September 12, 2024 05:36 PM
Himanshu Sharma
A UAE-based AI technology firm, G42, in collaboration with Microsoft, has unveiled a new ‘large language model,’ or LLM, named ‘Nanda’ customised for Hindi consumers in India. The launch was conducted during the UAE-India Business Forum in Mumbai with approval of the high officials of both friendly countries.
Nanda: A Bilingual LLM for India:
Nanda, which is derived from India’s second largest peak of Nanda Devi, is a 13-billion parameter LLM that has been trained on a humongous data set of 2.13 trillion. Hindi has been represented by these trillion tokens, along with English and what can be termed as ‘Hinglish.’
“We are completely confident that #Nanda can be woven into the fabric of India” The CEO of G42 India, Manu Kumar Jain in his tweet stated this on X. It means that more than half a billion of Hindi language speakers will be offered the chance to unleash generative AI potential.
Jais: Language preferences of previous G42’s Arabic LLM:
This is after G42releasedJais last year, an open-source Arabic LLM developed for Arabic natural language processing. Jais was released with different model sizes and one of the biggest model size being 70 billion parameters.
Microsoft's Investment in G42:
In April 2024 Microsoft invested one billion dollars. five billion dollars in G42, and grew to become one of its largest shareholders while gaining a position on its board. This investment also reflects the firm’s strategic objective of enhancing AI proficiency together with growth in new frontiers.
Use Cases and Availability:
G42 has not revealed concrete applications that Nanda can support, nor the extent to which it will be open to the public. I wonder if LLM is going to be released for the public use, or it is going to function only in governmental and enterprise sectors.
Even though it still has limitations, development of Nanda is a step towards making AI technology more accessible to the people who speak Hindi in the country hence creating new prospects for innovations and inclusion in the information society.