Perplexity CEO disagrees with Nandan Nilekani's stance on AI training
What's the story
Perplexity AI CEO Aravind Srinivas has publicly disagreed with Infosys co-founder Nandan Nilekani's opinion, that India shouldn't focus on building its own artificial intelligence (AI) models.
While Srinivas acknowledges Nilekani's immense contribution to the Indian tech sector through Infosys and UPI, he believes it is important for firms to train their own AI models while also building on top of existing ones.
AI strategy
Nilekani's advice to Indian AI start-ups
In October, Nilekani advised Indian AI start-ups against building large language models (LLMs) and instead focus on creating practical AI applications.
He suggested India should leave the development of LLMs to Silicon Valley giants who have billions at their disposal.
His strategy involves using these models for synthetic data creation, quick small language model development, and training them with relevant data.
AI implementation
Nilekani's focus on frugal data collection infrastructure
Nilekani stressed on the need for a strong infrastructure that can collect data to deploy and scale apps affordably.
"Let other people build LLMs, we will make sure it works for people," he said.
His approach indicates a shift in focus from building the model to applying and scaling AI technology in India.
Industry consensus
Other industry leaders share Nilekani's viewpoint
Nilekani's opinion on AI model development is echoed by other industry leaders in India.
TCS CEO K. Krithivasan echoed similar sentiments to The Hindu Businessline last week, questioning the incremental advantage of building new LLMs with so many already available.
He suggested Indian organizations could benefit more by integrating existing systems, just like Nilekani's advice for a practical approach toward AI technology implementation.