Earlier this month, IT entrepreneurs in Bangalore were joined by executives from Alphabet Inc.’s Google DeepMind, Microsoft Corp., and Meta Platforms Inc. to witness one of India’s leading AI firms launch a new product that might revolutionise the use of technology in the most populated nation in the world.
Software that allows businesses to communicate with clients using spoken voice instead of text was introduced by Sarvam AI, also referred to as India’s OpenAI. The technique was created using information from ten native Indian languages and was offered for one rupee per minute in order to gain traction. The billionaire venture capitalist and Sarvam investor Vinod Khosla stated in a video played during the event that “These voice bots have the potential to reach a billion people.”
Over the nearly two years since ChatGPT started, India has attempted to keep up with the global artificial intelligence frenzy. However, the absence of data on many of the country’s languages has often limited the capabilities of chatbots. While most Indians are not proficient in English, many people who live in large cities are able to type commands to a chatbot in the language. Increasingly, startups are placing their bets on the idea that speech bots developed using local language data can potentially appeal to people outside of India and cover a larger portion of the country.
While raising some safety concerns in other markets, these businesses may also make India a testing ground for what might be the next wave of generative AI products. Tech businesses want to build more dynamic, conversational services that can converse with users orally in real time and automate certain processes by integrating AI speech features. That’s already the case in India for a variety of consumer and corporate apps.
The biggest banks, insurers, and automakers in India rely on Samsung-backed Gnani AI to handle millions of voice chats every day. The state-owned railway company and a local police force can use voice bots provided by CoRover AI in 14 Indian languages. Additionally, Haloocom Technologies’ speech bot can assist with candidate screening and customer care chores in five Indian languages.
Voice is the most natural way to utilise technology, according to Ankush Sabharwal, co-founder and CEO of CoRover. “The world went from digital first to mobile first to AI first,” Sabharwal said.
This month, IRCTC, the Indian train reservation provider, launched CoRover’s Ask Disha voice bot. With voice commands alone, the bot can make rail ticket reservations and handle payments on behalf of users. According to Sabharwal, the nation requires AI agents that can do jobs in addition to providing information.
With the use of a bot provided by Gnani, lenders may interact with prospective borrowers to learn more about their financial situation, gather personal data, and assess their loan eligibility. In addition, the firm collaborates with Tata Motors Ltd., one of the biggest automakers in India, to market accessories and extended warranties as well as gather input on the newest car models.
In addition to handling multilingual conversations, Sarvam’s speech bots can perform tasks for clients including appointment scheduling and payment processing. About 50 clients make up the company, one of which being Sri Mandir, a devotional software that has amassed over 10 million downloads from the Android Play Store. The Sri Mandir app uses Sarvam’s voice software to direct users to particular temple rites and teach them how to pray for different kinds of blessings.
“At Sri Mandir, try hurling GPT-4 or Claude. With a reference to state-of-the-art AI models from OpenAI and Anthropic, Sarvam co-founder Vivek Raghavan said, “I can assure you that it won’t work.” According to him, spoken Indian language data—including regionally specific accents—is not sufficiently available to US corporations.
Though they have taken their time getting their technology to market, some of the top AI companies in the US, including OpenAI, have created technology that can produce realistic voices. In addition to announcing that it has taken precautions to avoid impersonations and the production of copyrighted sounds, OpenAI also issued a warning that users may grow emotionally dependent on its voice product. After a pause, the firm is now gradually handing out additional voice functions to select users.
The AI startups in India are upbeat about the technology despite the worries. Ganesh Gopalan, co-founder and CEO of Gnani, used a word that refers to AI systems falsifying facts to say, “AI made for specific use cases, languages, and audiences is more accurate, less expensive to run and has vastly reduced hallucinations.”
While many of these firms are concentrated on India, others have their sights set on other regions, such as the Middle East and Japan. Actually, a major Harley-Davidson lease company in California is using Gnani’s voice bots to help it reach Spanish-speaking clients in its backyard in Silicon Valley.