Anthropic 提出 AI 进度三大度量指标:自我改进、监督能力与算力集中度

Google DeepMind launched a new institute to widen the public debate on AGI governance. The move comes as frontier labs disagree on safety pacing. Meanwhile, Anthropic published three metrics for measuring AI progress: self-improvement share, oversight effectiveness, and compute concentration. These frameworks give enterprises concrete indicators for risk assessment and alignment compliance.

发表评论