Infosys co-founder Nandan Nilekani praises China's approach to AI, claims he is impressed with DeepSeek: 'They have done...'

Infosys co-founder Nandan Nilekani has raised concerns about the increasing complexity of artificial intelligence (AI) models, arguing that bigger models do not always mean better results. Speaking at the People+AI Mela in Bengaluru, he highlighted the lack of transparency in many AI systems, which are often controlled by private companies and hidden from public scrutiny.

“They keep making these models more complex, but that does not mean they are improving,” Nilekani said, as reported by The Economic Times. He warned that this secrecy limits innovation and prevents wider contributions to AI development.

Praising China’s approach to AI, Nilekani pointed to the DeepSeek project as an example of open AI development. He noted that China has built effective AI models with fewer resources while openly sharing their methods. “They have done the world a service,” he said, referring to China’s decision to publicly share AI-building practices.

Nilekani believes AI models are becoming commodities, with hundreds being developed worldwide. He emphasised that the real challenge is not just building models but using them effectively to create real-world value.

He also stressed the importance of high-quality datasets, particularly in Indian languages, for AI growth in India. He cited AI4Bharat, an IIT Madras initiative focused on open-source AI for Indian languages, as a crucial step in this direction. “The hard part is not building models, but using them to help people,” he said.

Nilekani also highlighted India's booming startup ecosystem, which has grown from 1,000 startups in 2016 to over 150,000 today. He believes AI should be used to solve practical problems rather than just focusing on building complex models.