Introduction

The BharatGPT group, comprising IIT Bombay and the Department of Science and Technology, is set to launch its first ChatGPT-like service named Hanooman next month.

Large Language Models (LLMs)

LLMs utilize deep learning methodologies to process extensive text data, enabling them to grasp linguistic nuances and semantic relationships.
These models are trained on vast datasets like Wikipedia and OpenWebText, allowing them to comprehend and generate natural language by discerning patterns and meanings from the provided text.

Multilingual Capability: Hanooman is a series of large language models (LLMs) proficient in 11 Indian languages initially, with plans to expand to over 20 languages, including Hindi, Tamil, and Marathi.
Functionality: Beyond a mere chatbot, Hanooman serves as a multimodal AI tool, capable of generating text, speech, videos, and more across various domains such as healthcare, governance, financial services, and education.
Customized Versions: One notable variant, VizzhyGPT, tailored for healthcare applications, showcases Hanooman’s versatility in fine-tuning AI models to specific sectors.
Scale: The size of these AI models ranges from 1.5 billion to an impressive 40 billion parameters, reflecting their robustness and complexity.

Quality of Datasets: Concerns regarding the quality of datasets in Indian languages, emphasizing the prevalence of synthetic datasets derived from translations, may lead to inaccuracies or distortions.
Competition: Alongside BharatGPT, several startups like Sarvam and Krutrim, supported by prominent VC investors such as Lightspeed Venture Partners are developing AI models tailored for India, indicating a burgeoning ecosystem in this domain.