Although recent advances in artificial intelligence have raised concerns about disinformation, deepfakes, and looming job losses, such risks must be weighed against the likely benefits. For a large, still-young developing country like India, the technology's potential to bridge current development gaps is difficult to overstate.
FAIRFAX, VIRGINIA – In wealthy advanced economies, conversations about artificial intelligence tend to focus on the dangers posed by machine learning. But in India, the bigger and more pressing question concerns human learning. With its young and growing population – which is already the world’s largest, and not expected to peak until 2065 – India’s main economic challenge is to build human capital, as is true of many countries in the Global South.
Here, AI already shows great promise. India’s education system is in crisis. Over half of fifth graders cannot read at a second-grade level, and merely a quarter can manage simple division. If these students had a personalized curriculum – taught in their native dialect, without caste-based or economic discrimination – they could catch up. While poor incentives for educators, state-level politics, bad curricula, and socioeconomic circumstances have stood in the way of this solution, AI could make such obstacles surmountable.
Imagine an AI tutor interacting with a student from India’s poorest state, Bihar, where learning scores are abysmal, in her native Maithili dialect. It would evaluate homework through images, correct pronunciation, teach other languages, integrate numeracy through games, and offer endless, patient repetition. The same approach also could be used to offer teacher training at scale, with large language models (LLMs), like the one that powers ChatGPT, aiding curriculum development in India’s 100-plus languages and more than 10,000 dialects, all at low cost.
These AI tutors will be affordable, partly because of India’s huge market. One in three Indian students already pays for private tutoring, and well before the recent AI breakthroughs, Indians dominated YouTube, where education playlists help students master various state examinations. All the data these students provide will train models for foundational-learning tutors that can be deployed across the Global South, where students face similar problems.
But while tech innovators and the market will deliver these AI tutors, it is incumbent on policymakers, civil society, and philanthropists to provide the relevant data, digitized at scale, to help train the models. In the process, they can also start to develop alternatives to a broken state-controlled education system.
The largest gains will go to poor and middle-class Indians plugged into the digital system. Almost two-thirds of Indians have access to a smartphone in their household, and that figure is projected to rise to 95% by 2040. Indians enjoy some of the cheapest mobile-data plans in the world. Yet while richer Indians send their children to elite private schools, the poor end up in budget private institutions or in a public-school system with rampant teacher absenteeism.
Young girls, whose opportunities for schooling and learning are often limited after puberty, will also benefit disproportionately. While an AI tutor cannot overcome all the effects of gender discrimination, it can eliminate the problem of physical distance in accessing high-quality education.
AI will also augment India’s human capital through the health-care system. While high-end private care in the country is excellent, the bottom four income quintiles must navigate a system that is chronically strained, expensive, and unfriendly to the point of cruelty. Training, hiring, and deploying doctors at scale is difficult. But hospitals could use doctors’ time better by introducing AI helpdesks or physicians’ assistants that can collect complete medical histories and record symptoms and other details.
With the ability to cover all native dialects, AIs might just be more humane than the current hospital interface. Most medical diagnostics, reports, prescriptions, instructions, and disclaimers in India are printed in English, even though only 15% of Indians speak some of the language, and only 4% are fluent. As a result, most patients struggle to take the correct medicine or dosage. My own literate, multilingual, highly resourceful grandmother needed me to accompany her during hospital visits because she didn’t speak English.
The language barrier often drives patients into the arms of quacks. Fortunately, current LLMs are quickly becoming adept at diagnosing medical issues and determining whether an urgent doctor visit may be required. Uploading the images of a prescription or medication with a simple voice prompt will provide an immediate response about dosing, contraindications, or related questions.
Nor is AI-augmented health care limited to chatbots. India manufactures 60% of vaccines globally, at costs low enough to serve developing countries. But few Indian labs develop new vaccines. This could change rapidly as AI-based computation tools accelerate important breakthroughs in protein folding or mRNA vaccine development. While such discoveries are also occurring in the United States and Europe, they are not accompanied by the same scale of physical manufacturing infrastructure.
