末日论心态：从核武器规划者到人工智能研究人员

一位作者通过对埃尔斯伯格著作《末日机器》的分析，揭示了冷战时期兰德公司员工与当代人工智能研究人员之间的心理相似性。1埃尔斯伯格及其兰德公司同事在20多岁时都拒绝了公司慷慨的退休计划，1其中一位同事表示"我们从字面上相信我们在努力拯救世界"。1埃尔斯伯格本人则对苏联威胁持悲观态度，认为"我无法相信世界能长期避免核浩劫"。1

作者指出，这种基于末日信念而放弃长期财务规划的现象如今在人工智能领域再次出现。1据报道，一些人工智能研究人员采取吸烟行为来表示他们相信人工智能会在吸烟杀死他们之前治愈癌症。1作者进一步推测，Anthropic等人工智能实验室可能会取消员工退休福利，以此展示对通用人工智能即将到来的信念。1

An author has drawn a parallel between the mindset of Cold War nuclear weapons planners and today's artificial intelligence researchers, arguing both groups share a conviction that transformative—or catastrophic—change is imminent. Drawing on Daniel Ellsberg's account in "The Doomsday Machine," the analysis reveals that Ellsberg and his RAND Corporation colleagues in their twenties rejected the company's generous retirement plans because they believed the world was headed toward nuclear destruction.1 Ellsberg himself stated, "I cannot believe the world can long avoid nuclear holocaust," reflecting the era's pervasive anxiety about Soviet threats.1

The author identifies a striking parallel in the contemporary AI research community, where similar fatalism appears to shape professional choices. AI researchers have reportedly adopted behaviors—such as smoking—that reflect their belief that artificial intelligence will cure cancer before tobacco does, suggesting confidence in near-term breakthroughs that would render conventional health concerns moot.1 The author speculates that AI labs such as Anthropic might eventually eliminate employee retirement benefits as a tangible expression of their conviction that artificial general intelligence is arriving soon enough to make long-term financial planning obsolete.1 In both historical and present contexts, the underlying logic remains constant: if transformative change is on the horizon, preparing for a conventional future becomes irrational.