智能体
www.aibase.com ·
哈佛心理学家平克：别渲染 AI 末日，真正该做的是安全工程
据 The Decoder 报道，哈佛大学心理学家 Steven Pinker 在 Quillette 发表致科技博主 Scott Alexander 的公开信，认为 AI 毁灭人类的风险被过度强调。他拒绝就 AI 生存风险进行公开辩论，称其只是'取悦观众的竞赛'，并反对'回形针最大化器'式末日场景，指出该假设混淆了'智能'与'攻击性'。Pinker 承认自己此前有两处误判：未料到大语言模型会如此强大，也未料到 AI 公司产品发布节奏如此之快。他表示真正担忧的是生物恐怖主义、网络攻击和无约束 AI Agent 三类风险，认为渲染末日无助于应对这些现实威胁，有效路径是落实审慎的安全工程、独立监督、问责机制与人类控制。
AI推荐理由 转述 The Decoder 对 Steven Pinker 在 Quillette 公开信的报道，核心新信息是 Pinker 承认自己此前两点误判（未料到大语言模型如此强大、未料到 AI 公司产品发布节奏如此之快），并明确区分'智能'与'攻击性'、反对'回形针最大化器'式末日叙事，主张把重点放在生物恐怖、网络攻击和无约束 AI Agent 三类现实风险及安全工程、独立监督、问责与人类控制上。属于观点性讨论而非技术或产品首发，事实依据为公开信与媒体转述，信息密度中等；标题中'哈佛心理学家平克'与正文'Pike'存在人名不一致，需以 Steven Pinker 为准。对 AI 安全治理议题有参考价值，但无新数据、无技术细节，故不属高优先级。
62
正文
Harvard University psychologist Pike: Don't spread rumors about the end of AI. What truly needs to be done is safety engineering
AIbase基地
Published inAI News · 3 min read · Sep 29, 2026
5
According to The Decoder, Harvard University psychologist Steven Pinker has publicly stated that the risk of AI destroying humanity is overly emphasized. In an open letter published in Quillette addressed to psychiatrist and technology blogger Scott Alexander, he explained his reasons.
The incident was triggered when Alexander issued a challenge to Pinker, calling for an open discussion on the survival risks posed by AI. Pinker declined, calling it merely a "competition to entertain the audience." He specifically disagreed with the apocalyptic scenario of a "paperclip maximizer," which assumes that AI would sacrifice all of humanity to produce more paperclips, pursuing a single goal so intensely that the system goes out of control. Pinker points out that this scenario conflates "intelligence" with "aggression."
The Greater Danger Lies in Reckless Announcements
Pinker has consistently held a positive view of AI in the past, but he admits he made two mistakes. First, he did not expect large language models to become so powerful. Second, he did not anticipate the extremely rapid pace at which AI companies have been shipping products. Ironicaly, many who are pushing forward this announcement competition belong to groups that have long warned about the risks.
There are also three other dangers that Pinker is genuinely concerned about: bio-terrorism, cyberattacks, and unrestricted AI agents. Pinker stated that emphasizing the end of the world does not help with these realistic threats. The effective approach is to actually implement careful security engineering, prioritize independent oversight, accountability systems, and human control.
On the Opposing Side, Extinction Is Estimated at 20%
AI New TermsPaperclip MaximizerHarvard UniversityLarge Language Model
This article is from AIbase Daily
Scan to view
Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you understand technical trends, and learning about innovative AI product applications.
—— Created by the AIbase Daily Team
© Copyright AIbase Base 2024, Click to View Source -