BharatGen geared to put a desi spin on ChatGPT, IIT-Bombay plays role

A team of experts – from IIT-Bombay, IIT-Madras, IIT-Mandi, IIT-Hyderabad, IIT-Kanpur, IIIT-Hyderabad and IIM-Indore – have completed the basic backend work

Mumbai: After observing the global buzz around ChatGPT, since it was released to public in November, 2022, IIT-Bombay professor Ganesh Ramakrishnan, who bears a strong interest in machine learning and AI algorithms, considered developing an indigenous solution that could use generative AI reflecting India’s cultural and linguistic diversity. Now, after spending a year on the project, aided by a consortium of experts from leading institutions, his efforts bore fruit as the Central government officially launched the initiative to develop BharatGen, an Indian version of ChatGPT, on September 30.

A team of experts – from IIT-Bombay, IIT-Madras, IIT-Mandi, IIT-Hyderabad, IIT-Kanpur, IIIT-Hyderabad and IIM-Indore – have completed the basic backend work for the initiative

The team of experts – from IIT-Madras, IIT-Mandi, IIT-Hyderabad, IIT-Kanpur, IIIT-Hyderabad and IIM-Indore – have completed the basic backend work for the initiative, including developing algorithms and providing technical support essential to its success. The team is now looking to collaborate with AI researchers, engineers, behavioural economists and design experts to create a multidisciplinary approach to solve India’s challenges through AI.

As they got to work on the project, they loosely titled it BharatGPT, which now has an official name – BharatGen.

So, what is the idea behind BharatGen?

“It is more than just foundational AI, it’s an inclusive and enabling ecosystem. From education to health, agriculture to industry, it opens new avenues for innovation. The goal is ambitious but clear: build AI for all, AI for every Indian,” said Ramakrishnan, the 46-year-old professor in the computer science and engineering department of IIT-B.

The project, however, was not without challenges. Ramakrishnan’s primary concern was how could generative AI models be designed to work for India, with its vast array of languages, scripts and cultural contexts. The consortium aimed to create India-centric, multimodal, and multilingual AI models, with focus on building technology that speaks to the heart of India’s diversity.

“The project celebrates India’s rich diversity by being mindful of its unique needs across various sectors,” said Ramakrishnan.

To that end, he said, each institution is looking into specific tasks to develop models in language, speech and vision that aim to enhance citizen engagement across the nation – from education and healthcare to agriculture and industry.

One of the biggest challenges BharatGen faces however is India’s linguistic diversity, as there are over 1,600 languages spoken across the country. “The diversity of languages and scripts in the Indian knowledge system is a key challenge. We are working closely with organisations and government bodies to ensure that BharatGen can accommodate these diverse linguistic needs,” said Ramakrishnan.

He added, by focusing on multilingual and multimodal AI models, BharatGen will address a critical gap in global AI models, which often overlooks Indian languages. “By training on multilingual datasets, BharatGen will ensure even underrepresented Indian languages are included in the AI ecosystem,” Ramakrishnan explained. “This approach not only reflects India’s cultural diversity but also promotes inclusivity by making AI accessible to people across the country, regardless of the language they speak.”