Also 'India makes its entry into the field of artificial intelligence. The Indian Republic (Bhārat in Hindi) is preparing to launch its Hanooman language model, a large language model (Llm) similar to ChatGPT, able to operate in 11 of the 22 official languages of the country. The official launch of Hanooman is scheduled for next month.
Presented by Seetha Mahalaxmi Healthcare (Sml) in collaboration with theBharatGPT ecosystem, led by IIT Bombay, led by IIT, Hanooman aims to revolutionize key sectors such as governance, healthcare, education and financial services. The BharatGPT ecosystem is a research consortium led by the Indian Institute of Technology (IIT) Bombay, supported by seven other IITs, besides the Department of Science and Technology, SML. In addition to the support of the government and Indian universities, the project also sees the involvement of a branch of Reliance Industries, Reliance Jio, one of India's main mobile operators, led by thirty-year-old Akash Ambani, son of Mukesh Ambani, the richest man in Asia.
India: what is Hanooman and how it works
Hanooman was named in honor of the Hindu deity Hanuman known for his strength and intelligence. The aim is to make advanced AI technologies accessible in up to 11 local languages, serving a diverse Indian demographic. It is a set of large indexed language models trained on 22 Indian languages, with variants of up to 40 billion parameters.
Hanooman has multimodal AI capabilities to generate content in different formats, such as text, voice and video. Ganesh Ramakrishnan, one of the initiators of the project and head of the Department of Engineering and Computer Science at IIT Bombay, said that Hanooman will enable the conversion of voice to text, thus facilitating access to information for millions of Indians who are unable to read or write. In the future, the model will also be extended to languages not considered in the initial version.
One of the major challenges in building large language models in India is obtaining high-quality datasets in local languages. Improving the quality of these datasets, which include text, audio and video in Indian languages, is a critical first step. Hanooman joins other language models in the AI race, such as Ola's Krutrim, SaravamAI's OpenHathi and IIT-Madras' Airavata model.
With regard to the regulation of AI, the Indian government is expected to come up with a draft AI regulatory framework by July 2024.
Artificial intelligence: India's challenges in the sector
India, despite being known for its inclination towards technology, had so far remained somewhat behind in the field of Artificial Intelligence. With the advent of Hanooman, this is expected gap is filled, marking an important step forward in the Asian country's goal to emerge as a global leader in AI.
Furthermore, the project represents a big challenge also in the linguistic field. The geographic diversification of the research consortium was deliberate to make Artificial Intelligence accessible to a large portion of India's non-English speaking population. Then there is also the native linguistic problem. India has around 22 languages. Currently, models can operate in 11 Indian languages which include some of the Indo-Aryan languages such as Hindi, Bengali and Marathi, spoken in Delhi, Kolkata and Mumbai respectively and also Southern Dravidian languages such as Tamil, Telugu and Malayalam. The future goal will be to extend the capabilities of the model to cover all 22 Indian languages.
There is no shortage of some doubts on the good intentions of the project. Strangely, among the first eleven languages inserted, theabsence of Urdu, the seventh most spoken language in India and predominant among the Muslim population. A signal, together with the choice of the name of the model and the consortium, which raised questions about the intentions of inclusion of the project. This lack is, in fact, particularly significant given that the Muslim community has lower social and income indicators and could benefit more from direct access to educational and health information, especially for women.
One of the project's promoters, Ganesh Ramakrishnan, head of the Department of Computer Science and Engineering at IIT Bombay, explained that Hanooman will offer the possibility of converting voice into text, to help the millions of Indians who cannot read or write , and in the future it will be extended to languages not covered in the initial release.
