Large language models for specific applications in Indian languages are being constructed by Sarvam AI startup with the goal of releasing apps at a cost that the 1.4 billion people living in the nation can afford. Peak XV Partners, formerly Sequoia Capital India & Southeast Asia, and Silicon Valley billionaire Vinod Khosla’s Khosla Ventures participated in the Series A fundraising round, which was headed by Lightspeed Venture Partners, Sarvam announced on Thursday.
Vivek Raghavan and Pratyush Kumar co-founded the Bangalore-based firm, which was operating in secret mode until recently. Its ambitious goals include developing tools for developing apps for the limited computational resources available to developers in India, as well as open source foundational AI systems. The startup, which just opened an office in the city, employs 18 people in total.
In a phone conversation, Raghavan stated, “What we are demonstrating is that you can build large language models with limited resources.” “We are constructing on a smaller scale and showcasing their exceptional cost and energy efficiency, making them accessible to all.”
Big language models are strong artificial intelligence (AI) systems that comprehend by learning from enormous volumes of heterogeneous data from the internet and elsewhere to summarize, translate, and create text, audio, and video for a wide range of applications. Examples of these models are OpenAI’s GPT-4 and Meta Platforms Inc.’s Llama. In Silicon Valley, there is a growing competition to create ever-more complex models, and billions of dollars have been invested in these attempts.
In the upcoming weeks, Sarvam will release its first open source model, which will be trialed by developers, entrepreneurs, government-owned entities, and companies in the top ten Indian languages. Largest models have been trained on trillions of parameters; Sarvam’s systems will be “much smaller” and in the “billions,” according to Raghavan.
According to him, Sarvam’s name is derived from an old Sanskrit term that means “all,” signifying its comprehensive goal. While Kumar is an ETH Zurich graduate and a graduate of the Indian Institute of Technology Bombay, Raghavan holds a PhD in computer engineering from Carnegie Mellon University.
Given their geopolitical significance, a number of nations are spearheading sovereign efforts to develop generative AI models, according to Khosla, a pioneer in Silicon Valley investment and an early supporter of OpenAI. In a statement, he stated, “We need businesses like Sarvam AI to develop deep expertise for building AI in and for India.”