Mukesh Ambani’s Reliance Industries Ltd. is all set to launch the India’s own ‘ChatGPT-like’ chatbot AI Hanuman, next month. Reliance is reportedly working 8 affiliated universities for this project which was named the ‘Bharat GPT Group’. The company unveiled the groundbreaking features of the Chatbot in a Technology conference on Tuesday. Lets delve into what they are:
AI Hanuman is a series of Large Language Models (LLM) specifically trained in 22 Indian Languages. This ambitious project was termed “BharatGPT” with the aim of making AI tools accessible to the diverse linguistic landscape of India. The tool isn’t just limited to text however as it features speech-to-text capabilities along with “multi-modal AI capabilities”, which means that it can create and understand content across different formats like text, photos and even videos.
Bridging linguistic gaps and going beyond language:
Due to the diversity of languages in India, AI development faces a unique challenge which AI Hanuman aims to address. It currently supports 11 languages with the aim of covering all 22. But beyond offering simple translation, it can also provide various online customer services in local languages. It can provide personalized education platforms which can adapt to individual learning styles. It can even provide AI-powered assistants assisting farmers with crop selection in their own native tongue.
Its features were boasted during the presentation as a motorcycle mechanic from Southern India ‘spoke’ to the chatbot in his own native language, Tamil. Also during the presentation, a banker interacted with the chatbot in Hindi and a developer from Hydrabad used the tool to write computer code. But building such a complex model will offer up tough challenges, such as sourcing quality data from multiple languages, which can be quite tough. It also face competition of other LLMs like Ola’s Krutrim and IIT-Madras’ Airavata.
What is an LLM?
A Large Language Model (example: ChatGPT) is a deep learning algorithm which is trained using large datasets from numerous sources. This enables them to translate, generate and subsequently predict text and other related content. LLMs are also known a Neural Networks, as they are computing systems inspired by the human brain. LLMs can trained for many tasks like generating content, understanding complex structures, writing softcare code etc.