AI安全公司Anthropic员工Jacob Coxon于2026年9月8日提交辞职信，在离职两个月前股权未成熟的情况下选择离开1。这份辞职信在一天内被浏览超过1亿次1，引发了对AI行业安全承诺真实性的广泛质疑。在辞职信中，Coxon声称AI行业同事们真诚相信AI可能在本十年内杀死所有人1。Anthropic研究员Evan Hubinger曾表示，人工智能在十年内杀死全人类的概率超过十分之一1。
这一事件将聚光灯投向了支撑AI安全论证的理性主义和有效利他主义社群。根据相关数据，LessWrong社区的人口构成中，男性占87%，约20%的成员有自闭症谱系诊断，近50%为抑郁症患者1。调查深入揭示了该社群内部存在的邪教特征和行为问题，其中Leverage Research组织曾发生心理操纵和性骚扰事件1，Zizian分裂教派与至少6起死亡事件有关，包括2022年8月的伪造自杀事件和2025年1月的谋杀案1。
这一社群与硅谷权力网络的关联进一步引发担忧。彼得·蒂尔向AI风险研究机构MIRI捐赠超过160万美元1，同时资助新反动派作家Curtis Yarvin的科技公司Tlon1。前加密货币企业家Sam Bankman-Fried通过FTX向Anthropic投入5亿美元，后因欺诈被判25年联邦监禁1。副总统JD Vance曾在蒂尔担任投资者的企业工作，蒂尔为其2022年参议院竞选投入1500万美元1。蒂尔在2025年秋进行了七小时的私人讲座，声称Eliezer Yudkowsky是"反基督的军团"1。此外，Open Philanthropy在2017年报告了约4000万美元与AI风险相关的拨款，其中3000万分配给OpenAI1。评估机构METR的员工与Anthropic等前沿实验室员工存在紧密的个人关系1，这引发了关于评估独立性的问题。
Jacob Coxon resigned from AI safety company Anthropic in September 2026, departing two months before his equity would vest.1 His resignation letter garnered over 100 million views in a single day.1 In his statement, Coxon disclosed that colleagues within the AI industry genuinely believe artificial intelligence could kill all humans within the current decade.1 The resignation has intensified scrutiny over the credibility of AI safety warnings and the networks promoting them.
The controversy extends beyond Coxon's departure to reveal a quarter-century history of the rationalist and effective altruism communities, which have cultivated deep ties to Silicon Valley AI laboratories and political figures, including Vice President JD Vance.1 Anthropic researcher Evan Hubinger has stated that the probability of AI killing all humanity within ten years exceeds one in ten.1 These claims originate from communities with documented internal problems: LessWrong, the rationalist hub, has a demographic composition of 87 percent male, approximately 20 percent with autism spectrum diagnoses, and nearly 50 percent reporting depression.1
The rationalist and effective altruism movements have been linked to troubling incidents and organizations. Leverage Research, an associated group, was the site of psychological manipulation and sexual harassment.1 The Zizian splinter sect has been connected to at least six deaths, including a staged suicide in August 2022 and a murder in January 2025.1 Philanthropist Peter Thiel donated over $1.6 million to MIRI, a rationalist-aligned AI safety organization, while simultaneously funding entrepreneur Curtis Yarvin's company Tlon.1 Sam Bankman-Fried channeled $500 million from FTX to Anthropic before being convicted of fraud and sentenced to 25 years in federal prison.1 Thiel employed Vance at one of his companies and invested $15 million in Vance's 2022 Senate campaign, then delivered a seven-hour private lecture in autumn 2025 characterizing AI researcher Eliezer Yudkowsky as leading a "legion of the antichrist."1
Funding for AI risk research has concentrated significantly within aligned institutions. Open Philanthropy reported approximately $40 million in AI-risk-related grants in 2017, with $30 million directed to OpenAI alone.1 Staff members at METR, an organization that evaluates AI systems, maintain close personal relationships with employees at frontier laboratories including Anthropic.1
评论
还没有评论，欢迎留下第一条。