Indian AI Model: AI i.e. Artificial Intelligence is the name of a technology which is spreading rapidly across the world. This technology is being used from America to India. Using this technology, leading tech companies around the world are also creating AI models that look and work like humans. India is also not behind in this work and an AI model has been prepared in India too, whose name is Hanooman i.e. Hanuman. Let us tell you the detailed information about this Made in India AI model.
Last Tuesday, BharatGPT, a group led by IIT Bombay and seven other engineering institutes of India, announced that it will launch India’s first ChatGPT-like service next month i.e. in March. Backed by Mukesh Ambani’s company Reliance Industries Limited and the Ministry of Science and Technology, the group has produced the ‘Hanuman’ series of Indic language models in collaboration with Sita Mahalaxmi Healthcare (SML).
What is Hanuman?
Hanuman is a series of Large Language Models (LLMs) that can respond to 11 Indian languages, such as Hindi, Tamil, and Marathi, with plans to expand to more than 20 languages. According to Bloomberg report, BharatGPT Group showed different people in different languages interacting with AI tools in a video last Tuesday. A banker spoke to this AI model Hanuman in Hindi language and a bike mechanic spoke in Tamil language. Made in India, this AI model Hanuman has been designed to work in four sectors including health care, governance, financial services and education.
It is worth noting that this series does not just work like a chatbot. According to BharatGPT, it is a multimodal AI tool that can generate text, speech, video and more in multiple Indian languages. One of the first customized versions is VizzhyGPT, which is an AI model. It is designed for health care using a range of medical data. The size of these AI models ranges from 1.5 billion to 40 billion parameters.
What is LLM?
Large Language Models (LLM) use deep learning techniques for large text content. By processing large amounts of text, they understand and learn its structure and meaning and then work accordingly. LLMs are ‘trained’ to identify meanings and relationships between words. The more amount of training data a model is given, the more efficient it becomes at understanding and producing text.
Training data are usually large datasets such as Wikipedia, OpenWebText, and Common Crawl Corpus. These contain large amounts of text data, which models use to understand and generate natural language.
Are there any other LLM?
Apart from BharatGPT, separate startups like Sarvam and Kritrim, backed by prominent VC investors like Lightspeed Venture Partners and billionaire Vinod Khosla, are also building customized AI models for India, Bloomberg reported.