Centre for Internet & Society

Radhika Radhakrishnan, attended a Roundtable Discussion on “The Future of AI Policy in India” organized by the Indian Council for Research on International Economic Relations (ICRIER) in New Delhi on July 1, 2019, to arrive at actionable recommendations for promotion of AI in India.

Radhika's inputs primarily focused on - capacity and skilling for AI adoption in India, sectoral opportunities for the adoption of AI, regulation of explanations for AI, fairness and bias in AI models, and actionable recommendations for government priorites for AI policies in India.

Concept Note

India’s Artificial Intelligence moment is truly here and now. At a time when a diverse range of applications based on AI are being developed, pushing its frontier further into uncharted realms of business and society, Indian policy makers are contemplating not just AI’s potential for growth and social transformation, but also its proclivity to create divides and inequality. Our study attempts to understand the impacts of AI and trace the pathways to realizing it.

AI’s transformational potential stems from its ability to lend itself to a diverse range of applications across a range of sectors. One can witness AI based applications in traditional spheres of manufacturing, which are transforming quality control, production lines, and supply chain management, and in services, which are creating personalized product offerings and high-quality customer engagement. AI applications are also common in sectors such as agriculture that have taken a back seat in technological innovations in the post-industrial world. AI also demonstrates potential for impacting developmental challenges by responding to societies’ immediate demand for healthcare, education and expanding access to finance and banking.

The consequences of AI diffusion stem from AI’s pervasiveness across society, its ability to trigger innovation, and its tendencies to undergo transformation and evolution. These are typical characteristics of a class of technologies that can be found across history, the emergence and diffusion of which have enabled the wealth of nations. These are called General Purpose Technologies (GPT). Technologies such as steam engine, electricity, computers, semi-conductors, and more recently the Internet, can all be conceived as belonging to the GPT class of technologies. Our study is based on the understanding that the implications of AI can be best understood by viewing AI as a GPT.

Historically, the economic impacts of GPTs have not been immediate but follow after its diffusion across the economy, i.e. over a period of time. There are two reasons that explain this phenomenon: firstly, in early phases of technology diffusion, an economy diverts part of its resources from productive activities to costly activities aimed at enabling the GPT. For instance, organizations adopting computers must also invest in training employees or hire computer scientists, re-arrange production activities or organizational structures to accommodate computer driven work-flows, all of which are costly economic activities. Secondly, it is only after the GPT is diffused and widely used in the economy that the statistics measuring GDP start counting and fully measuring the GPT.

Empirical research on GPTs such as AI, including ours, means confronting the challenge of measurement. Estimates on the economic impact of AI are bound to be imprecise because data on AI’s adoption is not available or adequately reflected in the data used to compute economic growth, at least not yet. Measuring the economic impact of AI is also difficult because of the magnitude of indirect effects on productivity that GPTs trigger. It is not therefore uncommon that studies on GPTs, while attempting to estimate their economic impacts, also engage in in-depth case studies and historical analysis of its impacts.

Our findings show unambiguous and positive impacts of AI on firm level productivity across sectors, although there is variation in the magnitude of positive impacts across sectors. We complement our findings with case studies that cover different firms that are developing AI based applications across a range of sectors to understand the underlying firm-level capabilities that drive innovations in AI based applications. Our study leads us towards high-level policy challenges facing organizations, civil society and government, and which when addressed enable the full realization of economic growth triggered by AI.

However, our conclusions are a step-away from actionable policy recommendations. Given your experience with and within India’s AI based ecosystem, we invite you to deliberate and recommend insights and strategies that can help us arrive at concrete and practicable policy recommendations towards achieving a growth and welfare enhancing AI-based ecosystem in India.

Proposed Questions for Deliberation

  • In which sectors do we observe an immediate opportunity for the adoption of AI? What could be the nature of these applications?
  • In which areas of AI development and application is there an immediate opportunity for governments, industry and academia to collaborate?
  • What should be the Government’s top five priorities in the next one year to catalyse the growth of AI in India?
  • How and what agencies of the Government should be involved in implementation of India’s National AI mission?
  • What aspects of the Government’s capacity requires enhancement to adapt to challenges of a growing Indian AI based ecosystem?
  • What measures can the Government take to regulate for AI safety and ethical use of AI?
  • What are the policy measures that the Government can undertake to safeguard against the consequences of AI based inequality?