AI labs should dedicate a larger share of compute to safety research

From the articleAnthropic says Claude now leads a quarter of work building its next AI models | ReutersAlso in Anthropic details practical metrics to help monitor the speed of AI development - SiliconANGLE