Can Socially Minded Governance Control the Artificial General Intelligence Beast?
Management Science · 2025 · Joshua S. Gans
中文摘要
本文稳健地得出结论: 不能。本文在理想化条件下构建了一个模型, 这些条件假定与通用人工智能 (AGI) 相关的风险是真实的、安全的 AGI 产品是可能实现的, 并且存在有社会责任感、愿意资助安全 AGI 的资助者, 即使这样做并不能实现利润最大化。本文证明, 由这类资助者组建的有社会责任感的实体无法将营利性企业发布的不受限制的产品可能造成的 AGI 危害降至最小。原因在于, 有社会责任感的实体只能在与营利性企业的事后 (ex post) 竞争中, 以令人望而却步的财务成本来最小化不受限制 AGI 产品的使用, 因而无法在事前 (ex ante) 阻止营利性企业开发 AGI。
Abstract
This paper robustly concludes that it cannot. A model is constructed under idealized conditions that presume that the risks associated with artificial general intelligence (AGI) are real, that safe AGI products are possible, and that there exist socially minded funders who are interested in funding safe AGI, even if this does not maximize profits. It is demonstrated that a socially minded entity formed by such funders would not be able to minimize harm from AGI that unrestricted products released by for-profit firms might create. The reason is that a socially minded entity can only minimize the use of unrestricted AGI products in ex post competition with for-profit firms at a prohibitive financial cost and so, does not preempt the AGI developed by for-profit firms ex ante.