In 2016, I wrote about why data science is a huge opportunity for India and how India can aspire to be a leader in data science services. That led me to start ml-india.org – a platform to get India’s machine learning community under one roof to connect and multiply their success. Today, we have enabled 27 machine learning meetups, featured 55 companies, 11 research groups, 146 machine learning professionals and we run a mailing list of 1806 people.
ml-india.org helped us learn that India needs to do a lot more – we rank quite low on our number of papers in top AI conferences. In fact, the whole of India publishes a lesser number of papers than a single Chinese university. None of our companies engage in cutting-edge machine learning that can have a global impact. There are just a few bright spots. We need to change our approach dramatically if we wish to have a real impact of artificial intelligence on our society and economy. Recently, I wrote a white paper on how this could be achieved with an annual outlay of $100M. The white paper is based on the principles laid out in my recent book, “Leading Science and Technology: India Next?”, that has many data-based insights into the research ecosystem.
Building the AI Ecosystem in India
Here are three things that India needs to do immediately if we indeed want to become a leader in Artificial Intelligence.
Critical Mass of AI Researchers
First, we need a critical mass of AI researchers. The number of productive AI researchers in India are countable on fingers. The number of Ph.D. students in the area is small. The AI research community will form the fountainhead for impact – they are the ones who will create and develop technology and train AI workforce and entrepreneurs. Without them, no other intervention will bear fruit. We should aspire to create a pool of 500 top AI researchers and 2000 Ph.D. students in India in the next 5 years. For this, we need to create AI faculty and Ph.D. tracks that attract the best from across the world. The government can enable this through a fellowship program for AI research faculty and Ph.D. students. The fellowship should promise great personal and professional returns, such as good salary and a great research environment. These fellows should be housed at top Indian universities through a continuous competitive process. (Details in 2.1-2.3 of AI white paper) Other than government, there is room for world-class private AI research institutes and universities in India. We need a vision of getting a critical mass of AI researchers under one roof in India in order to do disruptive research.
Unique AI Programs
Second, we need to create AI programs based on our strengths and priorities. Mimicking the west and doing derivate research will not help. We need to identify new ambitious questions and build AI-based systems as solutions. Consider, for example, a ‘swachchta’ robot that cleans roads as it autonomously moves among people. For solving such problems, we need to build differentiated data-sets relevant to our problems and interdisciplinary teams to work on them. This will not only help us lead but also spur collaboration- everyone in the world will like to work with us and our data-sets. India already has several advantages. We have no dearth of problems. In the last two decades, we have done immense digitization (even in areas where the West doesn’t have it- like Aadhaar). There is a huge amount of data with the government and companies. Additionally, we have the advantage of being able to collect data easily at low cost. Another interesting aspect is our huge space and military program which can provide several questions for AI and a platform for experimentation.
We need to form the connections and accumulate efforts. The industry, government bodies and academia need to vigorously collaborate. First, we need to create inter-university centers to facilitate industry interaction, and for creation/maintenance of data-sets and tools for industry usage. These centers act as a platform to get original questions and data from companies and also help in commercializing AI technology. Second, we need our funding agencies, defense and space agencies to launch various AI research programs. These programs should include areas of theoretical development, civic society programs, interdisciplinary programs, Indic projects, large public data-set creation, AI policy and projects to serve government priorities. A competitive process to involve researchers in these programs will drive excellence and an accumulation of effort of the AI community in a few well-defined areas. It is also critical that the industry is also allowed to participate in such programs.
Lastly, our companies need to proactively engage with university research programs and Ph.D.s if they want to do real innovation. For the last two decades, they have engaged with B.Tech. students through various hiring, internship and training programs. This is the time to do the same with Ph.D. students to create a demand, build capacity and develop high potential teams.
Training and Commercialisation
Lastly, we need the right policy and structures to get spill-overs to training and commercialization. We need data science and AI programs in various shapes and forms for different types of audience. Data science is emerging as a horizontal skill, very much like computer literacy. India should pioneer the curriculum development for data science in K12 schools (refer www.datasciencekids.org). Similarly, it needs to be introduced as a first-year course in all engineering disciplines. Finally, short-term courses, degree programs and bootcamps need to be available for people from industry and different domains of specialization. Our critical mass of researchers can offer such programs through their institutions. They should have freedom and incentive to do so.
Commercialization requires people, tools and connection. Training solves the ‘people’ problem. A critical mass of these trained data science professionals will take technology to the companies they work for. Some of them will become entrepreneurs leading to technology commercialization. Further, inter-disciplinary centers will serve as a pivotal agency for commercialization by maintaining and supporting tools built out of research. Private companies need to create open source tools and APIs using their core competency of writing good scalable software. Google, Facebook and Amazon have shown the way. The government can spur this by allowing companies to participate in government grant proposals and by providing incentives. University researchers and industry can come together as a team to build end-to-end solutions and to spur commercialization.
AI is a big opportunity for India. At the same time, it is a big risk. If we do not excel, our companies and economy cease to be competitive to global companies. We need to act now.