A recent artificial intelligence conference in Beijing highlighted the need for cooperation between the US and China on AI safety. Despite their intense rivalry, experts agree that the risks associated with advanced AI, such as cybersecurity threats and systemic failures, outweigh national security concerns. Stephen Casper, a computer scientist at MIT, emphasized that AI is a global technology with global benefits and harms, and that international collaboration is necessary to address these risks. The conference featured discussions on the universality of cyber challenges raised by advanced AI, including vulnerabilities in AI-generated code and novel attack methods. Experts like Lin Yun, a professor at Shanghai Jia Tong University, stressed the importance of shared safety principles and technical standards to reduce systemic risk. The development of open-weight AI models has become a pressing issue, as these models can be used to identify security vulnerabilities and wielded as cyber weapons. To balance openness with risk, the industry must devise new ways to guarantee that open models are up-to-date, free of backdoors and vulnerabilities, and meet safety standards. The US and China must put aside their rivalry and work together to address the global risks associated with AI, or risk facing catastrophic consequences.

Source: Wired