BharatGPT Hanooman: AI model developed by Reliance and 9 IITs

BharatGPT Hanooman, Reliance Industries Ltd has teamed up with 9 Indian Institutes of Technology (IITs), including IIT Bombay, to create a new AI (Art

BharatGPT Hanooman

Reliance Industries Ltd has teamed up with 9 Indian Institutes of Technology (IITs), including IIT Bombay, to create a new AI (Artificial Intelligence) model "Hanooman". This big project, supported by the Department of Science and Technology, is designed to solve important AI problems and help India become a leader in the world of AI technology.

The BharatGPT group declared the upcoming launch of Hanooman AI, scheduled for next month. This initiative has garnered support from two major entities: Reliance Industries Ltd and the Department of Science and Technology. Together, they have developed the 'Hanooman' series, which focuses on models that process Indic languages. This development was made possible through a partnership with Seetha Mahalaxmi Healthcare (SML).

The model is one of the biggest in its field, with its complexity varying from 1.5 billion to an impressive 40 billion parameters. This large scale and detailed design enable it to provide detailed and appropriate responses in Indian languages.

Key Points of BharatGPT Hanooman:

Initiative by BharatGPT Group: Spearheaded by IIT Bombay and seven other leading engineering institutes in India.

Upcoming Launch: A service similar to ChatGPT, named 'Hanooman,' focusing on Indic languages, set to launch next month.

Support and Collaboration: Backed by Reliance Industries Ltd and the Department of Science and Technology, in collaboration with Seetha Mahalaxmi Healthcare (SML).

Focus on Indic Language Models: The 'Hanooman' series aims to enhance the processing and understanding of Indic languages through advanced AI models.

This initiative marks a significant step towards promoting the development and use of artificial intelligence in processing and understanding Indic languages, potentially transforming various sectors by making technology more accessible to a wider audience in India.

What is Hanooman?

Hanooman is a project developing advanced AI models capable of understanding and responding in 11 Indian languages, such as Hindi, Tamil, and Marathi, with the goal to eventually support over 20 languages. It was demonstrated through a video by the BharatGPT group, showing people using this AI tool in various languages. The BharatGPT initiative, led by IIT Bombay in collaboration with seven additional IITs, is supported by the Department of Science and Technology, SML, and Reliance Jio.

Key features of Hanooman include:

• Initially supports 11 Indian languages, with plans to expand.
• Aimed at improving services in healthcare, governance, financial services, and education.
• More than just a chatbot, Hanooman can create text, speech, videos, and more in several Indian languages. An example is VizzhyGPT, tailored for healthcare through the analysis of extensive medical data.
• Its AI models vary in complexity, ranging from 1.5 billion to 40 billion parameters.

However, challenges have been noted in the quality of datasets for Indian languages, mainly due to reliance on synthetic datasets. These datasets, often created through translation, may not accurately reflect real-world language use, potentially leading to errors or distortions in the AI's understanding and output.

Besides BharatGPT, there are several other new companies, such as Sarvam and Krutrim, that are creating their own AI models specifically for India. These startups are getting support and funding from well-known investors, including Lightspeed Venture Partners and a fund owned by the billionaire Vinod Khosla. This shows a growing interest in developing technology tailored to Indian needs and languages.

BharatGPT Ecosystem?

The BharatGPT ecosystem refers to a collaborative initiative focused on developing advanced artificial intelligence (AI) technologies, specifically large language models (LLMs) similar to ChatGPT, for the Indian context. This ecosystem is spearheaded by IIT Bombay along with seven other leading Indian Institutes of Technology (IITs). It's supported by significant entities such as the Department of Science and Technology, SML (Seetha Mahalaxmi Healthcare), and Reliance Jio.

The BharatGPT ecosystem aims to create AI models that can understand and generate content in multiple Indian languages, thereby addressing the linguistic diversity of India. The initiative is focused on developing AI tools and technologies that can be applied across various fields such as healthcare, governance, financial services, and education. The ecosystem is designed to foster innovation in AI and make advanced AI technologies accessible and useful for a wide range of applications in India, ultimately contributing to the country's technological advancement and digital transformation.

What are LLMs?

Large language models (LLMs) as super-smart robots that read and learn from a massive pile of books, articles, and websites. They look at all this information to figure out how words and sentences fit together and what they mean. The more stuff they read, the better they get at understanding and making their own sentences.

These models learn from huge collections of text from sources like Wikipedia and various other big websites. By studying this information, they learn how to talk or write in a way that makes sense to us.

Challenges for building Indian LLMs

Building Indian LLMs (Large Language Models) faces significant challenges, especially when it comes to finding good quality data in Indian languages. Here's a simple explanation of the main challenge:

For a language model to understand and generate text, audio, or video accurately, it needs a lot of data to learn from. In the case of Indian languages, finding large amounts of high-quality data can be difficult. This data needs to be not just big in volume but also rich in quality, meaning it should accurately represent the language in various contexts without errors or biases. Since Indian languages are diverse and complex, gathering, cleaning, and preparing this data for training AI models is a tough task. This step is crucial because the better the data quality, the smarter and more accurate the language model will become.

Hanooman AI capabilities

Hanooman is designed with advanced multimodal artificial intelligence (AI) capabilities, enabling it to handle and generate various forms of content. Its capabilities include:

Text-to-Text: Hanooman can generate written content based on text inputs. This means it can take written questions or prompts and produce written answers or content in response.

Text-to-Speech: The AI can convert written text into spoken words. This feature allows Hanooman to read out written text in a natural-sounding voice, making content accessible in audio form.

Text-to-Video: It can create video content from text inputs. This involves generating visual content, possibly including animations or visualizations, that corresponds to the information or narrative provided in the text.

Vice Versa Content Generation: Besides the above capabilities, Hanooman can also perform the reverse operations, such as converting speech to text, video content to text, or video content to speech. This flexibility enables it to function in diverse scenarios, catering to different user needs and preferences.

These multimodal capabilities make Hanooman a versatile AI tool, capable of serving various purposes across sectors like healthcare, education, governance, and financial services. Its ability to understand and generate content in multiple formats allows for more interactive and accessible communication and information dissemination.

The applications of Hanooman, given its advanced AI capabilities, span across various sectors. SML (Seetha Mahalaxmi Healthcare) is exploring partnerships with different industries to leverage Hanooman's potential. Here's how Hanooman can be applied:

BFSI (Banking, Financial Services, and Insurance) Enterprises: Hanooman can be used to create AI-driven solutions for customer service, such as chatbots that handle queries, provide financial advice, or assist in the processing of transactions and claims. It can also analyze large volumes of financial data for insights, enhancing decision-making processes.

Healthcare Organizations: In healthcare, Hanooman can support by offering personalized medical advice through AI chatbots, analyzing patient data to assist in diagnostics, or even providing training and educational content for medical professionals. Its multimodal capabilities allow for diverse applications, such as creating visual content for patient education or converting medical texts into speech for accessibility.

Mobile App Providers: For mobile apps focusing on education, entertainment, or utility, Hanooman can offer personalized content generation, language translation services, and voice-enabled functionalities. It can enhance user engagement through interactive AI features and support local languages, making apps more accessible to a broader audience.

By offering the Hanooman series as a model-as-a-service or creating specialized models through fine-tuning, SML aims to provide tailored AI solutions that meet the specific needs of these sectors. This approach allows businesses and organizations to harness the power of AI for improving efficiency, customer satisfaction, and innovation.

Reference:

• Indian Express
• Times of India