There was another reason why Ford was receptive to higher wages. As Magnus Alexander, an electrical engineer who had helped design the production systems at Westinghouse and General Electric, put it, “Productivity creates purchasing power.” And purchasing power was vital for mass production.

彼得·舍戈尔德对1906年匹兹堡的22个职业（主要分布在制造业和建筑业）进行了具体研究，结果发现，最高技能的工人（例如砖瓦匠）和技能不太高的工人（普通劳动者）之间的小时工资差异达到2∶1。他的研究结果进一步说明，笼统地认为19世纪初美国工人阶层都在为生计挣扎，这个看法过于简单了。

匹兹堡的体力劳动者生活在两个世界……金字塔顶端是那些高技能、高收入的工人，他们有能力购买自己的房子，在室内浴室洗澡。他们的家庭能够享受到“美国标准”……金字塔底部是城市庞大的劳动人口，他们挤进简陋的旧式公寓和迅速老化的木质住宅，使用室外公厕和水龙头，隔壁就是钢厂和熔炉的噪声、强光和恶劣的气味。他们的孩子在坑坑洼洼的街道和开放的下水道旁边玩打仗，他们的休闲就是睡觉。

到1914年，制造业名义平均工资增长了30%，从每小时17美分增加到每小时22美分，即每天2.04美元。我们不妨想象一下亨利·福特在1914年初宣布将高地公园工厂的基本工资提高到每天5美元产生的轰动效应。当然，他不可告人的动机是为了降低劳动周转率，再加上一点点的利他主义。劳动周转率是一个普遍的地方性问题，部分原因是制造业工厂依赖那些尚未结婚的移民，这些人四处迁移，只为获得更好的工资和工作环境。举例来说，在宾夕法尼亚州西部的一个矿山负责人声称，他一年仅需要1000名工人，但前前后后共雇用了5000名工人。制造业工厂的无技能工作仅需要受过很少培训或根本没有受过培训的工人，这使得移民工人一旦对工作不满意，就马上辞职，并搬到另一个城镇，换份工作。

福特把工资提高了一倍多，但这样的慷慨不是没有条件的，他把利他主义和父爱主义结合起来。他新成立的社会部雇用了200名代理，他们要访问13000名员工的家庭，建议他们如何形成适当的生活方式。因为福特要求员工已婚，调查员家访的目的不仅要观察清洁标准，也要确保员工已婚。在家访中，单身男性员工经常借用年轻未婚女性扮演妻子。父爱主义遍及生活的方方面面。福特有法务人员帮助员工购置房屋；一支由10名医生和100名护士组成的医疗队提供医疗服务，特别是为那些受伤工人；同时还为移民提供语言培训。

但是，对于移民工人的那些非美国习惯，工厂并不宽容，100人因请假庆祝东正教圣诞节而被解雇。改革家塔贝尔（Ida Tarbell）参观了社区，打算揭露压迫性的福特体制，但是他最后总结说，“不管你们叫它什么——慈善、父爱主义或者独裁统治，结果是你的所有付出都物有所值”。

First, we look exclusively at the short‐run before any of the other factors have adjusted: Observation 1) Machine Labor and Factor Earnings (in the short‐run): adding a marginal unit of machine labor reduces human wages but increases returns of complementary factors in a zero‐sum manner. Intuitively, what happens if we add one unit of machine labor is that first, that unit will earn its marginal return, but secondly, there is also a redistribution from labor to capital, which now becomes relatively scarcer. The gains of capital are exactly the losses of the existing stock of labor. The redistribution generated by technological progress can be thought of as a pecuniary externality, as we emphasized earlier. The income losses of wage earners and the income gains of other factors owners are inefficient compared to the first‐best benchmark considered in Section 2.1. In the given example, the owners of capital have obtained windfall gains but have not done anything to earn these higher return. A compensatory transfer from capital owners to workers simply undoes these windfall gains and leaves them equally well off as they were before. More generally, adding machine labor creates a redistribution away from human labor toward complementary factors. More generally, the result holds no matter what the complementary factor, for instance whether it is capital or land or unskilled versus skilled labor or entrepreneurial rents. Policy can undo these redistributions by taxing windfall gains while leaving the price system to work at the margin. The result also holds for decreasing‐returns‐toscale production functions if we interpret the profits earned by the owner of the technology as compensation for the implicit factor “entrepreneurship,” which takes part in the zero‐sum redistribution.

