Global AI Conclave: AI for Bharat will need philanthropic investment in the initial stages, say Accel and Karya

The AI for Bharat models’ economics is not viable to be scaled in the initial stages and will require philanthropic investments, said panellists Manu Chopra of Karya.ai and venture capital firm Accel’s founding partner Prashanth Prakash

The AI for Bharat models’ economics is not viable to be scaled in the initial stages and will require philanthropic investments, said panellists Manu Chopra of Karya.ai and venture capital firm Accel’s founding partner Prashanth Prakash at the CNBC18 and Moneycontrol’s Global AI Conclave.

“What I have confidence is that frugal startups are emerging and they are mindful of the AI models they need to build at a cost. It is clear that long-term viability is important and we cannot do an AI operation at Rs 15 as of now,” Prakash said.

According to Prakash, Google has been intimidating for the poor and people with low literacy levels, and the AI models can distil and synthesise information at a simpler level by building new interfaces like chatbots that can communicate through voice in Indian languages.

The uptake in healthcare is expected to be highest as the perceived value is also higher. For instance, the AI chatbot developed by Karya for Tuberculosis has helped more than a million patients in Karnataka by delivering them information in Kannada.

Karya, which started at Microsoft Research a few years back before becoming a startup in 2021 has been working with disadvantaged and poor sections of society to build AI models for Bharat by paying them salaries that are 20 times that of the minimum wage of the country.

“The idea is to be transformative by reaching the communities that even the internet has not been able to serve them. But they should get to a point where they don’t need us,” said Chopra.

According to him, the internet has not been able to solve for Bharat and the information asymmetry continued.

“When I think about AI and uplifting Bharat, we cannot make the same mistakes we did with the internet with AI. So from Day 1, we have to think about AI solutions that we build are for everyone and not just a few in India,” Chopra said.

He said that while talent is equally distributed, opportunity is not. “Low literacy does not mean they don't have the skill to build AI models,” Chopra said adding that the company has been building livelihood opportunities for poor people in rural India by helping them to speak to AI in Indian languages to create LLM models.

According to Prakash, Indian AI companies should build for global from the country first and then adapt for India. “What is built for global, we need to adapt it for Bharat and build over a period for the Bharat use case. We cannot yet build AI business for Bharat with the existing VC model,” he added.