Zoho founder and Chief Scientist Sridhar Vembu is advising India to strategically avoid direct competition in the Large Language Model (LLM) race, which is currently dominated by Big Tech companies. Instead, Vembu suggests focusing on smaller, more efficient AI models that require less energy and capital.

Vembu's recommendation comes at a crucial time, with India's interest in AI rapidly increasing and the country preparing to host the India AI Impact Summit. His perspective aligns with the recent Economic Survey, which highlighted the challenges India faces due to limited access to advanced computing infrastructure and capital, making the pursuit of foundational models a potentially difficult AI strategy. The Economic Survey suggests a bottom-up approach to AI development, which aligns more effectively with India's realities.

Vembu points out the substantial costs, ranging from $50 to $100 billion, and the significant energy consumption associated with the large models developed by major technology companies. He suggests that "Sometimes, staying a little bit behind is a good idea" in this context. Instead of trying to replicate these energy-intensive models, Vembu advocates leveraging India's "brain power" to invest in research and development of smaller, more resource-efficient AI solutions.

He believes India should avoid entering the capital-intensive LLM race and instead pivot to smaller, distinct approaches. He stated that the supply chain issues of chips will be resolved and AMD is also coming with AI chips.

Zoho is already developing smaller AI models, based on 7 billion to 20 billion parameters, to address specific domain issues for their customers. Vembu stated at the CNBC-TV18 and Moneycontrol Global AI Conclave that they have discovered smaller models to be more effective for domain-specific issues, which is why they are currently not pursuing the 500 billion parameter models. Additionally, they are interested in acquiring their own GPU infrastructure, as it proves to be more cost-effective in the long run.

Vembu's emphasis on smaller AI models also reflects a broader discussion about the evolving landscape of software development and the potential impact of AI on coding jobs. He has advised programmers to consider alternative livelihoods as AI becomes increasingly capable of generating applications and websites, even without technical expertise from the user. He observed the rapid advancements in AI models, such as Anthropic's Claude, which recently created a C compiler.

While acknowledging the potential displacement of some coding roles, Vembu also envisions a future where AI could make technology more accessible and allow individuals to focus on other aspects of life, such as family, nature, and culture. He suggests that such a life is best lived in small, close-knit rural communities, noting that he already lives this way and sees it as deeply fulfilling if rural poverty can be addressed.

Zoho uses its own LLMs for internal product integration, which saves the company and its customers money. These are custom-trained models built from the ground up and are already in production. The company is also working on larger models, such as 100 billion parameter models, for internal use.

Vembu's advocacy for smaller AI models and a strategic approach to AI development in India highlights the importance of focusing on resource efficiency and leveraging the country's unique strengths. His vision emphasizes innovation and a balanced approach to AI adoption, considering both its potential benefits and potential disruptions.