高盛集团预测，未来十年人工智能可能取代美国约1000万个工作岗位，近五分之一的工人预计未来五年内其岗位将被AI消除。回溯2000年至2011年间，美国失去了近600万个工厂岗位（其中约100万个归因于中国进口竞争），但主要应对政策“贸易调整援助”（TAA）每年仅资助约4万名工人接受培训，远低于每年50万的制造业失业人数。如今，2023年联邦主要再培训项目仅支出1.7亿美元，覆盖约3.9万人，凸显出再培训体系的严重滞后与资金匮乏。

当前美国的职业培训体制面临严重的结构性失衡。全美约四成高中毕业生进入四年制大学，而每年服务800万人的社区大学职业项目仅获得约250亿美元的公共资金。此外，美国目前仅有约80万名注册学徒，不足人口仅为美国四分之一的德国的一半。研究表明，行业合作项目（Sectoral programmes）能将工人长期收入提高11%至40%，但此类项目在白领及非建造业领域的推广依然极为有限。

面对AI对白领及低学历女性岗位的剧烈冲击，各界正在探索跨领域合作机制。针对失业工人的短训计划预计到2034年每年仅能额外惠及10万人，平均补贴仅为2200美元。虽然Anthropic与OpenAI基金会分别承诺投入2.0亿和2.5亿美元用于资助劳动力转型项目（如RAISE US），但相比于OpenAI计划到2030年在算力上投入该金额3000倍的巨资，针对劳动力再培训的社会投资依然微乎其微。

Goldman Sachs estimates that artificial intelligence could displace approximately 10 million American jobs over the next decade, with nearly one in five workers anticipating their job will be eliminated within five years. Historical precedents offer little reassurance: between 2000 and 2011, America lost nearly 6 million manufacturing jobs—including 1 million lost to Chinese import competition—yet the primary safety net, Trade Adjustment Assistance (TAA), provided retraining to just 40,000 workers annually against 500,000 annual job losses. In 2023, the main federal displacement program spent a meager $170 million on only 39,000 individuals, demonstrating a systemic funding deficit.

America's current workforce education system remains poorly aligned with rapid labor displacement. While four in ten high school graduates enroll in four-year universities, community colleges serving around 8 million students annually receive just $25 billion in public workforce funding. Additionally, America registers only about 800,000 apprentices—barely half of Germany's total despite America having four times the population. Although randomized trials demonstrate that targeted sectoral training programs boost long-term worker earnings by 11% to 40%, such structured retraining models remain rare outside of traditional construction trades.

As AI automation threatens entry-level office workers and women without college degrees, new multi-sector initiatives like RAISE US are attempting to bridge the gap. Legislative tweaks expanding Pell grants to short career courses are projected by the Congressional Budget Office to reach just 100,000 additional recipients annually by 2034, with average grants of only $2,200. Although AI leaders Anthropic and OpenAI Foundation have committed $200 million and $250 million respectively toward workforce transition research, OpenAI's planned infrastructure spending through 2030 is 3,000 times larger, underscoring the massive disparity between AI investment and worker protection.

Source: Can America retrain workers before AI leaves them behind?

Subtitle: Success will require investing in employees as seriously as America invests in chips

Dateline: 8月 06, 2026 05:10 上午