人类学家的首席执行官警告先进的AI 到2027年可能比人类更聪明,敦促全球规则防止伤害。 Anthropic's CEO warns advanced AI could outsmart humans by 2027, urging global rules to prevent harm.

人类学首席执行官达里奥·阿莫迪(Dario Amodeei)在一篇38页的论文中警告说,到2027年,先进的人工智能可能超过人类智能,导致在超过人类体制的数据中心中创建了“天才国家”。 In a 38-page essay, Anthropic CEO Dario Amodei warns that by 2027, advanced AI may surpass human intelligence, leading to the creation of a "country of geniuses" in a datacenter that surpasses human institutions. 他呼吁立即制定基于证据的条例,如透明度法、芯片出口管制和大赦国际的监测,列举诸如自治制度、专制政权的滥用、就业流离失所和获得生物武器等风险。 He calls for immediate, evidence-based regulations like transparency laws, chip export controls, and AI monitoring, citing risks like autonomous systems, abuse by authoritarian regimes, job displacement, and access to bioweapons. 他强调国际合作、道德界限和社会成熟度的必要性，以防止不可逆转的伤害，尽管他的公司仍在开发人工智能。 He emphasizes the necessity of international cooperation, moral boundaries, and societal maturity in order to prevent irreversible harm, even though his company is still developing AI.