The AI industry is racing to adapt chatbots to India’s many languages
Stay up to date with free updates
Simply sign in Artificial intelligence myFT Digest — delivered directly to your inbox.
Global tech companies and local startups are looking to unlock lucrative new markets in India with artificial intelligence platforms tailored to multiple languages and industries in India. most populous country in the world.
Microsoft, Google and startups including Sarvam AI and Silicon Valley-backed Krutrim – founded by Bhavish Aggarwal of Indian mobile group Ola – are all working on artificial intelligence Voice assistants and chatbots work in languages like Hindi and Tamil. The tools are aimed at India’s rapidly growing industries, such as the country’s large customer service and call center sectors.
India There are 22 official languages, with Hindi being the most common, but researchers estimate the number of languages and dialects spoken by its 1.4 billion people is in the thousands. Google on Tuesday launched its Gemini AI assistant in nine Indian languages.
by Microsoft Co-pilot AI assistant is available in 12 Indian languages, and the company is working on other projects tailored for India, including building “small” language models at its Bengaluru-based research center. These smaller alternatives to the expensive large language models that underpin general AI can run on smartphones instead of in the cloud, making them cheaper and potentially more suitable for countries like India where connectivity may be limited.
Microsoft wants to “do [AI] simple, easy to use and put it in the hands of all these customers and partners,” Puneet Chandok, president of Microsoft India and South Asia, told the Financial Times. He added that this involves “contextualizing it to the Indian context, making it more relevant, more accurate”.
Microsoft is also partnering with Sarvam AI. The Bengaluru-based company founded last year is developing a “full suite” of general AI tools for Indian businesses. The startup has raised $41 million from investors including Sequoia’s former Indian subsidiary Peak XV and Menlo Park-based Lightspeed Venture Partners.
Lightspeed partner Hemant Mohapatra said investing in local AI companies is becoming more and more important as governments look to develop “sovereign AI” trained and hosted within their borders. Surname.
“The AI supply chain is starting to fragment,” said Mohapatra. “If you are training a platform model in India based on data, audio, video, text, different languages of Indian citizens then it should be an Indian company, focused on Indian use cases, resident Indians, Indian founders, etc.
India’s AI race does not involve building an LLM from scratch to compete with leaders like Open AI. Investors assumed that the resources and capital required would be too great to comprehend.
Instead, companies like Sarvam AI are focusing on adapting existing LLMs for Indian languages and using voice data instead of text. This makes them more effective in a country where many people prefer to communicate via audio messages rather than by text.
“There is still a huge gap between these basic models and real-world use cases in complex countries like India,” said Bejul Somaia, partner at Lightspeed. “In a market like India, you will need to have a robust development ecosystem to enable companies to use the capabilities of the underlying model.”
Tanuja Ganu, director at Microsoft Research in Bengaluru, says another benefit of testing new technologies and tools in a country of India’s size and diversity is that they can be exported. other places.
“It is using India as a testing ground and validating some of the technology in India and seeing how we can expand it to other parts of the world,” she said.
