Google DeepMind launches institute to widen the AGI debate

Google DeepMind 成立新研究所，旨在拓宽 AGI 讨论范围

Google and Google DeepMind researchers launched the DeepMind Institute on Wednesday to advance the conversation around artificial general intelligence. The institute lists DeepMind co-founder Shane Legg, Google executive James Manyika, and Google DeepMind chair Demis Hassabis as directors, with Legg serving as managing editor. Google 和 Google DeepMind 的研究人员于周三成立了 DeepMind 研究所（DeepMind Institute），旨在推动关于通用人工智能（AGI）的讨论。该研究所由 DeepMind 联合创始人 Shane Legg、Google 高管 James Manyika 和 Google DeepMind 主席 Demis Hassabis 担任董事，Legg 同时担任执行编辑。

The new institute aims to surface differing views between Google, Google DeepMind, and the broader global research community around AGI. “They will not always agree, and they will likely change their minds, as more data and information comes to light at the fast-moving frontier,” the announcement read. 该研究所旨在呈现 Google、Google DeepMind 以及全球更广泛的研究社区在 AGI 问题上的不同观点。公告中写道：“随着这一快速发展的前沿领域涌现出更多数据和信息，他们不会总是达成一致，也可能会改变自己的想法。”

The inaugural collection of four essays covers a range of topics: economic policies for managing potential AGI disruption, preserving human-readable model reasoning, principles for human flourishing, and a framework for evaluating frontier AI models. 首批发布的四篇论文涵盖了一系列主题：管理潜在 AGI 颠覆性影响的经济政策、保持人类可读的模型推理、人类繁荣的原则，以及评估前沿 AI 模型的框架。

One essay, by DeepMind safety researchers Rohin Shah and Anca Dragan, argues that AI’s shrinking window of transparency—the ability to see and check a model’s step-by-step reasoning—is not inevitable. As new architectures make the most powerful models harder to monitor, the authors say developers and regulators should confront the safety trade-offs directly. That could mean limiting “opaque serial depth”—the amount of sequential computation a model can perform without producing a readable reasoning trace—or requiring developers to demonstrate that less transparent systems remain just as monitorable. DeepMind 安全研究员 Rohin Shah 和 Anca Dragan 在其中一篇论文中指出，AI 透明度窗口的缩小（即查看和检查模型逐步推理过程的能力）并非不可避免。随着新架构使最强大的模型变得更难监控，作者认为开发者和监管机构应直接面对安全权衡问题。这可能意味着限制“不透明序列深度”（即模型在不产生可读推理轨迹的情况下所能执行的顺序计算量），或者要求开发者证明透明度较低的系统依然具备同等的可监控性。

In another essay, Hassabis proposes a U.S.-led frontier AI standards body to evaluate the most advanced AI models. Under his framework, developers would initially submit models voluntarily for review up to 30 days before release. Once the evaluation system has proved effective, passing its tests could become a requirement for deploying frontier models in the United States. The body would at first design assessments in consultation with AI companies, but would eventually develop independent, undisclosed evaluations—what the essay calls “held-out” tests—to prevent labs from tailoring their models to known evaluations. 在另一篇论文中，Hassabis 提议成立一个由美国主导的前沿 AI 标准机构，以评估最先进的 AI 模型。根据他的框架，开发者最初将在发布前 30 天内自愿提交模型进行审查。一旦评估系统被证明有效，通过其测试可能会成为在美国部署前沿模型的必要条件。该机构起初将与 AI 公司协商设计评估方案，但最终会开发独立的、不公开的评估（论文中称为“留存”测试），以防止实验室针对已知评估来定制模型。

Hassabis said the framework could be “ratcheted up if the seriousness of the situation demands,” potentially including a coordinated slowdown among frontier AI developers. The essays arrive as the industry’s safety debate shifts from broad statements of concern toward concrete proposals for disclosure, outside scrutiny and, if safeguards fall behind, coordinated slowdowns. That shift accelerated this week as industry leaders endorsed elements of Anthropic CEO Dario Amodei’s call to “pace” frontier AI development. Hassabis 表示，如果情况严重，该框架可以“逐步升级”，可能包括前沿 AI 开发者之间的协同放缓。这些论文发布之际，行业内的安全讨论正从广泛的担忧声明转向具体的披露、外部审查建议，以及在安全保障滞后时采取协同放缓措施。随着行业领袖支持 Anthropic 首席执行官 Dario Amodei 关于“控制”前沿 AI 开发节奏的呼吁，这种转变在本周进一步加速。