Missing in India’s AI growth plan is private investment
The zeal to boost AI is welcome. But it is not sufficient. Our research indicates that the bottleneck for growing India’s AI capacity is private investment.
On artificial intelligence (AI), the government appears to be moving at a frenetic pace. This month, plans were announced to make large public datasets available to Indian businesses. The government also wants to embed AI in different parts of India Stack, and fund three centres of excellence for AI, housed within leading academic institutions.
The government also wants to embed AI in different parts of India Stack, and fund three centres of excellence for AI, housed within leading academic institutions. (Shutterstock)
The government’s zeal to boost AI adoption is welcome. But it is not sufficient. Our research indicates that the bottleneck for growing India’s AI capacity isn’t public funds, it’s private investment.
Take India’s investment in research and development (R&D). When compared with advanced economies and China, India is in the bottom tier in R&D spending, both in absolute terms and as a percentage of GDP. The US invests almost $670 billion or 2% of its GDP. China spends more than $500 billion, more than 2% of its GDP. R&D in India gets a paltry $68 billion or roughly 0.7% of the country’s annual output.
In innovative economies, R&D spending by the private sector swamps that by the government or academic institutions. In India, it’s the opposite. The public sector stumps up almost two-thirds of the country’s research funds. The indifferent record of India’s academic institutions in taking ideas from the lab to the real world is another reason why more public funds aren’t the answer.
R&D can be classified into three categories: Basic research, applied research, and experimental development. Ideally, R&D spending should follow a graded profile, with basic research at the bottom, followed by a larger share of applied research, and finally, experimental development. This indicates a good rate of technology transfer from the lab to the field. The US spends around 15% of its R&D budget on basic research, 20% on applied research, and nearly 65% on experimental development. India invests around 24% in basic research, 37% in applied research, and 32% in experimental developments. Given the low rate of tech transfer from academia to industry, the way to spur innovation isn’t by spending more on academic research.
Finally, leadership in AI is no longer being decided in universities. The computation power required for large-scale AI experiments has shot up by more than 300,000 times in a decade. This has imposed a high cost on academic institutions for research and experimentation. During the same period, contributions from academia have plummeted by 60%. This trend of cutting-edge research, moving from academia to industry, is most visible in the hottest form of AI right now — large language models (LLMs). These models underpin the intelligence of chatty bots from Open AI’s ChatGPT to Google’s Bard. By some estimates, the largest language models that are free and publicly available are smaller than those developed by industry by a few orders of magnitude. Access to large computing resources is a privilege only a few academic institutions can afford.
Disruptive change, it appears, will come from the private sector. But leadership, or even competence, in AI isn’t cheap. One way for India’s tech sector to make an immediate impact is by launching a privately-funded research lab that works on foundational models for AI. This lab, which we may call BharatAI, has the potential to become the hub of India’s AI innovation ecosystem.
To overcome the mismatch between high upfront costs and a long horizon to recoup the investments, we propose a pooled investment approach. BharatAI’s backers can include India’s large tech services companies, deep tech investment funds, and private endowments. The company should also have a platform partner — either Microsoft, Google, or Amazon — whose investment in BharatAI can be in the form of computer credits.
The company itself would focus on foundational AI problems with broad applicability and not attempt to develop end-to-end applications. It should instead provide tools through application programming interfaces (APIs) and open-source or licence its models.
BharatAI can serve as a talent magnet for high-quality, high-demand engineers. A research lab that is closely tied to the industry will also promote a culture of innovation that is privately led. By seeding talent and research capabilities for Indian technology companies, BharatAI can kickstart the AI flywheel.
A single company cannot alter India’s AI trajectory. But the payoff, if it succeeds, is worth considering this idea seriously.
Shailesh Chitnis is an ex-entrepreneur and a fellow in high-tech geopolitics at Takshashila Institution. This is the second of a two-part series on India’s AI preparedness
E-Paper
