Nandan Nilekani Advocates AI Infrastructure Over Large Language Models
BENGALURU: Prominent technocrat and Infosys co-founder Nandan Nilekani has reiterated his stance against India investing in creating its own large language models (LLM).
Responding to recent comments by Google Research India Director Manish Gupta, Nilekani maintained that India’s resources would be better spent on developing AI infrastructure and cloud computing capabilities rather than building costly foundation models.
Foundation models are not the best use of your money,” Nilekani stated during an interaction. If India has $50 billion to allocate, it should invest in building computing infrastructure and AI cloud services. These components are the resources and driving forces of this.
Nilekani’s remarks come amid ongoing debates in the Indian tech ecosystem about the nation’s priorities in artificial intelligence (AI) development. The discussion highlights contrasting visions for India’s future in AI—one emphasizing the creation of foundational technologies and another advocating for leveraging existing global models to build tailored applications.
Foundation models like those developed by OpenAI and Meta require significant investments, often amounting to billions of dollars. These models are trained on extensive datasets using advanced and costly computational infrastructure.
While they form the basis for innovations in natural language processing and other AI applications, they also demand continuous maintenance and updates, adding to their cost.
Last month, at the Bengaluru Tech Summit, Manish Gupta expressed his “respectful disagreement” with Nilekani’s position. Gupta argued that India would benefit from building its own foundation models to address unique challenges and innovate within the constraints of its ecosystem.
“Nilekani revolutionized India’s technology landscape by starting with the basics,” Gupta said, referencing Nilekani’s pivotal role in building Aadhaar. With Aadhaar, he began not with use cases but by foundational. We too must, using our constraints as ingredients for innovation.”
Nilekani, however, believes that India’s limited resources necessitate a more pragmatic approach. He has consistently advocated for focusing on building use cases for AI by leveraging existing LLMs and foundational technologies developed by global players.
“The value is in how you apply AI,” Nilekani emphasized. “India can create transformative solutions by building on top of global LLMs. Developing our own foundation models would consume immense resources without guaranteeing a competitive edge.”
According to Nilekani, prioritizing infrastructure—such as compute capabilities, AI clouds, and data pipelines—will position India as a leader in applying AI across industries like agriculture, healthcare, and education. This approach aligns with India’s long-standing focus on frugality and scalability, enabling solutions that address local challenges at scale.
The Nilekani-Gupta debate reflects a broader dichotomy in AI development strategies worldwide. Nations like the United States and China have heavily invested in building their own LLMs, enabling them to shape global AI standards and technologies. However, the cost-intensive nature of these endeavors makes them inaccessible for many developing countries.
Gupta’s argument is rooted in the belief that India should aim for technological sovereignty by creating its own foundational models. He suggests that such an approach would inspire innovation despite resource constraints, much like Aadhaar did for digital identity.
In contrast, Nilekani’s perspective focuses on pragmatism and leveraging India’s existing strengths. By adopting global LLMs and adapting them for local needs, India can bypass the heavy costs of model development while still reaping AI’s benefits.
Gupta’s comparison of foundation models to Aadhaar has sparked considerable discussion. Aadhaar, a digital identity system envisioned by Nilekani, laid the groundwork for India’s digital infrastructure, enabling a host of innovations in fintech, governance, and social welfare.
However, Nilekani argues that the analogy is not entirely applicable. “Aadhaar was about creating foundational digital infrastructure that was missing in India. In AI, foundational models already exist and are accessible. Our focus should now be on how we use them,” he explained.
Where Should India Invest?
Both Nilekani and Gupta agree that AI holds transformative potential for India. The crux of their disagreement lies in how to prioritize investments:
- Infrastructure: Nilekani’s proposal centers on strengthening India’s computational and data infrastructure. This includes expanding AI cloud capabilities, improving data collection systems, and enabling widespread access to AI tools.
- Foundation Models: Gupta’s vision involves building indigenous models tailored to India’s linguistic and cultural diversity, potentially unlocking innovations not feasible with globally developed LLMs.
The Indian tech community is divided on the issue. Proponents of Nilekani’s approach emphasize the need to avoid duplicating global efforts and instead focus on practical applications. “India’s strength lies in its ability to scale affordable solutions. Investing in use cases will drive immediate impact,” said a Bengaluru-based AI entrepreneur.
On the other hand, advocates of Gupta’s viewpoint argue that developing indigenous models could catalyze breakthroughs in AI research. “Building foundational models may be expensive, but it’s an investment in intellectual property and long-term innovation,” noted an academic from a leading Indian university.
Countries like the U.S., China, and the EU have dedicated significant resources to building foundational AI models. OpenAI’s GPT series, Google’s Bard, and Meta’s LLaMA are examples of such initiatives. These models have set benchmarks in natural language understanding and have sparked competition in the global AI race.
India’s AI journey, however, is marked by resource constraints and a focus on scalability. Nilekani’s vision aligns with these realities, emphasizing practical outcomes over prestige. Gupta’s proposal, while ambitious, suggests that India should aim to join the ranks of global AI leaders.
The debate underscores the need for a balanced strategy that leverages India’s strengths while addressing its challenges. Some experts suggest a hybrid approach—investing in foundational AI research while simultaneously building use cases on top of existing models.
“India doesn’t need to choose between infrastructure and models. It’s about finding synergies that deliver the greatest value,” said a policy analyst.
The discussion between Nandan Nilekani and Manish Gupta highlights the complexities of India’s AI aspirations. Whether India focuses on building foundation models or leveraging existing ones, the outcome will significantly impact its position in the global AI ecosystem.
For now, Nilekani’s pragmatic approach to strengthening infrastructure seems aligned with India’s immediate priorities, while Gupta’s vision offers a long-term pathway to technological independence.
As India navigates this critical juncture, collaboration among policymakers, industry leaders, and technologists will be key to ensuring that its AI investments deliver maximum societal benefit.
China Unveils $80 Billion Gold Deposit, Largest Ever Found
China Explore World Largest $80 Billion Gold Deposit in Hunan Province In a groundbreaking development, China has announced the discovery of what is believed to be the largest known gold deposit in the world. This historic find, located in Pingjiang County, Hunan Province, has been valued at an estimated $80 billion (approximately £63 billion). The …
Stay connected with NH Punjab on social media. Follow us on Facebook, Twitter/X, and Instagram for the latest news updates, behind-the-scenes content, and more. Engage with us online and be a part of our growing community.
