 "AI")
Despite these concerns, Indian AI startups are optimistic about the technology. Image: Shutterstock
By Saritha Rai
Earlier this month, executives from Alphabet Inc.’s Google DeepMind, Microsoft Corp. and Meta Platforms Inc. joined tech founders in Bangalore to watch one of India’s top AI startups unveil a new product that could change how the world’s most populous country uses technology.
Sarvam AI, often described as India’s OpenAI, has launched software for businesses that can interact with customers using voice rather than just text. The technology was developed using data from 10 native Indian languages and is being priced at a rupee per minute to capture the market. In a video presented at the event, Vinod Khosla, a billionaire venture capitalist and investor in Sarvam, said, “These voice bots have the potential to reach a billion people.”
Since ChatGPT launched nearly two years ago, India has tried to keep pace with the global AI craze, but chatbots have often been limited by a lack of data on many of the country’s languages. Many people in big cities can type messages to a chatbot in English, but most Indians lack the language skills to do so. Now, a growing number of startups are betting that voice bots built on local language data can reach more of India and perhaps even appeal to users in other countries.
In doing so, these startups could make India a testing ground for what could be the next frontier of generative AI products, even though the latter has raised security concerns in other markets. By integrating AI voice capabilities, tech companies hope to create more dynamic conversational services that can respond verbally to users in real time and automate certain tasks. In India, this is already happening in a wide range of consumer and commercial applications.
Samsung-backed Gnani AI conducts millions of voice conversations every day for India’s largest banks, insurance companies and carmakers. CoRover AI provides voice bots in 14 Indian languages to the state-run railway and a regional police force. And Haloocom Technologies’ voice bot can speak in five Indian languages to handle customer service tasks and help screen job candidates.
“The world has moved from digital to mobile to AI, but voice is the most intuitive way to use technology,” said Ankush Sabharwal, Co-Founder and CEO, CoRover.
CoRover’s Ask Disha voice bot went live this month for Indian train booking company IRCTC. The bot can book train tickets and make payments on behalf of a customer using just voice. The country needs AI agents that can perform tasks, not just provide information, Sabharwal said.
Gnani offers a bot to help lenders engage with potential customers to determine their financial needs, collect personal information, and determine their loan eligibility. The startup is also working with one of India’s largest automakers, Tata Motors Ltd., to get feedback on the latest car models and sell extended warranties and accessories.
Sarvam’s voice bots can handle conversations in multiple languages and take actions for customers, such as scheduling appointments and facilitating payments. The company has about 50 clients, including Sri Mandir, a devotional app that has over 10 million downloads on the Android Play Store. Using Sarvam’s voice software, Sri Mandir’s app can guide people to specific rituals at different temples and instruct them on how to seek different types of blessings.
“Try running GPT-4 or Claude on Sri Mandir. I can guarantee you it won’t work,” said Vivek Raghavan, co-founder of Sarvam, referring to cutting-edge AI models from OpenAI and Anthropic. American companies don’t have access to enough data on the Indian language spoken, he said, including accents that vary from region to region.
Some of the leading AI companies in the US, including OpenAI, have developed technology that can generate convincing voices but have been slow to commercialize them. OpenAI recently warned that users could become emotionally dependent on its voice product and also said it has taken steps to prevent impersonation and the generation of copyrighted audio files. The startup has begun rolling out new voice features to a limited number of users after a delay.
Despite these concerns, Indian AI startups are optimistic about the technology. “AI designed for specific use cases, languages and audiences is more accurate, less costly to operate and has significantly reduced hallucinations,” said Ganesh Gopalan, co-founder and CEO of Gnani, using a term that refers to AI systems that fabricate facts.
While these startups are focused on India, some are also targeting international markets, including the Middle East and Japan. In fact, Gnani’s voice bots are already being deployed in Silicon Valley, helping a major California-based Harley-Davidson leasing company reach Spanish-speaking customers.
First published: August 26, 2024 | 07:13 IST
