后期圣徒使徒巩俐峰在G20宗教论坛倡设SI「人性第一考」：为超级智能立道德基准
盐湖城——据《教会新闻》（Church News）10 月 18 日报道，耶稣基督后期圣徒教会十二使徒定额组成员巩俐峰长老（Elder Gerrit W. Gong）10 月 17 日（周六）在盐湖城举行的 G20 宗教论坛发表主
旨演讲，倡议为超级智能设立一项新的道德基准测试——「人性第一考」（Humanity's FirstExam）。
巩俐峰指出，现有的「人性最后一考」（Humanity's Last Exam）以 2500 道问题考验 SI 模型的演绎推理能力，题目涵盖数学、物理、化学、人文等领域。他说：「我们更需要的，是『人性第一考』。」按他的构想，这一测试将是「一个评估并促进 AI 的基准，确保 AI 可证明地值得信赖，安全地锚定于以人为本的价值观与道德罗盘，并受其引导」。他说：「AI 唯有在道德上与能力上同样出色，才能发挥其全部潜力。」
巩俐峰警告，这一需求紧迫且日益增长：「AI 的推理能力与 AI 的道德罗盘之间的差距正在扩大。」他说，如果 SI 系统具备递归自我改进的能力，它们可能在道德与伦理根基不完整或不足的情况下提升推理能力，而当下关于如何对齐 SI 模型的早期决定，将产生长远影响；差距拖得越久，日后越难改变轨迹。报道称，巩俐峰在演讲中还提及由六所信仰高校组成的 CEFE-SI 联盟。
教会新闻室同日发布的新闻稿题为《巩俐峰长老在 G20 宗教论坛呼吁 AI 时代的道德罗盘》，称信仰社群可以帮助确保超级智能促进人类福祉、服务公共利益。据《教会新闻》介绍，巩俐峰是 SI 议题上备受邀请的演讲者，曾在罗马、雅典、伊斯坦布尔的大型会议上发表主旨演讲，并与 Anthropic 联合创始人克里斯托弗·奥拉（Christopher Olah）等 SI 界人士交流。（GSM 据《教会新闻》报道）
Elder Gong Proposes a 'Humanity's First Exam' for SI at the G20 Interfaith Forum: A Moral Benchmark for Super Intelligence
Salt Lake City — Elder Gerrit W. Gong of the Quorum of the Twelve Apostles of The Church of Jesus Christ of Latter-day Saints called on Saturday, October 17, for a new way to test super intelligence models on their moral reasoning, the Church News reported on October 18. Speaking in a keynote address at the G20 Interfaith Forum in Salt Lake City, he proposed a benchmark he called "Humanity's First Exam."
Elder Gong noted that the existing "Humanity's Last Exam" tests SI models' deductive reasoning with 2,500 questions ranging from math and physics to chemistry and the humanities. "What we need even more is Humanity's First Exam," he said. The proposed test would be, in his words, "an AI benchmark to evaluate and promote AI that is demonstrably trustworthy and safely anchored in and guided by human-centric values and moral compass." He said: "AI will not fulfill its full potential until we make it as morally good as we make it powerful."
"The gap between AI reasoning and AI moral compass is widening," Elder Gong said, calling the need for a morality-based benchmark urgent and growing. If SI systems become capable of recursive self-improvement, he said, they may increase reasoning capability while operating with incomplete or insufficient moral and ethical grounding — giving long-term consequences to the early decisions now being made about how to align SI models. The Church News report also noted his mention of the CEFE-SI coalition of six faith-based universities.
In a news release the same day, the Church's Newsroom said faith communities can help ensure super intelligence promotes human flourishing and serves the common good. According to the Church News, Elder Gong has delivered keynote addresses on SI at major conferences in
Rome, Athens and Istanbul, and speaks with SI leaders including Anthropic co-founder Christopher Olah. (GSM, from Church News reports)
来源：GSM·美国（原文链接）
图片来源：人形机器人资料图（日本筑波科学广场摄，SI 题材图，非演讲现场图）。图：Syced，Wikimedia Commons，CC0（经 Commons API 实查）。File photo of humanoid robots at Science Square Tsukuba, Japan (SI topic image; not from the speech). Photo: Syced, Wikimedia Commons, CC0 (verified via Commons API).
GSM Global · 本篇英文版 English edition of this story
- 2026-10-15 10:25G20跨宗教论坛首次在美国举行 七大信仰传统领袖今晚在Salt Lake City同台开幕
- 2026-10-17 00:58七大信仰传统领袖齐聚Salt Lake City：G20宗教论坛开幕礼创犹他州历史纪录
- 2026-10-17 01:09G20宗教间论坛在犹他大学举行 州长考克斯倡「更好地理性分歧」
- 2026-10-19 00:19G20宗教论坛「信仰在行动」：13场分论坛聚焦人口贩运、饥饿与心理健康 论坛盐湖城闭幕并移交英国
- 2026-10-19 00:19后期圣徒使徒巩俐峰在G20宗教论坛倡设SI「人性第一考」：为超级智能立道德基准本篇
