Global AI Summit in Delhi

The fourth Global AI Summit is being held for the first time in the Global South, with an attendance of around 35,000 delegates. Hosted in Delhi, this summit reflects themes that highlight the orientation towards public purpose over private profit, marking a new approach to AI in India.

Digital Public Infrastructure (DPI)

• Concept: DPI emphasizes public interest rather than public ownership and aims to create more competitive markets. 
  • It is seen as a potential model for inclusive and representative AI.
• Role of the State: The state plays a decisive role through policy and funding, despite DPI's focus on markets. 
  • The tension arises in balancing public investments with the risk of crowding out private investment.

AI as Infrastructure

• Infrastructural Interpretation: Some AI technologies are foundational, supporting a wide range of downstream uses. 
  • They function as neutral, open platforms for innovation, akin to roads and bridges.
• Investment Strategies: Governments are advised to avoid focusing solely on frontier language models and data centers due to variable demand and illusory autonomy. 

High-Quality Datasets

• Importance: The curation of high-quality and representative datasets is emphasized, with a need for robust governance to protect privacy and rights. 
  • Such datasets could be sourced from DPI systems, like payment data.

Alternative Infrastructure Models

• Innovative Models: General-purpose translation modules and open speech-to-text systems for low-resource languages are identified as promising avenues. 
  • Examples include initiatives like Bhashini in India and Masakhane in Africa.

Targeted Public Funding

• Specific Applications: Targeted funding is justified in cases of market failure, like agriculture chatbots and health surveillance tools. 
  • Public funding should extend to training and organizational integration, beyond just procurement.

Conclusion

The spread of DPI offers lessons for global delegates, highlighting the importance of infrastructure-based investments. While infrastructure-first approaches are recommended, governments should also pursue targeted interventions for public-interest goals not addressed by private actors.

Authors: Narayanan is a professor of computer science at Princeton University, and Kapur is a Visiting Fellow at Princeton’s Chadha Centre for Global India.