Rs 10K crore, 38K GPUs, and 12 firms later, India’s AI mission is still looking for users

In a bid to harness the potential of artificial intelligence for the betterment of society, India has embarked on a mission that has allocated approximately Rs 10,000 crore in funding. The government has selected 12 firms, focusing on developing indigenous AI products tailored for government work and public welfare. One such project is Project Indus, an open-source large language model (LLM) developed by Tech Mahindra.

This model is designed to support Hindi and its dialects, with special attention to factors such as latency, dialect variations, cost, and accuracy.

However, despite the substantial investment and active development, the challenge of convincing government entities to adopt these AI solutions remains. During the initial phase of Project Indus, officials were more interested in understanding how the technology would fit into their existing workflows rather than simply evaluating the model's performance.

This shift in focus highlighted the need for more than just a technically proficient AI solution; it required a product that aligns with the functional needs of the government.

The government's ambitious plan, announced in March 2024, aimed to leverage AI for improving educational outcomes, assisting in medical diagnostics, and preserving Indic languages. While progress has been made, converting these technological advancements into practical, government-approved applications has proven to be a significant hurdle.

Qualcomm's AI hub has embraced a version of Project Indus, known as IndusQ, which is optimized for use with Qualcomm's AI chips. These chips are integral to the government's smart city initiative, which includes a wide range of applications like sensors and cameras.

Despite the advancements, the primary obstacle remains the integration of these AI solutions into the government's operational frameworks. The government has expressed a desire for AI that not only performs well but also seamlessly integrates into their existing systems, a requirement that many of the developed products have yet to fully meet.