Home / Technology / AI Leaders Call for Pacing Strategy
AI Leaders Call for Pacing Strategy
13 Sep
Summary
- AI firms urged to adopt a coordinated pacing strategy.
- Plan includes third-party evaluators and safety standards.
- US and global governments may mediate AI development deals.
Anthropic CEO Dario Amodei has proposed a "coordinated pacing strategy" for major US AI companies, urging them to slow the development of frontier models. This initiative, shared publicly this week, aims to dedicate more resources to AI alignment, testing, and interpretability. Amodei's plan includes three stages: embedding third-party evaluators with employee-like access for monitoring safety practices, establishing voluntary safety standards among top AI firms with US government involvement to address antitrust concerns, and forging global agreements to limit dangerous AI applications.
Amodei emphasized that regulation targeting all US frontier AI companies would be the most effective method for pacing development. This call has garnered support from other prominent AI leaders. OpenAI CEO Sam Altman agreed with the need to pace development and committed to implementing independent evaluators. Elon Musk and Google DeepMind's Demis Hassabis also voiced their support for a more deliberate approach to AI advancement, with Hassabis noting DeepMind's recent proposal for an industry-wide standards body.
However, not all figures in the AI space share this view. Former Stability AI CEO Emad Mostaque expressed skepticism regarding the feasibility and impartiality of proposed neutral evaluators. Meanwhile, Senator Bernie Sanders has criticized the current proposals, stating they are "a start, but it's not enough." Sanders advocates for a complete pause on advanced AI development and a ban on superintelligent AI, urging a treaty negotiation between the US and China to achieve this.
