JONAS NAHM2026年2月26日 Illustration by The New York TimesTesla’s factory in Shanghai produces far more cars per worker than its plant in California. The gap reflects something unsettling about China’s broader edge in manufacturing: It has figured out how to organize production around large-scale deployment of automation, robotics and artificial intelligence. The United States has not.特斯拉上海工厂的人均汽车产量远超其加利福尼亚州的工厂。这一差距反映出中国在制造业方面更广泛优势中令人不安的一点：中国已摸索出如何围绕自动化、机器人和人工智能的大规模部署来组织生产。而美国尚未做到这一点。Reindustrialization is one of the few economic goals that now commands bipartisan support. Successive administrations — first Joe Biden’s, now President Trump’s — have made rebuilding American manufacturing a priority. In Washington, the gap between American factories and global competitors is often explained as the product of unfair subsidies, distorted markets or other forms of cheating.再工业化如今是少数能够获得两党支持的经济目标之一。从拜登开始到现在的特朗普，几届政府都把重建美国制造业作为优先事项。在华盛顿，人们往往把美国工厂与全球竞争对手之间的差距解释为不公平补贴、市场扭曲或其他形式的作弊所导致的结果。Those factors matter, as does the power of China’s political structure to command fast change from the top down. But the central challenge for the United States is not that China bends the rules. Around the world, modern manufacturing no longer resembles the mid-20th-century factory floor. Robotics, automation and A.I. now make it possible to produce more with fewer human workers, though those who remain are more skilled and better paid. Unlike China, America has failed to reckon with this reality and organize manufacturing in ways that turn its own technological strengths into comparable gains.这些因素固然重要，但中国的政治体制自上而下迅速推动变革的能力同样关键。而美国面临的核心挑战也并非中国破坏规则。放眼全球，现代制造业已不再像20世纪中叶的工厂车间。如今，机器人技术、自动化和人工智能使得企业能够用更少的劳动力生产更多的产品，而留下的工人技术更好、收入也更高。与中国不同，美国未能正视这一现实，也没有建立起一种能够把自身技术优势转化为同等生产率提升的制造体系。Washington talks about A.I. as if it lives in research labs, venture capital portfolios and data centers. China treats it as factory work. Today, A.I. is embedded into China’s efforts to accelerate automation — guiding machines, scheduling work and detecting problems in real time. China has built more than 30,000 smart factories. More than half of all new industrial robots installed worldwide in 2024 went into Chinese factories. Research from Weijian Shan, an investor and economist based in Hong Kong, has found that, from sectors ranging from steel to shipbuilding, these factories now produce more per worker than comparable U.S. plants.华盛顿在谈论人工智能时，仿佛它只存在于研究实验室、风险投资项目和数据中心里。中国则将人工智能视为工厂里实实在在的生产工具。如今，人工智能已深度融入中国加速自动化的进程——用于引导机器运行、安排生产进度并实时发现问题。中国已建成3万多家智能工厂。2024年全球新安装的工业机器人中超过一半在中国的工厂里。香港投资人和经济学家单伟建的研究发现，从钢铁到造船等多个行业，这些工厂如今的人均产出已超过美国同类工厂。The shift is visible on the shop floor. By last year, the Chinese electric vehicle company Zeekr had over 800 robots in its factory in Ningbo. The company even experimented with putting humanoid robots to work on its factory floor lifting boxes, assembling components and performing quality checks. Rather than following fixed instructions, the robots use cameras, sensors and A.I. to respond to conditions on the line, much like driver-assistance systems that adjust to traffic. That flexibility can allow them to handle variation, work safely alongside human workers and absorb routine changes that would otherwise force production to stop. These are the kind of efficiency gains that can eventually increase productivity per worker and help ease shortages of skilled labor.