The Next DPI: Can India Build Public AI for All?

What is the core idea of an 'AI DPI'?

The central proposition is to extend the principles of India's successful Digital Public Infrastructure (DPI) model to the domain of Artificial Intelligence. The 'India Stack'—a set of open APIs and digital public goods—has already 'commoditised' three critical layers: identity (Aadhaar), payments (UPI), and data sharing (DEPA). The argument, as articulated by proponents like Srivatsa Krishna in The Hindu, is that 'intelligence' itself should be the fourth layer. This would involve creating foundational AI models, curated public datasets, and accessible computing infrastructure as a public utility. The goal is to lower entry barriers for AI development, allowing startups and researchers to build applications without the prohibitive costs of training large models from scratch. This approach aims to make AI affordable and ubiquitous, much as UPI made digital payments a near-zero-cost service.

What is the government's position and existing framework?

The Government of India has signalled strong intent to become a global leader in AI. This strategy is being operationalised through the IndiaAI Mission, which received Cabinet approval in March 2024 with a total outlay of ₹10,372 crore for five years. The mission's objectives include establishing large-scale AI compute infrastructure, developing indigenous Large Language Models (LLMs), and creating datasets platforms. According to a Press Information Bureau release dated March 7, 2024, the mission will establish a scalable AI computing ecosystem of at least 10,000 GPUs (Graphics Processing Units) through public-private partnerships. Furthermore, an 'IndiaAI Datasets Platform' will be developed to provide researchers and startups with access to quality non-personal datasets, a critical component for training AI models. The government's rationale is rooted in fostering domestic innovation, ensuring technological self-reliance (Atmanirbhar Bharat), and harnessing AI for public services in sectors like agriculture and healthcare.

What are the potential benefits of this approach?