Nathan Lambert · @natolambert · X·2026-09-09 01:39·1小时前
AI 导读

Nathan Lambert 转发 OpenAI 的演示并评论称,希望看到关于智能体数量的 scaling laws。OpenAI 称其内部模型组用约 10,000 个协作 AI Agent 在 88 小时内得出 Navier–Stokes 解,并全程保留了监控与隔离等前沿评估安全防护。他认为智能体集群数量的扩展是否与单模型 CoT 推理时计算的扩展同等关键,关系到许多关于 AI 未来的讨论。

Nathan Lambert@natolambert
53AI 编辑部评分,满分 100
2026-09-09 01:39· 1小时前
AI 导读

Nathan Lambert 转发 OpenAI 的演示并评论称,希望看到关于智能体数量的 scaling laws。OpenAI 称其内部模型组用约 10,000 个协作 AI Agent 在 88 小时内得出 Navier–Stokes 解,并全程保留了监控与隔离等前沿评估安全防护。他认为智能体集群数量的扩展是否与单模型 CoT 推理时计算的扩展同等关键,关系到许多关于 AI 未来的讨论。

Would love to see scaling laws on the number of agents. Seems very important to a lot of discussions about the future of AI -- if scaling the number of agents in a swarm is as crucial as something like length the original inference time scaling of CoT in solo reasoning models.

OpenAIThis model represents a step-function improvement on many benchmarks, and its training is ongoing. Our internal model group arrived at the Navier–Stokes solutio...