The AI challenge: How to maximise economic impact, spread gains evenly

Why in the News

The global Artificial Intelligence (AI) race is measured in chips, data centres and computing power, with trillions of dollars invested. The World Bank Group’s World Development Report 2026: The Promise of Artificial Intelligence argues larger gains will go to those who apply AI to practical problems at scale, a route open to developing countries like India.

Why may AI’s value lie in using it rather than building it?

1. Builder centred narrative: The prevailing view expects most AI value to go to frontier model builders and upstream chip, data centre and energy suppliers, sorting countries into winners and losers.
2. General purpose technology: AI is a general purpose technology, one that spreads across the economy, like electricity. Such technologies give their largest gains to downstream users who adopt and adapt them.
3. Value from application: AI can sustain growth, raise productivity, improve public services and expand opportunity through practical use, not the frontier race.
4. India’s starting point: India is among the fastest growing major economies and has repeatedly deployed technology at population scale.
5. The takeaway: For India, the prize is not owning the best model but putting AI to work across farms, clinics, classrooms and offices.

Will AI replace or complement workers in developing countries?

1. Low automation exposure: Only 4.5% of jobs in low and middle income countries face high automation risk from generative AI (AI that creates text, images or code), against 14.2% in rich countries.
2. Productivity gains: Another 16.2% of jobs could see meaningful productivity gains, close to 18.7% in advanced economies.
3. Complement, not replace: AI will mostly complement workers in developing countries, as cheap task specific tools help farmers get advice, health workers diagnose and teachers teach.

What could stop AI’s gains from spreading evenly?

1. Concentration of AI inputs: Chips, computing, data, talent and risk capital sit in a handful of firms and countries. Competition keeps models cheap now, but relying on few providers gives them market power to set prices.
2. Missing complementary inputs: AI needs electricity, connectivity, affordable devices, local language data and skills. Governments also need capacity to evaluate, procure and regulate it.

What can India’s digital public infrastructure teach AI adoption?

1. Tools already working: Locally adapted AI supports farm advice, weather forecasting and health diagnostics. Eg. Bharat VISTAAR, the Centre’s farmer platform where States build multilingual AI chatbots.
2. Adoption at scale: Wadhwani AI’s tuberculosis screening tools show similar promise. The real question is whether millions adopt them.
3. Digital public infrastructure: Shared platforms such as Aadhaar, the Unified Payments Interface (UPI), DigiLocker and data sharing systems lower costs and enable interoperability (systems working together).
4. Open architecture: UPI’s open design avoided dependence on one provider. AI systems should likewise avoid technological lock in (users unable to switch providers) and preserve competition.

How should India build AI capability and scale it?

1. Strategic capabilities: India need not match hyperscalers (giant cloud and AI firms). It should build research, Indian language models and shared computing through IndiaAI, the Research, Development and Innovation (RDI) Fund and the Anusandhan National Research Foundation (ANRF).
2. Design for scale: Every pilot needs a path to wide adoption, covering financing, institutional ownership and integration with existing systems.
3. Test rather than assume: India’s scale makes rigorous evaluation feasible and cheap.
4. Scale what works, stop what does not: Share results, failures included, across States and ministries.

Challenges

1. Imported compute: Reliance on foreign advanced chips exposes India to other countries’ export controls.
2. Thin language data: Most Indian languages have little digitised text, limiting Indian language models.
3. Privacy and bias: AI in public services risks data misuse and algorithmic bias against groups missing from training data.

Way Forward

1. Domestic chips: Link AI compute plans to chip capacity under the India Semiconductor Mission.
2. Open language datasets: Fund open speech and text datasets for low resource Indian languages.
3. Frontline AI skills: Train teachers, health workers and civil servants to use and question AI tools.

Conclusion

India’s AI choice is between chasing unaffordable frontier models and spreading useful tools across an economy it already digitises well. Whether pilots become tools used by millions, on open designs, is the measure.

AI in India

1. IndiaAI Mission (2024): Run by IndiaAI under the Ministry of Electronics and Information Technology (MeitY) with ₹10,371 crore, to make “AI in India” and “AI work for India”.
2. Seven pillars: Compute, applications, AIKosh datasets, foundation models, FutureSkills, startup financing, and safe and trusted AI.
3. Language platforms: Digital India Bhashini offers speech and translation tools across 22 Indian languages.

Matching Previous Year Question

“[2026] Which of the following statements with regard to Large Language Models (LLMs) used in machine learning is/are correct? 1. LLMs assign probabilities to the next possible words and then pick the one with the highest probability. 2. LLMs process data through mathematical optimization to minimise prediction errors. 3. LLMs produce unbiased outputs. (a) 1 only (b) 1 and 2 only (c) 2 and 3 only (d) 1, 2 and 3 ANSWER: B”