AI model backed by Asia’s richest person to launch in March

A CONSORTIUM backed by Mukesh Ambani’s Reliance Industries and India’s top engineering schools aim to launch its first ChatGPT-style service next month, a big step in the country’s ambitions to become a player in the field of artificial intelligence.

The BharatGPT group, encompassing an arm of India’s most valuable company and eight affiliated universities, offered a sneak peek of the large language model on Tuesday during a technology conference in Mumbai.

In a video played before delegates, a motorcycle mechanic in southern India queried an AI bot in his native Tamil, a banker conversed with the tool in Hindi, and a developer in Hyderabad used it to write computer code.

If successful, the model — dubbed Hanooman after the half-monkey Hindu deity — will represent an advance for India in the accelerating race to develop potentially transformative AI technology.

BharatGPT envisions the model working via 11 local languages in four main fields: health care, governance, financial services and education. It developed the model in collaboration with Indian Institute of Technology universities including in Bombay, backed by wireless carrier Reliance Jio Infocomm Ltd. and India’s government. 

A swath of startups such as Sarvam and Krutrim, backed by prominent VC investors such as Lightspeed Venture Partners and billionaire Vinod Khosla’s fund, are also building open-sourced AI models customised for India.

GET BT IN YOUR INBOX DAILY

Start and end each day with the latest news stories and analyses delivered straight to your inbox.

Sign UpVIEW ALL

While Silicon Valley companies like OpenAI are building ever-larger LLMs, those efforts involve workarounds because of computational constraints and simpler models affordable to smaller businesses and government departments.

“It’s a different genre of LLMs,” said Ganesh Ramakrishnan, chair of IIT Bombay’s department of computer science and engineering. 

Hanooman will also offer speech-to-text capabilities, making it vastly more user-friendly, he said in an interview on the sidelines of the annual Nasscom IT industry conference. In a country of 1.4 billion, millions cannot read or write.

Reliance Jio will build customised models for specific uses, he said. The telecom-to-retail conglomerate is already working on Jio Brain, a platform to use AI across a network of about 450 million subscribers. 

LLMs are systems that learn from vast quantities of data and generate natural-sounding responses. Such models harness generative AI, a newer type of artificial intelligence popularised by the success of OpenAI’s ChatGPT.

BharatGPT’s effort in this space is somewhat unique — it’s the first private-public partnership of its kind in the country, and involves major players in disparate fields.

“It’s like the Indian joint family,” Ramakrishnan said, referring to inter-generational family structures still common in the country. “We are interdependent, and we do better together.” BLOOMBERG