AI will also address other gaps left by weak state capacity. For example, the Indian judiciary has a backlog of more than 50 million cases pending at all levels, 85% of which are stuck in lower courts with severe staff shortages. Soon, AI can help train clerks and officers to deal swiftly with cases before they come to judges for a decision. Some countries are already allowing newer AI models to help write opinions and expedite the resolution of cases. India’s own Supreme Court is using AI to translate more than 36,000 opinions into multiple languages. Now, policymakers need to catch up by formulating a code to guide judicial officers using AI.
Weighing Risks and Rewards
Of course, listing these benefits paints too rosy a picture for those concerned about sudden changes to the Indian economy and society. One major worry, especially in the short term, is that AI is reducing the cost of producing disinformation, which spreads easily through cheap encrypted distribution networks like WhatsApp. This content often is well-written, persuasive, and can now be produced in seconds. Doctored photographs and lifelike deepfakes of public figures, including fake pornographic images of female journalists and activists, are being disseminated with impunity. Eventually, AI models will help detect fakes and false information, perhaps even tracing the origins of malicious content. But until then, more Indians will lose trust in information networks.
Another fear is job loss, given that India already struggles with youth unemployment over 40%. While India’s top computer scientists have been among those building LLMs in Silicon Valley, AI’s threat to information-technology services and the back-office sector in India has grown. Indian back-office workers may be significantly cheaper than their Western counterparts, but AI assistants will be hard to beat.
Before the pandemic, the top 20% of earners in India accounted for 45% of income and 36% of consumption expenditure. When these households started spending less during the lockdowns, India’s urban workers were severely affected. If AI forces large firms into mass layoffs, it could have a similar effect that extends beyond those with AI-vulnerable jobs.
But some of these fears are overblown. While there certainly will be some major changes in the labor market, India is no longer just a traditional outsourcing hub. India has 1,580 overseas firms whose “global capability centers” employ hundreds of thousands of people who provide more than just back-office services. They create and maintain software systems, pursue research and development, and create highly customized business solutions. The youth unemployment crisis has little to do with AI and more to do with regulatory problems and a labor force with inadequate skills. All told, India has far more to gain by adopting AI to invest in its labor force than it stands to lose from some jobs becoming obsolete.
FAIRFAX, VIRGINIA – In wealthy advanced economies, conversations about artificial intelligence tend to focus on the dangers posed by machine learning. But in India, the bigger and more pressing question concerns human learning. With its young and growing population – which is already the world’s largest, and not expected to peak until 2065 – India’s main economic challenge is to build human capital, as is true of many countries in the Global South.
Here, AI already shows great promise. India’s education system is in crisis. Over half of fifth graders cannot read at a second-grade level, and merely a quarter can manage simple division. If these students had a personalized curriculum – taught in their native dialect, without caste-based or economic discrimination – they could catch up. While poor incentives for educators, state-level politics, bad curricula, and socioeconomic circumstances have stood in the way of this solution, AI could make such obstacles surmountable.
Imagine an AI tutor interacting with a student from India’s poorest state, Bihar, where learning scores are abysmal, in her native Maithili dialect. It would evaluate homework through images, correct pronunciation, teach other languages, integrate numeracy through games, and offer endless, patient repetition. The same approach also could be used to offer teacher training at scale, with large language models (LLMs), like the one that powers ChatGPT, aiding curriculum development in India’s 100-plus languages and more than 10,000 dialects, all at low cost.
These AI tutors will be affordable, partly because of India’s huge market. One in three Indian students already pays for private tutoring, and well before the recent AI breakthroughs, Indians dominated YouTube, where education playlists help students master various state examinations. All the data these students provide will train models for foundational-learning tutors that can be deployed across the Global South, where students face similar problems.
But while tech innovators and the market will deliver these AI tutors, it is incumbent on policymakers, civil society, and philanthropists to provide the relevant data, digitized at scale, to help train the models. In the process, they can also start to develop alternatives to a broken state-controlled education system.
The largest gains will go to poor and middle-class Indians plugged into the digital system. Almost two-thirds of Indians have access to a smartphone in their household, and that figure is projected to rise to 95% by 2040. Indians enjoy some of the cheapest mobile-data plans in the world. Yet while richer Indians send their children to elite private schools, the poor end up in budget private institutions or in a public-school system with rampant teacher absenteeism.
