When Amazon’s chief executive, Andy Jassy, wrote last month that he expected the company’s use of artificial intelligence to “reduce our total work force” over the next few years, it confirmed the fear among many workers that A.I. would replace them. The fear was reinforced two weeks later when Microsoft said it was laying off about 9,000 people, roughly 4 percent of its work force.

亚马逊首席执行官安迪·雅西上个月写道，他预计公司未来几年使用人工智能将“减少我们的总劳动力”，这印证了许多员工对人工智能将取代他们的担忧。两周后，微软宣布裁员约9000人，约占其员工总数的4%，这种担忧进一步加剧。

That artificial intelligence is poised to displace white-collar workers is indisputable. But what kind of workers, exactly? Mr. Jassy’s announcement landed in the middle of a debate over just this question.

人工智能即将取代白领工作者是不争的事实。但究竟是哪类呢？雅西的此番表态出现在人们围绕这个问题展开辩论之际。

Some experts argue that A.I. is most likely to affect novice workers, whose tasks are generally simplest and therefore easiest to automate. Dario Amodei, the chief executive of the A.I. company Anthropic, recently told Axios that the technology could cannibalize half of all entry-level white-collar roles within five years. An uptick in the unemployment rate for recent college graduates has aggravated this concern, even if it doesn’t prove that A.I. is the cause of their job-market struggles.

一些专家认为，人工智能最有可能影响到新手员工，因为他们的工作通常是最简单的，因此最容易实现自动化。人工智能公司Anthropic的首席执行官达里奥·阿莫代伊最近告诉《Axios》，人工智能技术可能会在五年内取代半数初级白领职位。最近毕业的大学生（美国官方统计机构对于这个表述的定义是：年龄在22至27岁之间、拥有学士学位或更高学历的人——译注）失业率的上升令这种担忧加剧，即便这并不能证明人工智能是导致他们在就业市场上挣扎的原因。

But other captains of the A.I. industry have taken the opposite view, arguing that younger workers are likely to benefit from A.I. and that experienced workers will ultimately be more vulnerable. In an interview at a New York Times event in late June, Brad Lightcap, the chief operating officer of OpenAI, suggested that the technology could pose problems for “a class of worker that I think is more tenured, is more oriented toward a routine in a certain way of doing things.”

但人工智能行业的其他领导者却持相反的观点，他们认为年轻的员工可能会从人工智能中受益，而资深员工最终将更容易受到伤害。OpenAI首席运营官布拉德·莱特卡普6月底在《纽约时报》的一次活动上接受采访时表示，这项技术可能会给“我认为资历较深、习惯于按某种固定方式和流程工作的员工”带来麻烦。

The ultimate answer to this question will have vast implications. If entry-level jobs are most at risk, it could require a rethinking of how we educate college students, or even the value of college itself. And if older workers are most at risk, it could lead to economic and even political instability as large-scale layoffs become a persistent feature of the labor market.

这个问题的最终答案将产生深远的影响。如果初级岗位风险更大，可能需要重新思考大学的教育方式，甚至是大学本身的价值。如果年龄较大的员工面临的风险最大，随着大规模裁员成为劳动力市场的一个持续特征，这可能会导致经济甚至政治动荡。

David Furlonger, a vice president at the research firm Gartner who helps oversee its survey of chief executives, has considered the implications if A.I. displaces more experienced workers.

戴维·弗隆是研究公司高德纳的副总裁，负责该公司针对企业首席执行官的调查项目，他曾经思考过人工智能取代资深员工的后果。

“What are those people going to do? How will they be funded? What is the impact on tax revenue?” he said. “I imagine governments are thinking about that.”

“那些人今后怎么办？他们的生活来源在哪里？这会对税收产生什么影响？”他说。“我想各国政府正在考虑这个问题。”

Is A.I. Making Better Managers?

人工智能是否能造就更优秀的管理者？

Economists and other experts who study A.I. often draw different conclusions about whom it’s more likely to displace.

对于哪一方更有可能被人工智能取代，经济学家和其他研究人工智能的专家往往会得出不同的结论。

Zooming in on the fields that have deployed A.I. most widely thus far tends to paint a dire picture for entry-level workers. Data from ADP, the payroll processing firm, shows that in computer-related fields, employment for workers with less than two years of tenure peaked in 2023 and is down about 20 to 25 percent since then. There is a similar pattern among customer service representatives, who are increasingly reliant on A.I. as well.

