一项新的研究评估了社会可实施的十一项政策，以应对通用人工智能（AGI）可能带来的经济破坏1。该研究根据三种AGI经济场景——温和、中等破坏和结构性劳动力衰退——提出了分阶段的政策框架1。研究采用由51位真实经济学家设计的AI代理评估员对这些政策进行评分1。
在温和情景下，建议的干预措施包括扩展失业保险、劳动所得税抵免（EITC）和雇主主导的再培训1。中等破坏情景则建议将EITC转变为负所得税（NIT）1。如果出现结构性劳动力衰退的最坏情况，通用基本资本（UBC）被提议作为后备政策1。
研究框架从四个维度评估这些政策：福利与复原力、代理与民主参与、可行性与效率，以及跨AGI场景的耐久性1。现实中的政策参考包括丹麦失业保险体系，该体系向失业人员支付高达先前工资的90%，最高月额为3200美元，最长可持续两年1。阿拉斯加主权财富基金在2025年向合格居民支付1000美元年度股息，体现了资本分配政策的实践探索1。
公众态度调查显示，美国公众对工人再培训的支持率为85%，对失业保险的支持率为72%，对通用基本资本的支持率为54%1。
A research framework has assessed eleven economic policy proposals designed to mitigate potential labor market disruptions from artificial general intelligence 1. The evaluation employs AI agents modeled after fifty-one real economists to rate these interventions across three AGI economic scenarios: a mild case, moderate disruption, and structural labor decline 1.
The phased policy approach varies by severity of impact 1. Under mild scenarios, the framework recommends expanding unemployment insurance, expanding the Earned Income Tax Credit (EITC), and employer-led retraining programs 1. For moderate disruption, policymakers would transition EITC into a negative income tax (NIT) 1. In the event of structural labor force decline, universal basic capital (UBC) emerges as a backstop policy 1. The framework evaluates all proposals across four dimensions: welfare and resilience, agency and democratic participation, feasibility and efficiency, and durability across AGI scenarios 1.
Public support for these measures reflects varying confidence in different approaches 1. Eighty-five percent of Americans support worker retraining, while seventy-two percent support unemployment insurance expansion and fifty-four percent support universal basic capital 1. Real-world precedents provide data points for policy design: Denmark's unemployment insurance replaces up to ninety percent of prior wages, reaching a maximum of $3,200 monthly for up to two years 1. Alaska's Sovereign Wealth Fund distributed $1,000 annual dividends to qualified residents in 2025 1.
评论
还没有评论，欢迎留下第一条。