AI的社会后果并非既定：繁荣与失业可能并存。文章指出，关键在于AI主要是“替代劳动”还是“增强劳动”。在此前工业化浪潮中，机械化与电气化在短期内取代了大量岗位，却也创造了新的大量任务，随后劳动力需求回升并推动工资上升。历史经验显示，技术浪潮的净效应取决于自动化与新任务增生的比例，这一逻辑同样适用于AI。
David Autor等人的新论文使用基于人口普查的任务数据，区分“新工作”和“更多同类工作”，发现新任务主要由更年轻且受教育程度更高的人群承担，起初有更高工资溢价，但随着技能扩散，溢价会下降。新任务通常出现在需求最强地区，并能为现有劳动者带来更多专业化机会。Daron Acemoglu等人则警告：美国制度可能偏向自动化。AI使自动化成本下降，促使大型公司更倾向减少人工，且劳动税负高于资本税进一步强化了替代激励。
政策回应不是一刀切打压自动化，而是扩大AI创造新任务和新技能的能力。提高总生产率仍是前提，因此政策应并行推进两条路径：一是提升劳动者跨任务流动性与再培训能力，使人机协作更强；二是降低职业准入壁垒，因为从美发到医生等领域的执照规则限制了中高技能岗位扩展。同一背景中，文本还提到4.5×10^10美元（原文45 billion）与4.0×10^10美元（原文40 billion）级别交易，显示AI讨论发生在高流动性的全球资本市场环境中。AGI仍然尚远，而近期最可实现的增益仍是技能劳动者与AI协同。
The social consequences of AI are not predetermined: prosperity and unemployment may arrive together. The central question is whether AI mainly substitutes labor or augments it. In earlier industrial waves, mechanization and electrification displaced many jobs in the short run but also created many new tasks, after which labor demand recovered and wages rose. History suggests the net effect of a technology wave depends on the share of automation versus new-task creation, and the same logic applies to AI.
A new paper by David Autor and co-authors uses Census-based task data to distinguish “new work” from “more of the same,” finding that new tasks are disproportionately performed by younger and more educated workers, pay an initial wage premium, and then decline as skills diffuse. New tasks also cluster where demand is strongest and can create specialization gains for current workers. Daron Acemoglu and others warn that U.S. institutions may lean toward automation: AI lowers automation costs, leading major firms to seek less human labor, while higher labor taxes than capital taxes further reinforce replacement incentives.
The policy response is not to punish automation per se, but to expand AI’s capacity to create new tasks and new skills. Raising total productivity remains essential, so policy should proceed on two tracks: first, improve workers’ mobility across tasks and reskilling so humans collaborate better with AI; second, reduce occupational entry barriers because licensing rules in fields from hairdressing to medicine restrict middle- and high-skill job expansion. In the same context, the article also notes transactions of 4.5×10^10 USD (45 billion) and 4.0×10^10 USD (40 billion), indicating a global capital-market backdrop to the AI debate. AGI remains far off, while the nearer-term gains are in skilled labor combined with AI.