BharatGPT Group Reveals 'Hanooman': A Comprehensive Guide to the Indic AI Model

The BharatGPT group, spearheaded by IIT Bombay alongside seven other top-tier Indian engineering institutes, has made a groundbreaking announcement. They are set to launch their inaugural ChatGPT-like service next month. Supported by Reliance Industries Ltd and the Department of Science and Technology, the group has collaborated with Seetha Mahalaxmi Healthcare (SML) to develop the 'Hanooman' series of Indic language models. Here's everything you should know about this exciting development.

What Exactly is Hanooman?

Hanooman represents a collection of large language models (LLMs) proficient in responding across 11 Indian languages, including Hindi, Tamil, and Marathi, with plans for expansion to over 20 languages. As reported by Bloomberg, the BharatGPT group showcased various individuals interacting with the AI tool in different languages via video demonstration.

Designed to cater to diverse sectors such as Healthcare, Governance, Financial Services, and Education, Hanooman stands out as more than just a typical chatbot. It's a multimodal AI tool capable of generating text, speech, videos, and more in multiple Indian languages. Among the customized versions is VizzhyGPT, an AI model tailored for Healthcare, leveraging extensive medical data.

These AI models boast a parameter range spanning from 1.5 billion to an impressive 40 billion. Vishnu Vardhan, Founder of SML, highlighted the challenges associated with the quality of datasets in Indian languages during the Hanooman launch event. He emphasized the prevalence of synthetic datasets derived from translations, potentially leading to inaccuracies or distortions, as reported by the ANI news agency.

Are There Other Indian Language Models?

In addition to BharatGPT, numerous startups like Sarvam and Krutrim, backed by notable VC investors such as Lightspeed Venture Partners and billionaire Vinod Khosla’s fund, are also developing customized AI models tailored for India, according to Bloomberg.

Understanding Large Language Models (LLMs)

Large language models employ deep learning techniques to process extensive text data. They analyze vast amounts of text, discerning structure and meaning while learning from it. LLMs are trained to recognize meanings and relationships between words, with their proficiency increasing as they are fed more training data.

Training data typically consists of large datasets like Wikipedia, OpenWebText, and the Common Crawl Corpus, which provide ample text data for models to comprehend and generate natural language.