In US-China AI race, India's place will depend on capabilities built at home, Pg13

Key Highlights:

• The US and China held initial AI talks in September, agreeing on a dialogue regarding risks and benefits and a channel for incident notification.
• The two nations disagreed on terminology, with the US preferring "super intelligence" and China using "artificial intelligence".
• US President Trump views AI development as a race to be won, while Chinese President Xi Jinping emphasizes human control and cooperation.
• India faces concerns regarding G2 overlay in AI norms, a split universe of AI systems, and the appeal of China's open-weight models.
• India's response should focus on multilateral norms, human control over nuclear-use decisions, differentiated de-risking, and building domestic capabilities.

Detailed Insights:

• The US-China AI talks were initiated by US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng.
• The US proposed a channel to notify AI incidents with national-security consequences, aiming for a next meeting by November.
• China views the American safety debate as a form of containment, aiming to restrict its technological advancement.
• Beijing's primary AI risk concern is its "systemic" effect on "political security environment," ensuring Party control over labs.
• A missed opportunity was the lack of formalization of the November 2024 Lima understanding on human control over nuclear weapons decisions.
• Washington maintains its lead through chip controls and supply chain actions, while Beijing pushes for self-reliance and global spread of its open-weight models.
• India is concerned about the potential for discriminatory controls if AI risk reduction norms are set by the two leading powers.
• China is promoting its AI agenda through platforms like BRICS, proposing an open-source AI community within the grouping.
• India is unlikely to join China's proposed World Artificial Intelligence Cooperation Organisation (WAICO), viewing it as a unilateral, China-centric initiative similar to the Belt and Road Initiative.
• Chinese open-weight, open-source models like Alibaba's Qwen are attractive due to their competitive cost and adaptability, influencing countries like Singapore and Brazil.
• India advocates for multilateral norms on frontier AI and inclusion in incident-notification arrangements to reduce risk.
• India's strategy includes applying dual and differentiated de-risking to AI, keeping Chinese models out of critical government systems.
• It also emphasizes avoiding over-dependence on the American technology stack, citing past access cut-offs.
• Building domestic capability in compute, chips, models, datasets, and talent is crucial for India's AI sovereignty.
• India aims to shape BRICS AI initiatives to be consensus-based and technology-neutral, rather than ceding space to China's agenda.

Key Concepts Involved:

• Super intelligence: A hypothetical AI that surpasses human intelligence across all cognitive tasks.
• Export controls: Government regulations restricting the export of certain goods, technologies, or information to specific countries or entities.
• Open-source models: AI models whose source code, data, and documentation are publicly available for use, modification, and distribution.
• De-risking: A strategy to reduce exposure to geopolitical or economic risks, often by diversifying supply chains or limiting reliance on specific partners.