This is an extract from “The ecosystem“, the first report of the four-part series, “Asia’s AI agenda”, by MIT Technology Review Insights.
India in some ways has natural AI assets, including deep technology research and a large population, to rival those of China. But multiple structural issues, including a lack of coordination between development agencies, threaten to slow its progress.
“India’s AI is still in a nascent stage,” says B. Ravindran, professor and head of the Robert Bosch Centre for Data Science and AI at the Indian Institute of Technology, Madras, although he is quick to point out that the country’s indigenous technology expertise does provide some useful building blocks. “We have the world’s biggest IT services sector, which has provided Indian firms with a lot of experience in the development and use of automated customer service management systems. Similarly, this has helped firms develop machine learning processes and Indian language processing—both by Indian firms and multinationals like Microsoft and Google. If you move away from industry, you see a lot more language modelling and machine learning theory development in academic circles—not at the scale as in China, but steadily growing over the last decade.”
“The lack of structure and coordination around data presents a significant obstacle for the development of an Indian AI ecosystem. Even though India has a huge population and volumes of data, it is not recorded in formats amenable to sharing and use.”
B. Ravindran Professor and Head of the Robert Bosch Centre for Data Science and AI Indian Institute of Technology, Madras
India’s higher education system provides a solid base for AI, but could be further strengthened, notes Ravindran. “The top-tier research institutes in India are focused on math and theory. So, you have lots of engineers who are really good at math, but not necessarily the hardcore computational aspects of it. And this is what gives (AI research) its additional impetus. Without computers and data—two important resources which India lacks—doing research in theoretical machine learning is less productive.”
The lack of structure and coordination around data gathering and sharing present a significant obstacle for the development of an Indian AI ecosystem. Even though India has a huge population and we can read large volumes of data, the data is not recorded in formats amenable for sharing and use, he says. “Privacy and security issues in India mean that it is hard to get our hands on large volumes of external data, which limits the AI work that India can do. That’s changing, with more international partners coming to India willing to share data, and work with teams here,” and potentially share cloud computing resources.
Ravindran notes that in many ways India’s startups, particularly those aligned with global cloud computing giants like Amazon or Microsoft, have an advantage over academics. He advocates something akin to what Canada has developed: “build a nationwide computing infrastructure for AI. And so different research groups in the country can apply for time, and the amount of time they get is proportional to their contribution.”
There are signs that government readiness for driving AI solutions is increasing. “Not only is the government interested, but it is making a lot of noise, and I think mostly the right noises. Departments that were not willing to share data a couple of years back are now falling over themselves trying to help. The only concern that I have is that India has too many of these AI task forces; every single government department is putting out one. There has to be some kind of convergence to make them effective.”