深入研究迄今为止最广泛应用人工智能的领域，往往会给初级员工描绘出一幅可怕的画面。薪资处理公司ADP的数据显示，在计算机相关领域，工龄不足两年者的就业率在2023年达到顶峰，此后下降了约20%至25%。越来越依赖人工智能的客户服务代表当中也存在类似的模式。

Over the same period, employment in these industries has increased for workers with two or more years of job tenure, according to Ruyu Chen, a Stanford researcher who analyzed the data.

对这些数据进行了分析的斯坦福大学研究员陈汝宇（音）指出，在同一时期，在这些行业工作两年或两年以上者的就业率有所增加。

Other studies point in a similar direction, if in a roundabout way. In early 2023, Italy temporarily banned ChatGPT, which software developers there relied on to help them code. A team of researchers at the University of California, Irvine, and Chapman University compared the change in the productivity of Italian coders with the productivity of coders in France and Portugal, which did not ban the software, to isolate the impact of ChatGPT.

其他的研究也指出了类似的趋势，虽然是以迂回的方式。2023年初，意大利暂时禁止了ChatGPT，因为当地的软件开发人员依靠ChatGPT帮忙写代码。加州大学欧文分校和查普曼大学的一组研究人员将意大利程序员的生产力变化与没有禁止该软件的法国和葡萄牙的程序员的生产力变化进行了比较，以便研究ChatGPT的影响。

While the study did not look at job loss, it did find that the A.I. tool had transformed the jobs of midlevel workers in more favorable ways than the jobs of entry-level workers. According to the researchers, the junior coders used A.I. to complete their tasks somewhat faster; the experienced coders often used it to benefit their teams more broadly. For example, the A.I. helped midlevel coders review the work of other coders and suggest improvements, and to contribute to projects in languages they didn’t know.

虽然这项研究没有关注失业问题，但它确实发现，该人工智能工具对中级员工的工作改变比对初级员工更加有利。根据研究人员的说法，初级程序员使用人工智能只是为了更快完成工作；而经验丰富的程序员则往往用它来更广泛地造福团队。例如，人工智能帮助中级程序员审查其他程序员的工作并提出改进建议，还能让他们参与自己不熟悉的计算机语言项目。

“When people are really good at things, what they end up doing is helping other people as opposed to working on their own projects,” said Sarah Bana, one of the paper’s authors, adding that the A.I. essentially reinforced this tendency. Dr. Bana said the paper’s result suggested that A.I. would prompt companies to hire fewer junior coders (because fewer would be needed to complete entry-level tasks) but more midlevel coders (because A.I. amplified their value to their whole team).

该论文的作者之一萨拉·巴纳说：“当人们真的很擅长某件事时，他们最终做的是帮助别人，而不是做自己的项目。”她还说，人工智能本质上强化了这种倾向。巴纳说，这篇论文的结果表明，人工智能将促使公司雇佣更少的初级程序员（因为完成入门级任务所需的人数减少了），但会雇佣更多的中级程序员（因为人工智能放大了他们对整个团队的价值）。

On the other hand, Danielle Li, an economist at M.I.T. who studies the use of A.I. in the workplace, said there were scenarios in which A.I. could undermine higher-skilled workers more than entry-level workers. The reason is that it can, in effect, untether valuable skills from the humans who have traditionally possessed them. For instance, you may no longer have to be an engineer to code, or a lawyer to write a legal brief.

另一方面，麻省理工学院研究人工智能在工作场所应用的经济学家丹妮尔·李（音）表示，在某些情况下，人工智能对高技能员工的影响可能大于初级员工。原因是，它实际上可以从传统上拥有这些宝贵技能的人类手中夺走这些技能。例如，你可能不再需要一个工程师来写代码，或者不再需要一个律师来写法律摘要。

“That state of the world is not good for experienced workers,” she said. “You’re being paid for the rarity of your skill, and what happens is that A.I. allows the skill to live outside of people.”

“这种态势对资深从业者不利，”她说。“你因为技能的稀缺性而获得报酬，而人工智能使得技能脱离人体存在。”

Dr. Li said A.I. would not necessarily be good for less experienced workers, either. But she speculated that the uptick in unemployment for new college graduates resulted from employers’ expectations that they will need fewer workers overall in the age of A.I., not just fewer novice workers. An overall hiring slowdown can have a bigger impact on workers right out of college, since they don’t have a job to begin with.

