Anthropic's recent blog post outlines three metrics designed to assist AI companies in monitoring their development processes. This announcement comes shortly after CEO Dario Amodei advocated for a slowdown in AI advancements, a call that has garnered support from notable industry figures like OpenAI's Sam Altman and Elon Musk.
The metrics include an assessment of AI-led research and development, oversight mechanisms for AI agents, and a snapshot of compute allocation within the company. Specifically, Anthropic found that its Claude models are not fully autonomous in their research tasks, and approximately 30,000 AI agents are engaged in research and engineering at any given time.
Additionally, the company reported that only about 6% of its compute resources for AI research were dedicated to safety measures. By sharing these metrics, Anthropic aims to foster transparency and provide a framework for third parties to evaluate the pace of AI development, which is crucial as the industry navigates the balance between innovation and safety