The recent visits of Nvidia’s Jensen Huang and Meta’s Yann LeCun didn’t just highlight the importance of the Indian market to artificial intelligence (AI) majors. It made clear that India needs top-tier AI research talent — and not just AI infrastructure — if it wants to become an AI power and make its National AI Mission a success.
Yann LeCun, Chief AI Scientist at Meta, spoke at various educational institutions including IIT Delhi and IIT Chennai, among others. LeCun, winner of the Turing Prize in 2018, urged India to enhance its participation in the global AI research community, and not focus only on AI product development. A dearth of cutting-edge research opportunities in AI, and brain drain (of some of India’s best AI research talent), he pointed out, are among the primary challenges for India to build its own AI expertise.
In contrast, when they shared the stage at the Nvidia AI Summit last month, Nvidia’s Jensen Huang and Reliance’s Mukesh Ambani emphasised building affordable AI infrastructure for India. But surprisingly, Jensen made almost no mention of the criticality of top-tier research talent for India. Their emphasis on AI infrastructure is in line with the weight India has placed on compute infrastructure in its National AI Mission (NAIM), with half of the Mission’s funds allocated towards it.
Admittedly, the computer is a prerequisite for any meaningful AI research. India also recently announced the setting up of three AI Centres of Excellence (CoEs) focused on health, education and agriculture as part of NAIM and the Indian focus on #AIforAll. However, the 10,000 GPU compute infrastructure and the three sectoral COEs will not, on their own, kickstart cutting-edge AI research in India. Even as India focuses on acquiring GPUs over the coming months, a key element of building a competitive edge in AI is being neglected — top-tier research talent in AI.
To be sure, NAIM has talent and skills as one of its pillars. However, the Mission’s diagnosis of the constraints or gaps India faces in the talent pillar is misplaced. It doesn’t highlight the need for India to attract, retain and train top-tier research talent. Instead, it envisions an IndiaAI FutureSkills programme that would focus on increasing the number and accessibility of AI curricula at the graduate and post-graduate levels.
On its own, the FutureSkills programme will help increase AI awareness and education, but will not help build a pool of cutting-edge talent in India in the next two to three years. And without cutting-edge talent in AI right now, India will lose the AI game, as LeCun also pointed out.
Look at France, LeCun’s home country, for example. For some time, French leaders sensed that France was losing the competitive edge to the United States (US) and China yet again, this time in the latest AI tech wave. Overseen by President Emmanuel Macron personally, the French AI strategy revolves around attracting French talent working elsewhere back to France. It recently organised a gathering of the top French AI talent at the Elysee Palace. A French startup, Mistral, started just over a year ago by French founders who formerly worked with Google DeepMind and the Fundamental AI Research (FAIR) team at Meta, has emerged as one of the top competitors to OpenAI’s ChatGPT platform. This has given France’s position in the AI world a major fillip.
LeCun pointed out another reason behind France’s emergence on the world stage — the setting up of Meta’s FAIR team in France over 10 years ago. It trained and inspired many French researchers to seek careers in AI research, LeCun argued, which, in turn, has contributed to the success of French AI startups such as Mistral.
India needs to do the same. It is widely acknowledged that much of the top AI research talent in Silicon Valley is of Indian origin. Let’s consider just one or two examples. Transformers, at the core of ChatGPT, were originally part of a seminal Google research paper, “Attention is all you need”. Two of the co-authors of that paper, Ashish Vaswani and Niki Parmar, are of Indian origin. Vaswani did his BTech at the Birla Institute of Technology, while Parmar did hers at Pune Institute of Computer Technology. Similarly, Aravind Srinivas, also an IIT-Madras alum and formerly a researcher with OpenAI, has gone on to start Perplexity.ai, considered one of the hottest AI startups in the Valley currently.
India needs to bring — or retain — some of this talent back to India and give them the necessary research ecosystem to flourish here in Bengaluru, Gurugram, Hyderabad, Chennai, Mumbai or anywhere else in India. To do that, India requires a vibrant ecosystem with a potent combination of capital, talent, research and infrastructure. Much like the IIT Madras Research Park has shown, a vibrant ecosystem allowed to flourish even in small pockets can lead to amazing breakthroughs and many success stories. The same strategy applied to AI could thrust India into the major league of AI.
The current thrust of the Anusandhan National Research Foundation should have a major AI pillar. Along with major Indian corporates, it should fund at least three or four AI labs that could host some of this talent to do fundamental research. These labs could also be provided with critical AI infrastructure, including the compute infrastructure and advanced AI chips being proposed to be acquired under NAIM and that will be built by the likes of Nvidia, Reliance, Yotta, and others.
None of this is to say that research talent in AI doesn’t exist in India already. It does. Our universities continue to produce great researchers in AI and related fields. Many of the global AI majors, including Nvidia, have thousands of Indians in their AI research labs situated in India. This base, combined with attracting and retaining more of the best top-tier AI talent, could help unleash India’s animal spirits in AI and drive the success of NAIM. And the fruits of success in AI, by the estimates of many, would be very sweet indeed.
Anirudh Suri is the author of The Great Tech Game and a nonresident scholar with Carnegie India. The views expressed are personal
