对齐研究中心 (ARC) 的评估团队已正式分拆，成立了一个名为 METR (Model Evaluation & Threat Research) 的新的独立非营利组织。METR 将继续致力于评估前沿人工智能系统，重点关注其自主能力和潜在威胁，包括人工智能的自我改进和规避监管。该组织由 Beth Barnes 领导，此前已与 OpenAI 和 Anthropic 等领先的人工智能实验室合作进行评估，并旨在开发严格的测试方法，以确保人工智能在广泛部署前的安全性。 AI
排序理由 人工智能评估团队分拆成立新的研究组织。
在 METR (Model Evaluation & Threat Research) 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
<p><em>Model Evaluation & Threat Research, pronounced ‘meter’</em></p> <p>As we <a href="https://evals.alignment.org/blog/2023-09-19-spin-out-announcement/">announced two months ago</a>, ARC Evals is wrapping up our incubation period at ARC (the <a href="https://www.alignment…
<p>ARC was founded as a small alignment theory organization run by Paul Christiano. In 2022 it hired Beth Barnes and incubated ARC Evals to do exploratory work on independent evaluations of cutting-edge AI models; Paul remained focused on <a href="https://www.alignment.org/theory…
<p>We believe that capable enough AI systems could pose very large risks to the world. We do not think today’s systems are capable enough to pose these sorts of risks, but we think that this situation could change quickly and it’s important to be monitoring the risks consistently…