BENGALURU: AI search engine Perplexity AI cofounder and CEO Aravind Srinivas disagrees with Infosys cofounder and chairman Nandan Nilekani’s stance on bypassing model training skills and instead concentrating on developing applications using pre-existing models. “Nandan Nilekani is awesome, and he’s done far more for India than any of us can imagine through Infosys, UPI,” Srinivas acknowledged in a post on X. “But he’s wrong in pushing Indians to ignore model training skills and just focus on building on top of existing models. It’s essential to do both.” Perplexity AI closed a $500 million round of funding that tripled the company’s valuation to $9 billion, according to a Bloomberg report last month.
In another post, Srinivas pledged his personal support for initiatives aimed at building India’s foundational AI capabilities. “I am ready to invest $1 million personally and 5 hours/week of my time into the most qualified group of people that can do this right now for making India great again in the context of AI,” he wrote. “Consider this as a commitment that cannot be backtracked. The team has to be cracked and obsessed like the DeepSeek team and has to open-source the models with an MIT license.”
Srinivas’s vision emphasises the need for a dedicated and highly skilled team to take on the challenge of building globally competitive AI models, underpinned by a commitment to open-source principles. Drawing from his own experiences as a tech entrepreneur, Srinivas expanded on the argument in a follow-up post, where he compared India’s potential in AI to its achievements in space exploration. He noted that India historically excelled in delivering complex projects at a fraction of the cost of its global counterparts, citing Isro’s success as an example. “I feel like India fell into the same trap I did while running Perplexity, thinking models are going to cost a lot of money to train,” Srinivas wrote. “India must show the world that it’s capable of Isro-like feats for AI.”
Srinivas pointed to recent advancements by DeepSeek as evidence that significant progress in AI can be achieved without excessive spending. DeepSeek is a Chinese AI startup which unveiled DeepSeek V3, a large language model with 671 billion parameters. The model outperformed Meta’s Llama 3.1 and OpenAI’s GPT-4o in benchmarks for text understanding, coding, and problem-solving, marking a significant milestone for China’s AI sector. He encouraged India to focus on building foundational AI models that are not only effective for Indic languages but also competitive on global benchmarks. “I’m not in a position to run a DeepSeek-like company for India, but I’m happy to help anyone obsessed enough to do it and open-source the models,” he added.
Srinivas’s comments highlight a critical juncture for India’s AI ecosystem, as the country grapples with questions of resource allocation and strategic priorities. His argument underscores the importance of developing both foundational model training skills and application-based innovations to position India as a global leader in AI.
In a landmark announcement on Wednesday, OpenAI, Oracle, and SoftBank unveiled a joint venture named “Stargate,” committing up to $500 billion over the next four years to develop AI infrastructure in the United States. It aims to support the next generation of AI advancements and plans to create about 100,000 jobs.
Nilekani has time and again said that India needs to focus on building small language models instead of large language models (LLMs) in the space of generative AI, as the latter has gotten increasingly commoditised. “Foundational models are not the best use of your money. If India has $50 billion to spend, it should certainly build on compute, infrastructure, and AI cloud. These are the raw materials and engines of this game,” he told TOI in an interview last month.
In November, Manish Gupta, director of Google Research in India, said that he respectfully disagreed with Nilekani’s advice to India on prioritising use case building over and above building foundational models in artificial intelligence. “He is not preaching what he practised. He revolutionised India’s technology landscape by starting with the basics. With Aadhaar, he did not start with use cases, he started with building foundations. We too must, using our constraints as ingredients for innovation,” Gupta said at the Bengaluru Tech Summit last year.
