Stay up thus far with free updates
Simply register for Artificial intelligence myFT Digest – delivered straight to your inbox.
Global technology firms and native startups wish to tap into lucrative recent markets in India with artificial intelligence platforms adapted to the big selection of languages and industries on the planet's most populous country.
Microsoft, Google and startups like Silicon Valley-based Sarvam AI and Krutrim – founded by Bhavish Aggarwal of Indian mobility giant Ola – are all working on AI voice assistants and chatbots that work in languages like Hindi and Tamil. The tools are geared toward fast-growing Indian industries, reminiscent of the country's large customer support and call center sector.
India has 22 official languages, with Hindi being essentially the most widely spoken, but researchers estimate that the languages and dialects spoken by the 1.4 billion people number within the 1000’s. Google on Tuesday launched its AI assistant Gemini in nine Indian languages.
Microsoft's AI assistant Copilot is offered in 12 Indian languages, and the corporate is working on other projects tailored to India, including developing “tiny” language models at its research center in Bengaluru. These smaller alternatives to the expensive large language models that underpin generative AI can run on smartphones quite than within the cloud, making them cheaper and potentially higher suited to countries like India where connectivity will be limited.
Microsoft desires to “make AI easy and straightforward to make use of and accessible to all those customers and partners,” Puneet Chandok, Microsoft president for India and South Asia, told the Financial Times, adding that that is about “adapting AI to the Indian context and making it more relevant and accurate.”
Microsoft can also be working with Sarvam AI. Founded just last yr, the Bengaluru-based company is developing a full stack of generative AI tools for Indian firms. The startup has raised $41 million from investors including Peak XV, Sequoia's former Indian arm, and Menlo Park-based Lightspeed Venture Partners.
Hemant Mohapatra, partner at Lightspeed, said investing in local AI firms is becoming increasingly vital as governments seek to develop “sovereign AI” that’s trained and stored inside their borders.
“The supply chain for AI is beginning to fragment,” Mohapatra said. “If you train a base model in India with data from Indian residents, audio, video, text, different languages, then it needs to be an Indian company focused on Indian use cases, based in India, having Indian founders and so forth.”
India's AI race doesn’t involve constructing LLMs from scratch to compete with leaders like Open AI. Investors argue that the resources and capital required can be too high to make sense.
Instead, firms like Sarvam AI are specializing in adapting existing LLMs for Indian languages and using voice data as an alternative of text, making them simpler in a rustic where many prefer to speak via audio messages quite than written messages.
“In countries as complex as India, there continues to be a giant gap between these underlying models and real-world use cases,” said Lightspeed partner Bejul Somaia. “In a market like India, a small ecosystem must emerge that permits firms to leverage the underlying model capabilities.”
Tanuja Ganu, manager at Microsoft Research in Bengaluru, said one other advantage of testing recent technologies and tools in a rustic the dimensions and variety of India is that they will be exported elsewhere.
“We are using India as a test bed to validate among the technologies in India and see how we are able to scale them as much as other parts of the world,” she said.
