China Doesn't Care About AI Safety, It Just Wants to Win
Apart from gaining a competitive edge, China won't be too interested in anything else. The upcoming US-China talks are unlikely to save humanity. China will likely downplay AI safety risks again.
TL;DR
- Many advocate for US-China cooperation on AI risks, viewing it as a last defense against potential disasters.
- A former US envoy expresses skepticism about the upcoming AI talks, believing China is primarily focused on gaining a competitive edge.
- During past AI dialogues, Chinese officials reportedly avoided discussing AI risks and instead accused the US of hindering technological opportunities.
- The author anticipates China will again downplay safety risks in upcoming talks, potentially using cooperation as leverage for concessions like eased chip restrictions.
- Despite recent acknowledgments of AI risks by China's leadership, domestic companies reportedly lag in safety performance compared to US models.
- The author warns that influential US figures pushing for immediate AI alliances could be exploited by China if diplomatic efforts fail to yield comprehensive security agreements.
- True progress in AI safety requires developers to improve standards, invest in safer AI research, and enhance transparency, with governments later translating these into policy.
- The best-case scenario for future talks might be non-binding commitments to pressure companies on cybersecurity and biosafety standards, and transparency in reporting incidents.