'We Need More AI Researchers'

'20 per cent of the global AI researchers are of Indian origin, and we should bring them back.'

Not funding but lack of skilled researchers in artificial intelligence (AI) is a bigger hindrance Indian AI startups are facing in developing homegrown solutions, said investors and founders during a session at the Global IndiaAI summit in New Delhi.

"India doesn't have a paucity of capital, and today what is happening is that capital is being invested in the best founders who are building the most interesting companies. So in AI, what we are seeing is the beginning of an interesting wave of AI application companies," said Rajan Anandan, managing director, Peak XV, a leading venture capital firm.

Anandan said the real issue in the sector was the lack of qualified AI researchers.

"We need more AI researchers. Where are they going to come from? ... 20 per cent of the global AI researchers are of Indian origin, and we should bring them back.

"Most of them are in the US, but we should bring them back the way China did over the past 15 years," he added.

Anandan said India had the largest number of STEM (science, technology, engineering and mathematics) graduates.

"We just need to make sure we invest in them over a period of time."

Anandan said his firm had Rs 16,000 crore (Rs 160 billion) of capital.

"There's no shortage of capital. We just want more people starting up in AI."

Fellow panellist Abhishek Upperwal, founder and chief executive officer, Soket Labs, talked of the need for good researchers, especially when it came to building foundational AI models -- an area where Indian startups have to compete with big global players.

"I'm happy that people are talking about the research aspect when it comes to building foundation models because that is ultimately very important.

"Unless we break that particular barrier, we will mostly be executing and building on top of whatever has been done," he said.

Taking the example of OpenAI, Upperwal said because of the lack of resources, Indian firms would ultimately end up building on top of Western models like OpenAI.

"If we have to be at the frontier, we have to invest in research and break that particular barrier," he added.

The panellists were unanimous that homegrown AI talent and data contextualisation, to be used to train language models, were the key requirements for Indian AI startups.

Feature Presentation: Ashish Narsale/Rediff.com