The global AI race is pushing Western nations to evolve at a rapid pace, and India, as a result, has to up its game too. However, unlike American firms that are focusing highly on Large Language Models (LLMs), India could gain highly by working on more accessible Small Language Models (SLMs), says Infosys Chairman Nandan Nilekani.
Nilekani, in a recent interaction, said that India is exceptionally well-positioned to become a global leader in the application of artificial intelligence (AI) and frugal engineering. He outlined three key areas where this combination of AI and frugal engineering will be most impactful: the global capability centers of multinational firms, the work of IT service providers in assisting clients with their AI transformation, and India’s own digital public infrastructure.
India has a better shot at AI
To give an example of the potential for innovation at a population scale, Nilekani pointed to two of India’s most successful digital initiatives – Aadhaar and UPI.
“The idea of Aadhaar was that everybody should have a digital ID, which shall be the entry point of everything,” said Nilekani. For the Unified Payments Interface (UPI), he highlighted its design philosophy, saying, “it was to create a population-scale transaction with frugal engineering, which was inclusive and provided an equal opportunity for all to participate. It is interoperable, scalable, and privacy-protected.”
Nilekani said that the simplicity of the one-page API design for both Aadhaar and UPI was a key factor in their success.
SLM to become India’s strength
Commenting on the future of AI, Nilekani offered a positive view. He suggested that while large language models (LLMs) may become a commodity, India’s strength lies in developing “small language models (SLM) which are cheaper” and can be customised to provide a localised AI experience.
“If a farmer in Bihar is able to speak into his phone in his version of Hindi and gets real time information from his agent on best farming practices and market access, then we have done it,” Nilekani gave an example.
