The vital question in the present context is: Where is backing from the Indian government essential for the advancement of Indian LLMs, and can Indian AI startups feasibly create LLMs without governmental assistance?

Moreover, this is a comprehensive contemplation regarding the question of whether large-scale AI development necessitates governmental backing.

Let us consider the ground-level reality in light of the global precedent.

Related: Rise and Rise of Indian Defence Startups, Op Sindoor Fuelled Ambition

In reality, Global Large Language Models such as GPT-4o, Gemini, Llama, Claude, and Mistral have been predominantly developed and invested in by private companies and startups, with or without venture capital funding.

Typically, the creation and implementation of these large language models have been driven by either a major tech corporation (e.g., Google) or a well-capitalized startup (e.g., OpenAI).  For instance, OpenAI is significantly supported by Microsoft, while Llama boasts a substantial open-source ecosystem and a foundational contribution from Meta.

The reality discovered above mirrors the position, bolstering the claim that developing resourceful LLMs is feasible even in the absence of government backing.

When a government sees a sovereign AI necessity to provide citizens at large with economical LLM resources, this could serve as a powerful suggestion for that government to back the infrastructure needed to create such sovereign LLMs, including GPU support.

Thus, what I actually support is a moderate position: the ecosystem should cultivate a healthy mix of sovereign and commercial models.

Nonetheless, I firmly believe that the comprehensive support required for a robust startup ecosystem focused on building LLMs goes beyond merely offering hardware assistance.  It encompasses a vast array of requirements (in addition to significant hardware infrastructure support like a GPU bank) that range from –

• The necessity of a robust AI research ecosystem
• The necessity of well-curated localised data repositories
• A robust AI governance mechanism to oversee, monitor, and enhance the models
• Lastly — A strong human resource ecosystem to produce manpower ranging from data labelers, data preprocessors, and data engineers to ML algorithm specialists, ML deployment engineers, AI policy specialists, and ultimately AI managers and chief AI officers.

We are confident that a robust AI research ecosystem is emerging in India, consisting of work from IITs, IIITs, NITs, and commercial research labs such as Google Research India, Microsoft Research India, and IBM Research India.  According to a study by the research agency Itihaasa, India is ranked third globally for high-quality AI research publications, but there is a considerable gap between it and the world leader, China.

Although it may be behind the USA and China, we are witnessing the emergence of a robustly developing AI research ecosystem encompassing both academic and industry research.

Concerning the data repositories, the government can make a substantial contribution through its systematic initiative at AIKosh, which serves as a repository for Indian context data sets.

This work is greatly valued and will revolutionize the development of LLM for AI startups and other ecosystem participants. Therefore, making a major effort to increase the size, diversity, and accuracy of the datasets in AIKosh will greatly aid in the simplification of Indian LLM initiatives.

In terms of new initiatives for AI governance, the IndiaAI mission under MietY (Government of India) will shortly introduce a framework tailored to India that will prioritize responsible and safe AI without limiting innovation.

Using the EU AI Act and UNESCO’s current frameworks as a model, several MietY round tables have been held with stakeholders from academia, government, and entrepreneurs to design the same.

Regarding skills, I believe India has the potential to become the global skill powerhouse.  This encompasses outputs from extensive AI/ML degree programs in engineering and management institutes, AI/ML certification courses for professional continuing education, and specialized CXO awareness initiatives on AI/ML.  These initiatives contribute to India’s emergence as the global hub for AI skills.

In addition to the infrastructure aspect, there is a clear consensus that government support can significantly accelerate AI and LLM innovations.

To foster the development of LLMs in India, it will be essential to provide multi-faceted support across all four areas mentioned—data repositories, AI governance, AI research, and AI skill development—as well as ensuring that infrastructure is available.

Given the widespread presence of LLMs in our daily lives and the necessity to tailor them for India-specific data and use cases, these LLMs are an essential requirement at this moment.