The race for AI supremacy, with its implications for the fortunes of countries and organisations, is a continuous process. Like the Western world and China, India too needs to build large models that enable it to have an independent sovereign AI foundation. This foundation will give nations and organisations the advantage to shape their future in the global context. It will also involve massive and sustained investments.
For India, a consistent focus and investment in small AI models could deliver significant economic advantage. Small AI models aimed at distributing intelligence, rather than producing intelligence, could hold greater promise. Distilled from large models and trained for specific domains, with local factors and vernacular languages, they could hold the key to the economic dividend.
In this context, the focus of AI has to be more on economic development than extraordinary reasoning ability. A small manufacturer being able to plan inventory with just-in-time replenishment, a woman artisan being able to predict demand for her products, or a farmer being able to maximise yield through the right mix of crops and weather predictions are examples where AI based on local factors can be transformational.
The MSME segment, accounting for 31% of India’s GDP, could immensely benefit from specialised small AI models. Its problems are already well understood, and there is a concerted effort to develop small AI solutions for different industries. These solutions, customised for each industry and focused on production, sales, distribution and product development, need to be affordable and easy to adopt. That can translate into significant productivity gains.
Instead of one large AI assistant capable of handling every query, individual assistants could be equipped to handle specific functions of an organisation. This would have been unthinkable for small firms. Small AI models could address these needs precisely and make them affordable. But this will happen only when AI models are designed around small businesses, rather than taking large AI models to small companies and expecting them to adapt.
The arrival of generative AI has also raised concerns about job losses. However, the recent World Development Report presents some interesting facets. It states that 14.2% of jobs in high-income countries are significantly exposed to generative AI, compared with only 4.5% in low- and middle-income economies. At the same time, 16.2% of jobs in developing countries could benefit significantly from AI through productivity gains.
For India, therefore, the question is not how many jobs AI will eliminate. It is how many ordinary workers and enterprises can become more productive with its help. For small AI models to work here, innovation in practical adoption will be important. Solutions need to be developed within limited budgets and made available in vernacular languages. They should work on low-cost devices, with limited connectivity and through voice rather than text. In many cases, data also needs to reside within the organisation rather than in the cloud. This can lower hosting and running costs while addressing concerns around data ownership and exposure.
India has the talent and capability to develop small AI models. Just as it has demonstrated its ability to develop and implement digital financial applications such as UPI at scale, it could become a beacon for other developing countries. Many of them face similar challenges in agriculture, healthcare, small businesses and education.
India may not need to build the biggest AI models to capture the biggest AI opportunity. It needs to make intelligence accessible, affordable and useful at scale.
The writer is chairperson, GTT Foundation.