Let us remember that the automatic machine, whatever we think of any feelings it may have or may not have, is the precise economic equivalent of slave labor. Any labor which competes with slave labor must accept the economic co11ditions of slave labor. It is perfectly clear that this will produce an unemployment situation, in comparison with \vhich the present recession and even the depression of the thirties will seem a pleasant joke. This depression will ruin many industries -possibly even the industries which have taken advantage of the new potentialities. Ho\vever, there is nothing in the industrial tradition which forbids an industrialist to make a sure and quick profit, and to get out before the crash touches him personally.

Thus the new industrial revolution is a two-edged sword. It may be used for the benefit of humanity, assuming that humanity survh·es long enough to enter a period in which such a benefit is possible. If, however, \Ve proceed along the clear and obvious lines of our traditional behavior, and follow our traditional worship of progress and the fifth freedom ~the freedom to exploit- it is practically certain that we shall have to face a decade or more of ruin and despair.

We further study how technological choices interact with market power. When workers have market power, profit-maximizing firms pursue innovations that erode workers’ market power by making them more easily replaceable, even at the expense of production efficiency. A social planner who values worker welfare would instead employ technologies that preserve workers’ market power. Similarly, when employers have monopsony power, they choose technologies that expand this power beyond what a social planner would consider optimal.

“That’s why Meta’s Mark Zuckerberg and others have said that mid-level coders will be unnecessary soon, perhaps in this calendar year.

Zuckerberg, in January, told Joe Rogan: “Probably in 2025, we at Meta, as well as the other companies that are basically working on this, are going to have an AI that can effectively be a sort of mid-level engineer that you have at your company that can write code.” He said this will eventually reduce the need for humans to do this work. Shortly after, Meta announced plans to shrink its workforce by 5%.”

“Make no mistake: We’ve talked to scores of CEOs at companies of various sizes and across many industries. Every single one of them is working furiously to figure out when and how agents or other AI technology can displace human workers at scale. The second these technologies can operate at a human efficacy level, which could be six months to several years from now, companies will shift from humans to machines.”

This could wipe out tens of millions of jobs in a very short period of time. Yes, past technological transformations wiped away a lot of jobs but, over the long span, created many and more new ones.

This could hold true with AI, too. What’s different here is both the speed at which this AI transformation could hit, and the breadth of industries and individual jobs that will be profoundly affected.

You’re starting to see even big, profitable companies pull back:

Microsoft is laying off 6,000 workers (about 3% of the company), many of them engineers.

Walmart is cutting 1,500 corporate jobs as part of simplifying operations in anticipation of the big shift ahead.

CrowdStrike, a Texas-based cybersecurity company, slashed 500 jobs or 5% of its workforce, citing “a market and technology inflection point, with AI reshaping every industry.”

Aneesh Raman, chief economic opportunity officer at LinkedIn, warned in a New York Times op-ed (gift link) this month that AI is breaking “the bottom rungs of the career ladder — junior software developers … junior paralegals and first-year law-firm associates “who once cut their teeth on document review” … and young retail associates who are being supplanted by chatbots and other automated customer service tools.

Adding new tasks will expand productivity, and this effect could be in principle larger than the cost savings due to automation and task complementarities. In addition, new tasks will tend to increase wages. In particular, using the same steps as before, d ln w dN = 1 σ d ln Y dN + 1 σ γL(N )σ−1 ( ∫N I γL(z)σ−1dz ) +σ−1 σ B′(N ) B(N ) . (17) This is always positive and could be large. Note also that the wage and productivity impact of new tasks can be potentially larger than cost savings in existing tasks, and this is particularly likely to be the case when new tasks improve the entire production process (as captured by the B term), or when they add new sources of cost improvements or complementary functions. Despite new tasks’ central role in wage and productivity growth and in reducing labor income inequality (see Acemoglu and Restrepo, 2018, and Autor et al., 2024), I will not focus

AI is also predicted to widen the gap between capital and labor income.

“I finally note that consumer welfare in this economy is proportional to TFP, and in particular, GDP increases overstate improvements in consumer welfare, because investment comes at the expense of consumption”

““Cancer is cured, the economy grows at 10% a year, the budget is balanced — and 20% of people don’t have jobs.” That’s one very possible scenario rattling in his mind as AI power expands exponentially.”

