《Pacing model development in an era of cyber-critical capabilities》

OpenAI的Astra模型能力已经危险到需要暂停训练两周来提高安全措施的级别了。
“These safeguards require meaningful compute. Our current estimates put monitoring overhead at roughly 20% of the inference compute being monitored, though the cost varies substantially across training and evaluation workloads. ”
并非scaling law失效，反倒是前沿模型继续训练还需要一份额外的防御性监控算力。