BLACK FRIDAY SALE: Subscribe for as little as $34.99
Subscribe now to gain access to insights and analyses from the world’s leading thinkers – starting at just $34.99 for your first year.
Subscribe Now
Young girls, whose opportunities for schooling and learning are often limited after puberty, will also benefit disproportionately. While an AI tutor cannot overcome all the effects of gender discrimination, it can eliminate the problem of physical distance in accessing high-quality education.
AI will also augment India’s human capital through the health-care system. While high-end private care in the country is excellent, the bottom four income quintiles must navigate a system that is chronically strained, expensive, and unfriendly to the point of cruelty. Training, hiring, and deploying doctors at scale is difficult. But hospitals could use doctors’ time better by introducing AI helpdesks or physicians’ assistants that can collect complete medical histories and record symptoms and other details.
With the ability to cover all native dialects, AIs might just be more humane than the current hospital interface. Most medical diagnostics, reports, prescriptions, instructions, and disclaimers in India are printed in English, even though only 15% of Indians speak some of the language, and only 4% are fluent. As a result, most patients struggle to take the correct medicine or dosage. My own literate, multilingual, highly resourceful grandmother needed me to accompany her during hospital visits because she didn’t speak English.
The language barrier often drives patients into the arms of quacks. Fortunately, current LLMs are quickly becoming adept at diagnosing medical issues and determining whether an urgent doctor visit may be required. Uploading the images of a prescription or medication with a simple voice prompt will provide an immediate response about dosing, contraindications, or related questions.
Nor is AI-augmented health care limited to chatbots. India manufactures 60% of vaccines globally, at costs low enough to serve developing countries. But few Indian labs develop new vaccines. This could change rapidly as AI-based computation tools accelerate important breakthroughs in protein folding or mRNA vaccine development. While such discoveries are also occurring in the United States and Europe, they are not accompanied by the same scale of physical manufacturing infrastructure.
AI will also address other gaps left by weak state capacity. For example, the Indian judiciary has a backlog of more than 50 million cases pending at all levels, 85% of which are stuck in lower courts with severe staff shortages. Soon, AI can help train clerks and officers to deal swiftly with cases before they come to judges for a decision. Some countries are already allowing newer AI models to help write opinions and expedite the resolution of cases. India’s own Supreme Court is using AI to translate more than 36,000 opinions into multiple languages. Now, policymakers need to catch up by formulating a code to guide judicial officers using AI.
Weighing Risks and Rewards
Of course, listing these benefits paints too rosy a picture for those concerned about sudden changes to the Indian economy and society. One major worry, especially in the short term, is that AI is reducing the cost of producing disinformation, which spreads easily through cheap encrypted distribution networks like WhatsApp. This content often is well-written, persuasive, and can now be produced in seconds. Doctored photographs and lifelike deepfakes of public figures, including fake pornographic images of female journalists and activists, are being disseminated with impunity. Eventually, AI models will help detect fakes and false information, perhaps even tracing the origins of malicious content. But until then, more Indians will lose trust in information networks.
Another fear is job loss, given that India already struggles with youth unemployment over 40%. While India’s top computer scientists have been among those building LLMs in Silicon Valley, AI’s threat to information-technology services and the back-office sector in India has grown. Indian back-office workers may be significantly cheaper than their Western counterparts, but AI assistants will be hard to beat.
Before the pandemic, the top 20% of earners in India accounted for 45% of income and 36% of consumption expenditure. When these households started spending less during the lockdowns, India’s urban workers were severely affected. If AI forces large firms into mass layoffs, it could have a similar effect that extends beyond those with AI-vulnerable jobs.
But some of these fears are overblown. While there certainly will be some major changes in the labor market, India is no longer just a traditional outsourcing hub. India has 1,580 overseas firms whose “global capability centers” employ hundreds of thousands of people who provide more than just back-office services. They create and maintain software systems, pursue research and development, and create highly customized business solutions. The youth unemployment crisis has little to do with AI and more to do with regulatory problems and a labor force with inadequate skills. All told, India has far more to gain by adopting AI to invest in its labor force than it stands to lose from some jobs becoming obsolete.