The Great Inversion in Artificial Intelligence

AI leadership is no longer about who spends the most, but about who innovates smarter, scales faster and democratizes the technology for all citizens

- Amitabh Kant

The Paris AI Summit earlier this month inked an agreement on inclusive and sustainable Artificial Intelligence (AI). With 61 nations and organizations signing the agreement, the real success of this summit was its ability to align global efforts and reinforce AI's diversity, both in terms of geography and perspectives.

While the absence of the US and UK has been a talking point, the focus should be on the broader impact of global AI cooperation. Disagreements on governance are natural in such transformational discussions. However, each dialogue and declaration is moving the needle forward in shaping AI's future. A pivotal moment of the summit was when Prime Minister Narendra Modi outlined a bold and inclusive vision of AI and elaborated on how AI development is deeply interdependent across borders and governed collectively, and why governance must not solely focus on managing risks but should also promote innovation, ensure inclusivity and drive equitable access.

In this context, he advocated open and ethical AI. India's Digital Public Infrastructure (DPI) stands as a model for scalable, open and cost-effective AI deployment.

In the corridors of Silicon Valley's most powerful AI labs, a cardinal assumption—the scaling law, which dictates that an AI model's performance improves with more computing power, larger data sets, and increased parameters (more capital)—has just been upended. The breakthrough came with DeepSeek, a Chinese AI startup that developed and open-sourced a frontier model—Rl—for just $6 million—a stark contrast to the over $100 million reportedly spent on training OpenAI's GPT-4.

Sam Altman, OpenAI's CEO publicly reacted by stating that he believes they are on the wrong side of history in the open-source debate.