Nandan Nilekani Advocates for Practical AI Applications in India, Discourages Pursuit of Large Language Models

Why Infosys co-founder Nandan Nilekani feels India can't afford to make it's own ChatGPTs, DeepSeeks

Nandan Nilekani emphasizes the high costs of building core AI models and the need for practical AI applications in India. He highlights the importance of making AI accessible to improve productivity for users like farmers, and urges Indian AI firms to focus on real-world challenges, rather than competing with major global companies.Infosys co-founder and chairman, Nandan Nilekani, recently discussed India's AI aspirations, emphasising that while core AI models are crucial, their high computational costs pose a significant challenge. In an interview with MoneyControl, he pointed out that building these models can be prohibitively expensive, often reaching around $50 million, a cost that many companies can afford but still requires strategic investment. Nilekani stressed that the real challenge lies not just in developing AI models, but in making them scalable and affordable for widespread use. "You can have all the AIs in the world, but if you can't get five people to work together and collaborate, then you cannot go anywhere," he explained. The focus should be on building systems that are inexpensive to operate, ensuring that AI becomes accessible to all, particularly in rural India. He further said, "...And I think now the building of core models is now at a quality and cost point, which anybody can do it. $50 million, so many companies can spend that. So I think now is the time to do that. But again, the challenge with AI is, how do you make it work? How do you make it work at population scale? How do you make sure that it's very cheap to operate? Because they say inference and reasoning and all kinds of things, but they're very expensive from a compute perspective. And we can't afford those costs." He elaborated on this vision, suggesting that for AI to truly serve India's needs, it must be affordable at a granular level. "If we want a farmer in a village to speak to his phone in Hindi or Bhojpuri and receive practical, actionable advice on improving his productivity, all for just a few rupees per transaction, that's a whole different ballgame," Nilekani said. On the impact of AI on employment, Nilekani acknowledged that some jobs would be affected, but expressed an optimistic outlook. "Some jobs will be affected -- certain tasks will be automated -- but very few jobs will be entirely eliminated," he stated. He also believes AI will increase human productivity and create entirely new roles that are yet to be imagined. Earlier, Nilekani had also urged Indian AI companies to focus on practical applications rather than competing in the crowded market of large language models (LLMs). At Meta's Build with AI summit in Bengaluru, he pointed out that India's priority should be to solve real-world problems, not enter the arms race of building the next big LLM. "Let the big boys in Silicon Valley do it, spending billions of dollars," Nilekani remarked. He also praised Meta's decision to make its foundational LLMs open source, calling it a "game changer" for India, as it enables developers to use Meta's Llama models to create and train new AI solutions. Nilekani's vision for India's AI future is focused on innovation, collaboration, and accessibility, emphasising that India's role should be to leverage AI to address the country's unique challenges.

Why Infosys co-founder Nandan Nilekani feels India doesn't need a China-type DeepSeek AI

Nandan Nilekani believes AI will enhance rather than replace core human abilities like empathy, leadership, and creativity. He emphasizes the need for adaptive skills and first-principles thinking. While some jobs might be affected, AI will create new opportunities and increase productivity.Infosys co-founder and chairman Nandan Nilekani while talking about what kind of AI model the country requires pointed out that but building core AI models are very expensive from a compute perspective and the country can't afford those costs. "You can have all the AIs in the world, but if you can't get five people to work together and collaborate, then you cannot go anywhere," Nilekani said in conversation with MoneyControl. "And I think now the building of core models is now at a quality and cost point, which anybody can do it. $50 million, so many companies can spend that. So I think now is the time to do that. But again, the challenge with AI is, how do you make it work? How do you make it work at population scale? How do you make sure that it's very cheap to operate? Because they say inference and reasoning and all kinds of things, but they're very expensive from a compute perspective. And we can't afford those costs," said the Infosys chairman during the interview. Commenting on what is needed in the country, the Infosys co-founder said "So if we want a farmer in a village to speak to his phone, in Hindi or Bhojpuri or Maithili or whatever, and get a very reasoned answer on what he needs to do so that he can improve his productivity, and I want to do that at a few rupee per transaction, that's a whole different ballgame. And that's where we have to go." While acknowledging that AI will impact employment, Nilekani maintained an optimistic outlook. "Some jobs will be affected -- certain tasks will be automated -- but very few jobs will be entirely eliminated," he told MoneyControl. "AI will make humans more productive and create new jobs we haven't even thought of yet." Previously, Nandan Nilekani, chairman and co-founder of Infosys, urged Indian artificial intelligence (AI) companies to focus on creating practical applications rather than competing in the saturated market of large language models (LLMs). Speaking at Meta's Build with AI summit in Bengaluru, he stressed on the need for India to lead in AI use-cases that address real-world challenges and foster innovation across various sectors. "Our goal should not be to build one more LLM. Let the big boys in the (Silicon) Valley do it, spending billions of dollars," Nilekani said. "We are not in the arms race to build the next LLM... We are here to make a difference," he remarked. Nilekani also praised Meta for its decision to make its foundational LLMs open source, calling it a "game changer for us in India." This move allows developers to utilize synthetic data generated by Meta's Llama models to create and train other models, thereby enhancing innovation in the AI space.