This leads to the important question: How costly is redistribution in practice? As we noted earlier, markets don’t exist in a vacuum. They are structured by laws and regulations and how those laws and regulations are enforced. The outcome is the so‐called “market” distribution of income, which is then subject to taxes and transfer, leading to an after‐tax distribution of income. But this conventional distinction may not be quite accurate: the rules of the game concerning redistribution affect the market income distribution. See Piketty et al. (2014) and Stiglitz (2017). The points that we have denoted Eo and E1 describe the initial outcome and the outcome after‐technological change, assuming that laws, regulations, institutions, etc. remain unchanged. But, of course, it is not reasonable to

here is a growing consensus that one of the sources of the growth of inequality is the growth of rents, including the rents that innovators earn in excess of the cost of innovation (see e.g. Korinek and Ng, 2017). Taxing and redistributing such rents has an important role in ensuring that AI and other advances in technology are Pareto improving. Also, anti‐trust policies may lower such rents, ensuring that the benefits of innovations are more widely shared, as more competition lowers consumer prices from which all benefit. From the perspective of low skilled workers who losing from innovation, targeted expenditure programs financed by high rent taxes may be of greater benefit than the lowering of prices, the benefit of which will go disproportionately to those who otherwise have spending power.

Unemployment is one of the most problematic societal implications of technological progressnew technology often implies that old jobs are destroyed and workers need to find new jobs. Economists, of course, understand the “lump‐of‐labor fallacy” – the false notion that there is a fixed number of jobs, and that automating a given job means that there will forever be fewer jobs left in the economy. In a well‐functioning economy, we generally expect that technological progress creates additional income, which in turn can support more jobs. However, there are two sound economic reasons for why technological unemployment may arise: first, because wages do not adjust for some structural reason, as described e.g. by efficiency wage theories, and second as a transition phenomenon. The two phenomena may also interact in important ways, e.g. when efficiency wage considerations slow down the transition to a new equilibrium. We discuss the two categories in turn in the following subsections. The unemployment implications are especially problematic when technological progress is labor‐saving, which – by definition – requires that either wages have to fall or that other complementary factors like capital have to adjust enough for labor market equilibrium to be restored.

I argue that the advent of AGI could fundamentally alter the dynamics of our economic system, necessitating a comprehensive reassessment of economic theory and policy.

We need a dual approach to AI: steer technology in the short term while building new systems for the long term.

Our perspective is that technological progress does not happen by itself but is driven by human decisions on what, where, and how to innovate. It would be misplaced to view our fate as pre-determined by blind technological forces and market forces that are beyond our control, as some techno-fatalists suggest. We as a society have the power actively steer the path of technological progress in AI to address the challenges posed by our technological possibilities. Moreover, our material condition is shaped jointly by the technological innovations that we humans create and by the social and economic institutions that we collectively design and within which these innovations take place.

In the event that the adoption of this technology is both rapid and widespread—far from certain yet certainly possible—preparing the economy for a major national production shock is a first order policy concern. The 2024 Economic Report of the President dedicates an entire chapter to artificial intelligence (AI), stating: “The Federal Government’s role goes beyond ensuring that the gains brought about by AI are widely shared. It must also ensure that the costs to harmed individuals are addressed” (p. 273). While previous work has assessed which occupations are most likely to be exposed to LLMs (Brynjolfsson et al. 2018; Felten et al. 2018; Webb 2019; Eloundou et al. 2023), either as complements or substitutes in production (Acemoglu et al. 2022), historical experience with major shocks to the structure of the economy suggests that any initial employment impact is only the beginning of a process with many disparate downstream effects on geographic inequality (Connor et al. 2023), education, population (Berry & Glaeser 2005), and political behavior (Rodr ́ıguezPose et al. 2023).

日益加剧的不平等、步履蹒跚的教育体系、人口阻力以及可能性极大的财政转向，这四大阻力的综合影响意味着未来实际可支配收入中位数的增长速度比过去要缓慢得多。如果再考虑到1970年以来创新对生产率的较弱影响，经济增长将几乎没有任何空间。考虑到所有这些阻力，实际人均可支配收入中位数的未来增长很难是正的，并将远远低于19世纪以来几代美国人可支配收入中位数的增长率。

不平等、教育、人口和政府债务这四大阻力的共同影响大致可以量化。但更难以评估的是美国社会的崩溃迹象。无论是以单亲家庭抚养的孩子比例，还是以低收入家庭学龄前儿童词汇缺陷的百分比，或者以白人和黑人年轻人在监狱服刑时间的百分比来衡量，21世纪初美国社会衰败的迹象比比皆是。