这种转变在生产车间里清晰可见。截至去年，中国的电动车企业极氪在宁波的工厂部署了超过800台机器人。该公司甚至尝试使用人形机器人参与生产，比如搬运箱子、组装零部件和执行质检工作。这些机器人不再只是遵循固定指令，而是利用摄像头、传感器和人工智能对生产线上的状况做出响应，就像能根据路况做出调整的辅助驾驶系统一样。这种灵活性使它们能够处理变化，与人类工人安全协作，并适应那些原本会导致生产中断的常规变化。此类效率提升最终将提高人均生产力，并有助于缓解熟练劳动力短缺问题。These gains are not limited to experimental systems. At Midea, one of the world’s largest home appliance manufacturers, an A.I.-driven control system coordinates robots, sensors and machines at its Jingzhou plant. A company official said the system has reduced response times from hours to seconds.这些成果并非仅限于实验性系统。在全球最大的家电制造商之一美的集团位于荆州的工厂里，一个由人工智能驱动的控制系统协调着机器人、传感器和各类机器设备。公司的一位管理者表示，该系统将响应时间从几小时缩短到了几秒。Productivity gains come from multiple forms of A.I. Software can analyze camera feeds so that defects can be removed from production. Scheduling algorithms can automatically balance production, inventory and logistics — Lenovo, for example, says it has used such systems to cut production scheduling times from hours to minutes. A.I. can also analyze streams of production data in real time and highlight small inefficiencies before they slow the entire line. The technology company Xiaomi says it used smart manufacturing and over 700 robots to produce a car every 76 seconds on average in its Beijing plant.生产率提升来自多种形式的人工智能。软件可以分析摄像头画面，从而把有缺陷的产品从生产过程中剔除。调度算法能够自动平衡生产、库存和物流——例如，联想集团表示，它利用此类系统把生产调度时间从数小时缩短到了几分钟。人工智能还可以实时分析源源不断的生产数据，在小问题拖慢整条生产线之前将其识别出来。科技公司小米表示，依靠智能制造和700多台机器人，其北京工厂平均每76秒就能生产一辆汽车。For a decade, Beijing has pursued factory modernization as a national project, driven by all levels of the government, beyond flagship factories like Zeekr and deep into China’s manufacturing supply chains.十年来，北京一直将工厂现代化作为国家工程来推进，各级政府参与其中，其范围不仅限于极氪这样的旗舰工厂，而是深入中国制造业的供应链之中。Provinces fund local companies developing A.I and automation technologies. Government ministries coordinate standards so that suppliers and manufacturers can share data and solutions. The government promotes programs to train workers to use these technologies. Smaller manufacturers can plug into shared, government-supported digital networks that collect production data, coordinate schedules and monitor equipment, allowing them to adopt A.I. tools without building from scratch.各省为开发人工智能和自动化技术的本地企业提供资金支持。国家部委协调技术标准的制定，以便供应商和制造商能够共享数据和解决方案。政府还推动培训项目，帮助工人掌握这些技术。规模较小的制造商可以接入由政府支持的共享数字网络，这些网络收集生产数据、协调生产调度并监控设备，使它们无需从零开始就能采用人工智能工具。On the other side of the Pacific, U.S. policy on A.I. emphasizes frontier research in A.I. and large language model development. While America leads in these fields, this focus neglects other practical areas for using A.I. and automation, leaving American manufacturers struggling to use digital tools on the factory floor. Only 18 percent of manufacturers surveyed by the Manufacturing Leadership Council said they have formal A.I. strategies for their operations; two-thirds said they were struggling to scale A.I. test projects into production.而在太平洋另一侧，美国的人工智能政策则强调前沿研究以及大语言模型的开发。虽然美国在这些领域处于领先地位，但这种侧重忽略了对人工智能和自动化的其他实际应用，导致美国制造商在工厂车间使用数字化工具时举步维艰。制造业领导力委员会的一项调查显示，只有18%的制造商表示他们为运营制定了正式的人工智能战略；三分之二的企业表示，他们难以把人工智能试验项目推广到实际生产中。