李博士表示，人工智能对经验较少的员工来说也不一定是好事。但她推测，新近毕业的大学生当中失业率上升的原因是，雇主们认为，在人工智能时代，他们需要的劳动力总量将减少，而不仅仅是新手减少。整体招聘放缓可能对大学刚毕业者的冲击更大，因为他们一开始就没有工作。

Robert Plotkin, a partner in a small law firm specializing in intellectual property, said A.I. had not affected his firm’s need for lower-skilled workers like paralegals, who format the documents that his firm submits to the patent office. But his firm now uses roughly half as many contract lawyers, including some with several years of experience, as it used a few years ago, before the availability of generative A.I., he added.

罗伯特·普洛特金是一家专门从事知识产权业务的小型律师事务所的合伙人，他说，人工智能并没有影响他的事务所对律师助理等低技能员工的需求，这些员工负责按照要求整理提交给专利局的文件。不过他也表示，在生成式人工智能出现后，他的律所现在使用的合同律师人数，包括一些拥有多年经验的合同律师人数，大约是几年前的一半。

These more senior lawyers draft patent applications for clients, which Mr. Plotkin then reviews and asks them to revise. But he can often draft applications more efficiently with the help of an A.I. assistant, except when the patent involves a field of science or technology that he is unfamiliar with.

这些资历较深的律师为客户起草专利申请，然后普罗特金进行审核并提出修改意见。但在人工智能助手的帮助下，他通常可以更有效地起草申请，除非专利涉及他不熟悉的科学或技术领域。

“I’ve become very efficient at using A.I. as a tool to help me draft applications in a way that’s reduced our need for contract lawyers,” Mr. Plotkin said.

“我已经可以非常高效地使用人工智能作为工具，帮助我起草申请，这减少了我们对合同律师的需求，”普洛特金说。

Some of the companies at the cutting edge of A.I. adoption appear to have made similar calculations, laying off experienced employees rather than simply hiring fewer entry-level workers. Google, Meta and Amazon have all done layoffs since 2022. Two months before its most recent layoff announcement, Microsoft laid off 6,000 employees, many of them software developers, while the July layoffs included many middle managers.

一些处于人工智能应用前沿的公司似乎也做了类似的考量，这些公司解雇了有经验的员工，而不是简单地减少招聘初级员工。自2022年以来，谷歌、Meta和亚马逊都进行了裁员。在最近一次宣布裁员的两个月前，微软解雇了6000名员工，其中许多是软件开发人员，而7月份的裁员包括许多中层管理人员。

The Value of Inexperience

新手的价值

Gil Luria, an equity analyst who covers Microsoft for the investment bank D.A. Davidson, said one reason for layoffs was that companies like Microsoft and Google were cutting costs to prop up their profit margins as they invested billions in chips and data centers to develop A.I. But another reason is that software engineers are susceptible to replacement by A.I. at all skill levels — including experienced engineers who make a large salary but are reluctant to embrace the technology.

吉尔·卢里亚是投资银行D.A.戴维森研究微软的股票分析师，他说，裁员的一个原因是，像微软和谷歌这样的公司在投资数十亿美元开发人工智能芯片和数据中心的同时，也在削减成本以支撑利润率。但另一个原因是，软件工程师无论资历深浅都容易被人工智能取代——包括那些拿着高薪却不愿接受这项技术的资深工程师。

Microsoft “can do math quickly — see who’s adding value, who’s overpaid, who’s not overpaid, who’s adapting well,” Mr. Luria said. “There are senior people who have figured out how to get leverage out of A.I. and senior people who are insistent that A.I. can’t write code.”

微软“可以快速计算——看看谁在增加价值，谁的薪酬过高，谁的薪酬并不过高，谁适应得很好”，卢里亚说。“有些资深人士已经找到了如何利用人工智能的方法，也有些资深人士坚持认为人工智能不会写代码。”

Harper Reed, the chief executive of 2389 Research, which is building autonomous A.I. agents to help companies perform a variety of tasks, said the combination of higher salaries and a reluctance to embrace A.I. was likely to put the jobs of experienced coders at risk.

哈珀·里德是2389研究公司的首席执行官，该公司正在开发自主人工智能体，以帮助公司执行各种任务。他表示，更高的工资和不愿接受人工智能的结合可能会让经验丰富的程序员面临失业风险。 

“How you decrease cost is not by firing the cheapest employees you have,” Mr. Reed said. “You take the cheapest employee and make them worth the expensive employee.”

“降低成本的方法不是解雇最廉价的员工，”里德说。“而是使用最便宜的员工，让他们创造昂贵员工的价值。”