The development of these models aims to address the need for more localized experiences, as foundation models often struggle with non-English languages. Tech Mahindra is working on an open-source large language model under The Indus Project, while Reliance Industries is partnering with Nvidia to build AI infrastructure. Other players in the field include Sarvam AI, which has built a 7 billion parameter Indic language model, and Google-backed Corover, which claims to have built an indic language model supporting 22 languages for its platform for conversational enterprise chatbots.
Key takeaways:
- Indian startup Ola Krutrim, founded by Bhavish Aggarwal, is making strides in generative AI and is set to compete with OpenAI’s GPT models. The company has debuted its first language model, Krutrim base, and a chatbot built on top of it.
- Other Indian companies, including Tech Mahindra and Reliance Industries, are also in the race to develop generative AI models, with a focus on improving localized experiences and handling queries in non-English languages.
- Ola Krutrim's model understands 20 Indian languages and generates 10, including Hindi and English. The company claims its performance across Indic languages is already better than GPT-4, although its English quality performance is still behind.
- As the generative AI ecosystem in India evolves, more sophisticated closed and open-source Indic language models are expected to emerge, improving internal enterprise workflows and leading to better applications for organizations operating across different sectors.