AI In Hindi — Nvidia CEO Announces New Model, Says Language Hardest To Map

Artificial Intelligence (AI) computing giant Nvidia is launching an AI model for the Hindi language, CEO Jensen Huang announced in Mumbai at the Nvidia AI Summit.

“India is the hardest regional language model area in the world. Once Hindi is figured out, we can do it for any other country,” he said. 

The AI model released is a small language model and is available as an NVIDIA NIM microservice. The model is dubbed Nemotron-4-Mini-Hindi-4B. It can be deployed on any NVIDIA GPU-accelerated system for optimised performance.

IT firm Tech Mahindra is the first to use the model to develop an AI model called Indus 2.0, which is focused on Hindi and dozens of its dialects.

Innovators, major enterprises and global systems integrators across India are building customised language models using NVIDIA NeMo, said the company. Sarvam AI, Gnani.ai, Flipkart, Krutrim, and Zoho are some of the companies using the model. 

India’s systems integrators are also offering NVIDIA NeMo-accelerated solutions to their customers. Infosys, TCS and Wipro are the companies that have been using the model for various business functions. Huang at the summit said, “Wipro, Infosys will help take Nvidia's AI models to the globe.” 

The Nvidia CEO is his address said that while India exports software, in the future it will export AI.

To produce AI, infrastructure is needed, which requires energy. The future will have a new industry called production of intelligence. Further, the production of AI will be the new industrial revolution, he said.

“India was focused on delivery of software but the next generation will be about the delivery of AI, these are fundamentally different. I don't know if anyone else has the ability that India has to help every single country of the world to enjoy the benefits of AI,” Huang said.