AI 导读
据 Bloomberg 报道，Sam Altman 向 OpenAI 员工表示，公司可能放缓前沿模型开发，并可能与其他 AI 实验室同步。OpenAI 此前已暂停面向部署的 RL 训练两周，其最大规模的前沿 RL 训练也一直保持暂停，以便测试更强的安全防护措施。OpenAI 还估算其监测系统会额外消耗约 20% 的 inference 算力，使先进模型的测试在计算上更加昂贵。
推荐理由
Bloomberg 报道给出 OpenAI 暂停 RL 训练与监测算力的具体数字，可供判断安全测试对前沿研发节奏的影响。
正文
Sam Altman told OpenAI staff the company may slow frontier-model development, potentially alongside other AI labs, Bloomberg reports.
OpenAI had already paused deployment-focused RL training for 2 weeks and kept its largest planned frontier RL run on hold while testing stronger safeguards.
OpenAI also estimated its monitoring system uses roughly 20% extra inference compute, making advanced-model testing more computationally expensive.
来源：@rohanpaul_ai · x.com