Some American manufacturers are experimenting with similar tools. Auto manufacturers like Ford are experimenting with A.I.-enabled visual detection systems to identify defects on assembly lines. They also investing in systems that can detect equipment problems before they shut down production lines. Yet these efforts remain fragmented.一些美国制造商正在尝试类似的工具。福特等车企正在试验基于人工智能的视觉检测系统来识别装配线上的缺陷。他们也在投资那些能在设备故障导致生产线停工前发现问题的系统。然而，这些努力仍然零散，不成体系。The barriers are practical rather than technological, especially for small and midsize companies. In many factories, production data is incomplete or still recorded manually, making it impossible to use digital tools that rely on continuous information. Three-quarters of manufacturers surveyed struggle to connect older machines to systems that could help them run more efficiently. And eight in 10 report shortages of workers who can use A.I.-powered manufacturing tools. More than half say the upfront cost of A.I. projects is prohibitive.障碍更多来自实操层面，而非技术本身，尤其是对中小企业而言。在许多工厂，生产数据不完整或者仍依靠人工记录，使得依赖持续数据流的数字化工具难以投入应用。调查显示，四分之三的制造商难以将旧设备连接到能够提高效率的系统中；此外，八成企业表示缺乏能够操作基于人工智能的制造工具的工人。超过半数企业表示，人工智能项目的初期投入成本过高，难以承受。Yet the U.S. policy response to China’s rise in manufacturing targets trade flows rather than factory performance. A familiar mix of tariffs, trade investigations and restrictions on imported components and technologies looks tough but does little to close the productivity gap.然而，美国针对中国制造业崛起的政策回应主要瞄准的是贸易流动，而不是工厂本身的生产效率。关税、贸易调查以及对进口零部件和技术的限制组成的惯用手段，看似强硬，但对缩小生产率差距几乎没有帮助。These policies actually leave American manufacturers in a bind. When they attempt to automate, they often rely on imported robotics, sensors and machinery. Yet the Trump administration has opened a national security investigation into these robotics supply chains, potentially leading to tariffs that would raise the cost of such modernizing equipment.这些政策实际上让美国制造商陷入了两难境地。当他们尝试进行自动化改造时，往往依赖进口的机器人、传感器和机械设备。然而，特朗普政府已对这些机器人供应链启动了国家安全调查，这可能导致加征关税，从而提高此类现代化设备的成本。If the goal is to bring manufacturing back from overseas, American policymakers should focus less on protection and more on helping manufacturers deploy digital tools. That means connecting legacy equipment to digital systems so they can generate usable data. It means modernizing older plants so equipment can integrate with sensors, software and analytics. It means investing in work force training so employees can use A.I. tools. And it means federal and state governments supporting shared digital infrastructure that allows small and midsize manufacturers to adopt advanced tools.如果目标是真正把制造业从海外带回美国，那么美国的决策者就应该少强调保护主义，多帮助制造商部署数字化工具。这意味着要把传统设备接入数字系统，使其能够产生可用的数据；要改造老旧工厂，让设备能够与传感器、软件和数据分析系统整合。这意味着要投资劳动力培训，使员工能够掌握人工智能工具。这还意味着联邦和州政府需要支持共享的数字基础设施，使中小型制造商也能够采用先进工具。This is what China is doing. And other advanced economies in Germany, Japan, South Korea and elsewhere are responding with similar strategies.这正是中国正在做的事情。而德国、日本、韩国等其他发达经济体也正在以类似的战略作出回应。America has emphasized invention and breakthroughs over deployment. But sometimes, technology needs to be treated as factory work — unsexy and perhaps boring, but essential for being competitive.美国历来重视发明和突破，而轻视技术的应用。但有时，技术需要被当作工厂里的平常工作来对待——也许不光鲜，甚至有些枯燥，却是保持竞争力所必不可少的。Jonas Nahm是约翰·霍普金斯大学副教授，在拜登政府时期担任白宫经济顾问委员会产业战略高级经济学家。翻译：纽约时报中文网点击查看本文英文版。获取更多RSS：https://feedx.net https://feedx.site