The next DPI — how India can commoditise AI, Pg10

Key Highlights:

• India aims to replicate its success in Digital Public Infrastructure (DPI) for identity, payments, and data by commoditizing Artificial Intelligence (AI).
• The country's current position in the global AI race is described as a "digital quarry," exporting raw intelligence and importing finished AI at high costs.
• The IndiaAI Mission has been launched with an outlay of approximately ₹10,372 crore to build a public-private partnership model for AI infrastructure.
• Under the mission, over 38,000 GPUs have been onboarded to provide affordable compute access to startups and researchers.
• A Unified Intelligence Interface (UII) is proposed as an API gateway for AI, similar to UPI, to ensure open and standardized access to AI models.

Next DPI.jpg

Detailed Insights:

• India's unique DPI stack integrates Aadhaar for identity, the Unified Payments Interface (UPI) for payments, and DEPA/Account Aggregator for data, enabling widespread digital adoption.
• The cost of data in India significantly decreased, bringing 500 million people online and powering the digital economy.
• The article argues that India should focus on driving down the cost of using AI models (inference) rather than competing in AI training, similar to how it commoditized data.
• The current global AI economic model is extractive, with India supplying talent, data, and usage, then importing AI services at high costs.
• The IndiaAI Mission aims to provide cheap compute power, with eligible entities accessing GPUs at a fraction of global market rates.
• Integrating AI load into the National Electricity Plan and dedicating renewable energy to inference is suggested to ensure cheap and sustainable power for AI.
• India advocates for foundational AI models to become public goods, leveraging anonymized public data (including all 22 official languages) for open-source model development.
• Any AI model developed with state-subsidized compute or public datasets would be released under an open-weights license, shifting economic value to application builders.
• The proposed Unified Intelligence Interface (UII) would provide an open, standardized interface for applications to access various AI models, abstracting complexity.
• A "freemium" model is suggested, offering free API tokens to verified Indian startups and students, subsidized by the state, with commercial tiers for scaling businesses.
• Commoditizing intelligence is expected to transform sectors like healthcare, education, and agriculture by providing affordable AI tools and services.

Key Concepts Involved:

• Digital Public Infrastructure (DPI): Open, interoperable digital systems built by the public sector or in public-private partnerships, enabling essential services at scale.
• Aadhaar: India's unique 12-digit identification number issued to residents, serving as a foundational digital identity.
• Unified Payments Interface (UPI): An instant real-time payment system developed by NPCI, facilitating inter-bank peer-to-peer and person-to-merchant transactions.
• DEPA/Account Aggregator: Data Empowerment and Protection Architecture (DEPA) is a framework that enables consent-based data sharing, with Account Aggregators being RBI-licensed entities facilitating this sharing of financial data.
• IndiaAI Mission: A government initiative with a significant outlay to build a comprehensive AI ecosystem in India, focusing on compute capacity, innovation, and ethical AI.
• Large Language Models (LLMs): Advanced AI models capable of understanding, generating, and processing human language, often used for tasks like translation and content creation.
• Unified Intelligence Interface (UII): A proposed API gateway for AI, designed to provide standardized and open access to various AI models, similar to UPI for payments.