Yet, the conversation has focused relatively less on how emerging and developing countries could scale up and tailor AI to their own distinctive context.
India, like all countries, faces several choices relating to the development and utilization of AI. Prime Minister Narendra Modi, on his recent visit to the US, stressed India’s policy is to promote “AI for all.” This would entail making AI inclusive for workers and firms.
Like any technology adoption problem in economics, AI faces the quintessential risk-return trade-off, compounded by concerns of equity. Here, I will not dwell on the development of AI, an area that is better left for technologists, and will instead focus on its economic implications in the Indian context.
As the most populous country on the planet, the first implicative question for India is: How to make AI inclusive, and a growth tool for the vast pool of its workers?
India is a labour-abundant country, and will still be expanding its productive cohorts, even while most developed and even emerging countries like China will be contracting theirs in the coming decades.
Basic economic thinking, therefore, would entail a choice of products and production methods that are comparatively labour intensive. Yet, ironically, India is increasingly using capital-intensive technologies, and labour in comparison has contributed less to India’s growth trajectory.
India’s export baskets, in fact, have undergone a shift from labour-intensive products like textiles to capital-and skill-intensive ones such as engineering goods.
AI may replace human jobs. This debate, however, is about the magnitude of it and how much of it would be Schumpeterian creative destruction that would bring greater good over time. There is much uncertainty around estimates of AI’s impact.
It is impossible even for AI to get a one-handed economist for tackling this question! Estimates of the International Monetary Fund, for example, suggest that 40% of global employment is exposed to AI.
Economist Daron Acemoglu, on the other hand, is way more conservative and estimates that only a quarter of AI-exposed tasks will be cost-effective to automate within the next 10 years, implying that AI will impact less than 5% of all tasks.
Hence, the principal question is: How to make AI complementary to labour and not a substitute? This is an empirical question, where the net effect would reflect an amalgamation of several forces. Economic history is replete with such challenges.
Earlier automation had raised similar if not identical questions. Robotics affected countries differently based on each country’s management.
Evidence pointed towards large job losses at least in the short-run in the US, while South Korea managed it more deftly with skilling and retraining programmes. Today’s challenges may be even more profound for developing economies with profuse labour.
How about employers? According to the 2023 IBM Global AI Adoption Index, 59% of Indian enterprises with over 1,000 employees have actively deployed AI. This is the highest adoption rate among the countries surveyed.
The share of large firms in India, however, is minuscule. There are more than 64 million micro, small and medium enterprises (MSMEs) in India, with micro-enterprises accounting for 97% of them.
MSMEs contribute one-third of India’s GDP and close to half its manufacturing output, providing employment to 110 million people. AI has a scale problem in India. How can AI be engaged by MSMEs?
There are interesting examples within India and from the rest of the Global South. Almost half of India’s working-age population is employed in agriculture, where AI is already playing a complementary role to enhance farm productivity. The PM-Kisan chatbot, for example, launched by the Indian government, is encouraging.
This AI chatbot makes India’s direct benefit transfer scheme for farmers more accessible through voice-enabled provision of information on their eligibility, status of applications, payments and grievance redressal.
The government could invoke this in an array of programmes, like the Public Distribution System (PDS), where eligibility, access and digital literacy are critical issues.
AI has also started to figure in what is called as ‘agripreneurship.’ Regen Crops is an example of an AI-first agritech startup; it uses AI to develop models that can certify the use of sustainable practices by farmers, while helping them get better prices and secure carbon credits.
These are all important for addressing productivity and remuneration problems in agriculture, for enticing youth back to farming (not merely as a fallback option), and to help in meeting the UN’s Sustainable Development Goals by means of a sustainable and healthy food system.
This is a dramatic improvement over old processes that require auditors to manually assess a harvest through its entire cycle, and gains can be made on the ease of doing agri-business.
There are examples outside India too. Several sub-Saharan African countries have used AI to help the poor. In Sierra Leone, for example, AI is used effectively in education. TheTeacher.AI has tailored ChatGPT to the local curriculum and made instructions accessible to the poor even where (and when) internet connections are poor.
Training and skilling teachers was critical to increase their productivity and make the initiative successful. ChatGPT is not all about doing aided assignments without doing brick-and-mortar or pen-and-paper work.
Machine learning, in fact, is often used across Africa to target cash transfers. In all these cases, the successful use of AI required adapting the new technology to local needs and circumstances.
These anecdotes of success aside, the fundamental question is: How can AI-use successes be scaled up, and that too in a cost-effective way? A scale-up of AI technology can be costly, with huge amounts of computing and power infrastructure needs, more so as everyone chases this upsurge of AI.
The problem then is of opportunity cost, which is of particular relevance to emerging economies where electricity, like other resources, is scarce.
How can AI become a tool to complement and make labour an attractive factor of production, especially in developing countries facing an employment challenge? The Global South needs to scale up economic research on AI even as it tries to scale up the technology’s adoption itself.
The role of policies is critical. Any structural change—let alone one arising from a truly life-changing invention—is bound to create winners and losers. This is not the first time we have faced this conundrum. How to compensate losers is a problem faced in policy perhaps on a daily basis.
Even for reforms in common areas such as trade, globally, adjustment programmes to support the losers have been limited or ineffective, because of design or resource issues.
Not surprisingly, this led to a significant social and political backlash against globalization; and we have seen policymakers around the world resorting to trade restricting measures, with concerns about fragmentation on the rise.
If the AI experiment must be successful, equitable and sustainable in the end, policymakers should pay careful attention to policies that can help adjust to this truly transformative change. This is especially important in the context of a developing economy.
These are the author’s personal views.