. The question is not only whether a machine can produce something creative, but whether people who have ideas can find the space to develop them.

The internet, AI and political expression

I also think about Season 1 of the Cockroach Janata Party (CJP) movement in New Delhi. The internet and social media have changed how political ideas, humour, slogans and dissent can travel beyond traditional media and established political organisations.

I am interested in Abhijeet Dipke, the CJP founder associated with this movement, and how a young political voice can reach people without the resources or networks of an established party.

I don’t know whether Dipke or his team used AI, so I wouldn’t claim it contributed to the movement. What interests me is the wider possibility: could digital tools help people new to politics research unfamiliar issues, develop communication strategies, and think through the responsibilities of leadership?

Imagine a young person who unexpectedly finds themselves with a following and asks an AI tool how to understand public policy, organise ideas or evaluate different strategies. The tool would not make that person a good leader. But it might provide a starting point that would otherwise require access to experienced advisers.

Digital platforms can widen participation, but they can also spread misinformation, enable manipulation and create new forms of control. These risks need to be taken seriously. Still, we should not discuss the power of technology only in terms of what corporations and established institutions can do with it. The debate should also examine how people outside those institutions might use it.

What the East India Company can teach about technology and power

This is where I find myself questioning the idea that AI can be understood only as capitalism’s final destination.

History offers a useful, though imperfect, comparison. The East India Company came to India as a commercial enterprise and gradually acquired political and territorial power. Its activities served its own interests and those of British imperial expansion, not the emancipation of Indians. Colonial rule involved extraction, coercion and profound inequalities.

Yet the history of colonialism also reveals a complicated relationship between systems of power and the tools or institutions that people later use in unexpected ways. Modern education expanded under colonial rule for reasons that included the British Empire’s administrative needs. But education was not forever confined to the purposes its colonial architects intended. Indians used it in different ways, including to challenge colonial rule and social hierarchies.

For Dalits, access to education became essential to the struggle against caste oppression. The life and work of BR Ambedkar demonstrate how education could become a means of intellectual independence and political resistance.

This is not to suggest that the East India Company was comparable to an AI corporation, or that colonial exploitation should be understood as a necessary route to progress. Nor should we overlook the work of reformers and communities who fought to make education accessible. The comparison is narrower: a system designed to serve powerful interests can sometimes provide tools that people use for purposes its creators did not intend.

Could something similar happen with AI? A technology developed and controlled by powerful corporations might still help a first-generation learner understand an academic paper, help a young researcher formulate a question or help an aspiring filmmaker learn a concept without paying for a course.

The fact that a technology has unequal ownership does not automatically mean that every use of it reinforces that ownership. Equally, the fact that it helps an individual does not make the larger system just. Both realities can exist at once.

The problem is not always a lack of ideas

For some people, the problem has never been a lack of curiosity or ideas. It has been a lack of guidance, opportunity, and access.

Imagine a student from a village in Maharashtra who wants to study abroad but has nobody to explain the application process. Imagine a graduate who understands a subject but struggles to express their knowledge in English. Imagine an aspiring filmmaker who wants to explore a theory but cannot afford a course or find someone willing to discuss it.

AI cannot guarantee their success. It cannot pay application fees, provide a scholarship, repair an underfunded school or replace public institutions. Nor should governments use it as an excuse to withdraw investment from education.

But it can help someone take a first step: understand a requirement, practise a skill, organise an idea or learn how to ask a more precise question. For a person who has always had teachers and mentors, this may seem ordinary. For someone who has had to figure things out alone, it can make a real difference.

That difference does not dismantle structural inequality. But neither should it be dismissed because it falls short.

I do not want AI to think for me

I do not want AI to think for me. I want it to help me think better.

There is a difference between asking a machine to make every decision and using it to understand an argument, challenge an idea or find language for something a person is trying to express. My use of AI to understand Roy’s essay did not decide what I should think about it. It helped me reach a point where I could form my own response.

Access to information has never been enough on its own. People also need ways to understand it, question it and apply it. AI can sometimes help with that process, but it also demands critical thinking. It can reproduce stereotypes, invent information and make confident errors. AI literacy, accountability, and systems that let people challenge its harms are all needed.

These are reasons to demand better and more equitable technology – not to dismiss the experiences of people who are finding value in it.

AI and inequality: who decides what it means for humanity?

Perhaps the question is not whether AI is good or bad for humanity. It is who benefits, who bears the costs, and who gets to decide what it becomes.

Society must question the corporations that own these systems. It must demand accountability for the use of personal data, protect workers, ensure environmental responsibility, and safeguard against surveillance and discrimination. It must ask who profits from human knowledge, who pays the environmental costs of data centres and whether the benefits will reach people who cannot afford the best tools.

But it must also ask what happens if access to AI becomes expensive or restricted to people who already possess the resources to use it effectively. An hour saved by a privileged professional and an hour reclaimed by an overworked employee may look similar on paper, but their consequences can differ greatly.

I began reading Roy’s essay to understand her argument. I turned to ChatGPT because I needed help with the language. With that support, I could continue reading, decide where I agreed and disagreed, and develop my own response. AI did not give me a finished opinion. It helped me reach a point where I could form one.

I do not expect AI to solve the inequalities I have lived with or replace education, friendship, political struggle, and collective action. I do, however, want the freedom to use it to learn, think and make room for the creative work I want to do.

Anyone fighting for a more equal world must be careful not to condemn a technology without listening to the people for whom it has made some part of that world more accessible.

The challenge is not to defend AI corporations or reject AI altogether. It is to ensure that people who have historically been denied knowledge, resources and opportunity are not excluded from shaping this technology’s future. Their voices should not be an afterthought in a debate about humanity. They should be at the centre.