As China's DeepSeek shakes up the AI race, here is what experts say about building India's own LLMs

DeepSeek, an open-source ChatGPT rival from China, has fundamentally changed the race for global Artificial Intelligence (AI) supremacy. The chatbot, which was released some two weeks ago, poses a serious threat to the American monopoly in the sector.

The manner in which DeepSeek exploded onto the scene has also raised some serious questions in India¡ªif China can do it, why can't India build its own LLMs that could cater to the specific needs of Indian languages?

CREDIT: UNSPLASH

'India will develop own AI model'

Earlier this week, Union Minister for Information Technology Ashwini Vaishnaw said that India will soon develop a world-class foundational AI model that will compete with the likes of ChatGPT and DeepSeek.

India needs its own LLMs

Entrepreneurs and experts also argue that it is high time India built its own LLM.

"India must develop its own LLMs because they are a foundational capability, and existing models are not well-suited for India. The prohibitive cost of training and running these models, along with their lack of support for local languages, makes them less viable for India¡¯s needs," Vinci Mathews, the co-founder of Nofrills AI, a search engine designed to be conversational and multi-modal and support local Indian languages, told Indiatimes.

CREDIT: ADOBE STOCK

He said that if India did not develop its own LLMs, the country risked being left behind¡ªjust as it was in the semiconductor industry.

India will be left behind without own LLMs

According to Abhivyakti Sengar, a Senior Analyst at Everest Group, building an indigenous LLM is an investment in India's future, enabling innovation at scale and safeguarding the country¡¯s place in the AI-driven world economy.

¡°If India does not invest in building its own LLMs, it risks becoming a mere consumer in the global AI economy, losing both digital sovereignty and economic opportunities. Indigenous LLMs are not just about technological advancement¡ªthey are critical for addressing India's unique linguistic diversity, empowering industries with localised solutions, and ensuring competitiveness on the global stage," he said.

CREDIT: SCIENCE VIEW

DeepSeek busted high cost myth

According to Mathews, from the start, the prevailing narrative was that India should focus on building applications and business use cases on top of models from OpenAI, Microsoft, and Google rather than investing time in developing its own models.

"The high cost of training has often been cited as the primary reason for this. However, DeepSeek has demonstrated that architectural innovations can significantly reduce costs while enabling more extensive training," he said.

Mathews also pointed out that DeepSeek was not the first to prove this point¡ªMistral demonstrated the power of better architecture with its mixture-of-experts (MoE) models.

With DeepSeek now proving that training AI models doesn't necessarily take a fortune, Mathews said he is confident that there will be a renewed push to build India's own LLM.

"Hopefully, this shift in perspective will help reshape the narrative, encouraging more investment in startups like ours, which are focused on architectural innovations that make LLMs more affordable and better suited for India¡¯s unique requirements," he said.

CREDIT: NOFRILLS AI

AI search engine for Indian languages

In 2024, Nofrills AI launched India¡¯s first live AI search engine interface that provides real-time answers to user queries in both English and local languages, facilitating hassle-free discovery of information, products, and services online.

"At Nofrills AI, we have already demonstrated that architectural innovations can yield better results for local languages. The way forward is clear: India must focus on developing more efficient model architectures to reduce both training and operational costs," he said.

For more news and current affairs from around the world, please visit Indiatimes News.