BENGALURU: In what is touted as India’s answer to ChatGPT, ride-hailing firm Ola on Friday unveiled the country’s first multi-lingual AI model Krutrim, which means artificial in Sanskrit.
While ChatGPT and other major AI models can respond to queries only in English, Krutrim understands as many as 22 Indian languages and can give responses in 10 languages — English, Hindi, Tamil, Bengali, Gujarati, Marathi, Malayalam, Telugu, Kannada, and Odia. Ola CEO Bhavish Aggarwal claims Krutrim will be able to process more Indian languages going forward.
During the launch event, Aggarwal demonstrated how Krutrim can churn out stories and poems in Indian languages. Explaining the AI’s features, he said other models are trained largely in English. Also, foreign AI models can’t capture the nuances of Indian languages and culture. In this context, he said, Krutrim surpasses all others as it has been trained on unique data sets specific to the Indian scenario.
In technical terms, Krutrim is a family of large language models (LLM). It has been trained on over 2 trillion ‘tokens’ by a team of scientists based in Bengaluru and San Francisco. Tokens are the basic units of text or code that an LLM AI uses to process and generate language.
Krutrim will be available in two versions – the base model and Pro. The latter version will have multimodal, larger knowledge capabilities and other technical advancements for inference.
The beta version will be out in January for enterprises and developers, while Krutrim Pro is expected to be available in Q4 of FY24.