Anthropic CEO Maps Plan to Slow AI Development, Commits to Independent Oversight
Anthropic CEO Dario Amodei calls for slowing the pace of AI development in a new blog post, outlining three broad strategies for "pacing the frontier." He says two developments convince him a more cautious approach is needed: the OpenAI-HuggingFace hack and the fact that AI capabilities are advancing drastically faster, particularly in their ability to build the next generation of AI. Other tech leaders respond positively, with OpenAI CEO Sam Altman agreeing the industry needs to pace the frontier and Elon Musk posting that "Dario is right."
The post arrives amid an intensifying debate over AI safety. Researcher Jacob Coxon resigns from Anthropic this week, writing that leading AI companies are "gambling with our lives" and that people building the technology earnestly believe it could kill us all by the end of the decade. While Amodei doesn't mention Coxon's resignation directly, he writes that "we must slow the pace at which we improve the capabilities of AI models," adding that progress will still seem fast and that the industry must make wise use of the time it gains.
Amodei's first proposed step involves "embedded evaluators" from third-party organizations like METR, who would verify that AI companies follow their pacing and safety commitments and ensure safety incidents get reported. He compares the evaluators to regulators embedded with bank employees, and says Anthropic is unilaterally committing to hosting them, with badges, desks, laptops, and access mostly comparable to internal risk assessment teams. He also calls on governments to require other frontier companies to match this commitment.