BharatGPT designed to work in 11 Indian languages

BharatGPT is a large language model (LLM) developed by a consortium led by the Indian Institute of Technology-Bombay (IIT-B) and involving academics and experts from multiple IITs. “The consortium is creating Large Language Foundational Models with a focus on India,” Vishnu Vardhan, CEO of Seetha Mahalaxmi Healthcare (SML), the business that has so far funded the consortium.

It's designed to work in 11 Indian languages, with plans to expand to 22 eventually. This is a significant milestone for India's artificial intelligence journey, as it marks the development of a homegrown LLM that caters to the specific needs and nuances of the Indian market.

BharatGPT can understand and generate text in 11 Indian languages, including Hindi,Bengali, Tamil, Telugu, Marathi, Kannada, Malayalam, Gujarati, Punjabi, Odia, and Urdu. This makes it accessible to a large portion of the Indian population who may not be comfortable with English.

BharatGPT is not just limited to text. It can also process and generate images, documents, and audio files. This makes it a versatile tool that can be used for a variety of tasks, such as machine translation, chatbots, and content creation.

BharatGPT is trained on a massive dataset of Indian text and code, which allows it to better understand the cultural context and nuances of the Indian language. This makes it more accurate and relevant for Indian users.

It has designed from the ground up to be multi-modal, which means that input does not have to be just text. It may be a photograph, a paper, or an audio file.. Sources said, while this is a government-backed effort, monies from the government are not expected for several months.

It is funded by Seetha Mahalaxmi Healthcare (SML), a Bengaluru-based firm.The initial phase of development involved importing 256 GPUs at a cost of 16 million dollars. Going forward, efforts are on to reach an arrangement with a big chipmaker for another 1,000 GPUs. This would cost an additional US$120 million.

Prof Ganesh leads the BharatGPT consortium, which comprises prominent academic institutions such as IITs, IIIT-Hyderabad, IIM Indore, and Bhashini. Reliance Jio said on Wednesday that it is also cooperating with IIT-B on BharatGPT, and Nasscom is also collaborating with the consortium.

“BharatGPT seeks to leverage AI to empower India’s diverse communities, bridging the digital divide. This has transformative potential to reshape India’s digital landscape and unlock opportunities for the entire population,” Ramakrishnan said.