[ad_1]
BENGALURU:Ola founder Bhavish Aggarwal on Friday announced a “Made for India” ChatGPT equivalent that he said will support 10 Indian languages. Several ventures in India are working on similar initiatives, based partly on the government’s Bhashini mission that is building an Indian language database and which can be used to train AI models.
Called Krutrim, Sanskrit for artificial, Aggarwal’s AI company will have an identity separate from Ola’s mobility and EV businesses.Aggarwal said the company was incorporated as Krutrim SI Designs in April with teams based in Bengaluru and San Francisco.
Krutrim will debut with a beta version in January that enterprises and developers will be able to use to build chatbots for their specific purposes. The demonstration showed the system can switch between languages and discuss nuanced topics ranging from poetry in Bengali, to Bollywood movies, to creative masala dosa recipes.
Aggarwal said models trained specifically on Indic languages have performed better than open-source large language models. “It is our time to realise the dream of Viksit (advanced) Bharat and lead the AI vision over the next 25 years. AI is going to define the future paradigm of economy and culture. AI is going to be one of the most important technologies of the future. We believe India needs to build its own stack,” said Aggarwal. “India’s linguistic diversity and cultural context is very different from the West and that has implications on how AI is architected. With an India first cost structure, we can reach 1.4 billion people,” he said.
Their system, he said, has been trained on datasets that are proprietary and those obtained from partners.
Aggarwal didn’t disclose the pricing, but said it’s a family of models and each one has a certain price-performance balance built into it. “In addition to these language models, we also have speech models. So, you can talk to the AI models,” he added.
Aggarwal also has ambitious plans to build their own AI chips in two years. “We have a reasonable amount of GPU resources already. Eventually we will be building our own AI supercomputer and our silicon. We’re also building our own technology for data centres; it’s very important to make it more efficient to bring down costs,” he said.
Called Krutrim, Sanskrit for artificial, Aggarwal’s AI company will have an identity separate from Ola’s mobility and EV businesses.Aggarwal said the company was incorporated as Krutrim SI Designs in April with teams based in Bengaluru and San Francisco.
Krutrim will debut with a beta version in January that enterprises and developers will be able to use to build chatbots for their specific purposes. The demonstration showed the system can switch between languages and discuss nuanced topics ranging from poetry in Bengali, to Bollywood movies, to creative masala dosa recipes.
Aggarwal said models trained specifically on Indic languages have performed better than open-source large language models. “It is our time to realise the dream of Viksit (advanced) Bharat and lead the AI vision over the next 25 years. AI is going to define the future paradigm of economy and culture. AI is going to be one of the most important technologies of the future. We believe India needs to build its own stack,” said Aggarwal. “India’s linguistic diversity and cultural context is very different from the West and that has implications on how AI is architected. With an India first cost structure, we can reach 1.4 billion people,” he said.
Their system, he said, has been trained on datasets that are proprietary and those obtained from partners.
Aggarwal didn’t disclose the pricing, but said it’s a family of models and each one has a certain price-performance balance built into it. “In addition to these language models, we also have speech models. So, you can talk to the AI models,” he added.
Aggarwal also has ambitious plans to build their own AI chips in two years. “We have a reasonable amount of GPU resources already. Eventually we will be building our own AI supercomputer and our silicon. We’re also building our own technology for data centres; it’s very important to make it more efficient to bring down costs,” he said.
