Sarvam AI can train competitive models for a fraction of Big Tech’s cost, co-founder Pratyush Kumar says

At the India AI Impact Summit in New Delhi, Pratyush Kumar, co-founder and CEO of Sarvam AI, delivered a compelling message about the evolving landscape of artificial intelligence (AI). He emphasized that the belief that only a select few tech giants can create advanced AI models is becoming outdated. In a conversation with Business Today, Kumar highlighted how his Bengaluru-based startup has demonstrated that powerful AI systems can be developed with significantly less investment than previously thought. Kumar asserted that advancements in technology have made it feasible to create effective models at a reduced cost, stating, "We are in a deflationary world of AI itself." Over the past two weeks, Sarvam AI has rolled out an extensive range of products, including speech and vision systems, developer tools, state partnerships, and consumer hardware, all strategically timed for the summit. Among its latest offerings are large language models (LLMs) aimed at global competitiveness while prioritizing Indian languages and data. Kumar challenged the notion that AI development is monopolized by a few companies, suggesting that these barriers are diminishing. He believes the focus for India should not be merely on catching up with the United States or China but rather on ensuring that AI technology is widely accessible to the general populace. "The goal is that the technology should be accessible to all," he remarked. To facilitate this access, Sarvam is committed to integrating AI into daily life through both cloud-based systems and innovative hardware. Recently, they unveiled smart glasses designed and produced in India, which Kumar believes can empower individuals with AI capabilities wherever they go. Beyond hardware advancements, Sarvam is expanding its presence in the AI industry by launching a vision-language model specifically for document understanding, trained on diverse multilingual data. The company is also forming partnerships with state governments to create AI-optimized data centers for public sector applications, alongside releasing open-source foundational models aimed at developers and enterprises. Kumar emphasized the importance of controlling training data to develop reliable systems that cater to local needs. He explained that the quality and biases of a language model are directly influenced by the data it is trained on. Despite focusing on Indian datasets, he reassured that these models are not isolated from global knowledge, stating, "These are general-purpose models…" He noted the necessity of aligning models with societal values after exposing them to worldwide information. With India being a voice-first nation, Kumar highlighted the need for AI systems that support multiple languages and function effectively in noisy environments. He anticipates a swift increase in enterprise adoption of AI technology, particularly in automated voice interactions, such as customer service. "Voice AI will become an infrastructure," he predicted. Although Sarvam's ambitious growth plans will require substantial capital, Kumar assured that the company is being judicious in its funding decisions. He sees the mission of Sarvam not just as creating a singular breakthrough product but as positioning India as a significant player in the global AI arena. "India should be building its own thing," he concluded, signaling a new era where AI influences every industry.