Unlocking India's AI Potential: Rise of Localized AI Models

From Krutrim to OpenHathi: How Localized AI Models are Revolutionizing India's Tech Landscape and Driving Innovation Beyond Borders.

Dear Rise & Shine Community,

In the ever-evolving landscape of technology, the pendulum is swinging back toward local innovation.

The rise of artificial intelligence (AI) is ushering in a new era where nations are asserting their technological independence, moving away from the global dominance of Big Tech.

This shift is evident in the emergence of localized AI models that cater to the unique needs and languages of individual countries.

Deja Vu in the World of AI

Just as regulations once sought to limit Big Tech's expansion into critical sectors like finance and healthcare, the current focus has shifted to the realm of artificial intelligence.

Microsoft, Google, and Meta remain at the forefront, but now, local contenders are rising to the challenge worldwide.

In India, Ola founder Bhavish Aggarwal introduced 'Krutrim,' an AI platform supporting 10 Indian languages.

Earlier this week, Lightspeed and Peak XV-backed startup Sarvam AI announced its first large language model (LLM) based on Hindi. Sarvam’s LLM has been launched under the OpenHathi (elephant in Hindi) series.

Meanwhile, European players like Silo AI in Helsinki unveiled 'Poro,' an AI language model based on Finnish.

The Language of Independence

The recent Global Partnership on Artificial Intelligence (GPAI) Summit in New Delhi highlighted the importance of local language-based AI models.

Experts argue that models from global tech giants may struggle to capture the nuances embedded in Indian languages and cultures. This need for independence stems from concerns that current large AI models rely heavily on web-crawled data, leading to a convergence primarily in English.

To truly set Local Language Models (LLMs) apart, access to proprietary data, including national language libraries, is crucial.

Building Value, Protecting Culture

Minister of State for Electronics and Information Technology, Rajeev Chandrasekhar, emphasized the importance of "trusted" AI models with access to India's datasets. This aligns with the broader conversation about creating trusted datasets globally.

Speaking from Helsinki, Silo AI's CEO, Peter Sarlin, stressed the significance of building AI models within one's country to ensure that the value created remains within national borders.

As LLMs form the foundation for AI development, their ability to foster homegrown solutions is essential.

The European Perspective

In Europe, Rajan Anandan, Managing Director of Peak XV Partners, highlighted the necessity of open-source models to retain value within the continent. The concern is that relying on closed-source models from Big Tech could result in value creation flowing out of the country.

This emphasis on retaining value is not just a matter of economic interest but also a way to preserve cultural identity and foster innovation within regional boundaries.

Challenges on the Horizon

While non-English language AI models claim superiority in processing low-resource languages, their true test lies in real-world applications. As these models evolve, they face challenges in providing solutions for public services, particularly in healthcare, education, and financial services, reaching remote areas.

Additionally, access to computing power poses a hurdle for startups, as training these models requires substantial resources.

The contrast between Finland's supercomputer training Silo AI's 'Poro' and India's computational capabilities highlights the potential gap in this aspect.

In navigating the future of AI, the balance between global collaboration and local autonomy is delicate. The journey involves not only overcoming technological challenges but also addressing economic, cultural, and accessibility concerns.

As the world embraces the era of localized AI, it raises fundamental questions about the future of innovation, data governance, and the role of nations in shaping the digital landscape.

Thanks for being a part of the Rise & Shine community!

---

Thank you for reading our newsletter!💛

Join our community on LinkedIn.