Global AI Cooperation Urged Amid Rising Cybersecurity Risks

A recent artificial intelligence conference in Beijing's high-tech district, Zhongguancun, brought together top experts in the field, including legends like Whitfield Diffie and Andrew Barto. The event highlighted the need for international cooperation on AI development, particularly between the US and China, to mitigate growing cybersecurity and systemic risks. According to Stephen Casper, a computer scientist at MIT, AI's global nature means its benefits and harms will be shared worldwide, and new capabilities will inevitably spread. The US has imposed restrictions on China's AI development, citing economic and national security concerns, but experts argue that cooperation is essential to prevent catastrophic failures or cyberattacks.

Casper compared the current situation to the US and Soviet Union's cooperation on nuclear dangers during the Cold War, saying that international collaboration on AI risks outweighs national security concerns. Lin Yun, a professor at Shanghai Jia Tong University, emphasized the importance of shared safety principles and technical standards, even in the face of competition. As AI models become more advanced and intertwined with daily life, the risks of cyberattacks and failures will grow, making cooperation between the world's AI superpowers increasingly necessary. The development of open-weight AI models, popular in both the US and China, poses a particular challenge, as they can be used to identify security vulnerabilities or wielded as cyber weapons.

Experts warn that the next generation of open-weight models may be as capable as the most powerful models, and the industry must devise new ways to guarantee their safety and security.

Source: Wired