我们的使命是确保通用人工智能——比人类更聪明的 AI 系统——惠及全人类。_更新于 2025 年 10 月 28 日：本文包含关于我们结构的过时信息。请参考以下页面获取最新信息。_ 我们的使命是确保通用人工智能——比人类更聪明的 AI 系统——惠及全人类。
Our mission is to ensure that artificial general intelligence—AI systems that are generally smarter than humans—benefits all of humanity. _Updated October 28, 2025: This post contains outdated information about our structure. Please refer to thefollowing pagefor updated information._ Our mission is to ensure that artificial general intelligence—AI systems that are generally smarter than humans—benefits all of humanity.
核心贡献 · Key contributions
提出逐步部署 AGI，使其与社会共同演化，以最小化突然的破坏。 Proposes gradual deployment of AGI to enable co-evolution with society and minimize sudden disruption.
倡导通过快速学习和谨慎迭代的紧密反馈循环来应对部署挑战。 Advocates for tight feedback loop of rapid learning and careful iteration to navigate deployment challenges.
强调对齐技术与能力同步发展，利用 AI 帮助评估复杂模型。 Emphasizes developing alignment techniques alongside capabilities, using AI to help evaluate complex models.
呼吁就 AGI 的治理、利益分配和访问权进行全球对话。 Calls for global conversation on governance, benefit distribution, and access to AGI.
通过非营利治理和利润上限来构建组织，使激励与安全和公共利益一致。 Structures organization with nonprofit governance and profit cap to align incentives with safety and public good.
强调需要独立审计、公共标准以及政府对大规模训练运行的监督。 Highlights need for independent audits, public standards, and government insight on large-scale training runs.
局限 · Limitations
假设渐进式过渡可行，但能力突然跃升可能绕过计划的安全措施。 Assumes gradual transition is feasible, but sudden capability jumps may bypass planned safeguards.
依赖 AI 来评估 AI，可能引入循环验证问题。 Relies on AI to help evaluate AI, which may introduce circular validation issues.
全球治理和公平分配仍属愿景，缺乏具体实施路径。 Global governance and fair distribution remain aspirational with no concrete implementation path.
利润上限和非营利结构可能无法完全防止竞争环境中的激励失调。 Profit cap and nonprofit structure may not fully prevent misaligned incentives in competitive landscape.
在关键时刻减速需要协调，但地缘政治竞争可能破坏这种协调。 Slowing down at critical junctures requires coordination that may be undermined by geopolitical competition.