摘要：Young workers are getting hit in fields where generative-AI tools such as ChatGPT can most easily automate tasks done by humans, s

AI Begins to Disrupt Job Hunters' Prospects

AI开始冲击求职者前景

Younger workers see rising competition in fields like software coding, translators

By Justin Lahart

Artificial intelligence is profoundly limiting some young Americans’ employment prospects, new research shows.

最新研究显示，人工智能正严重限制部分美国年轻人的就业前景。

Young workers are getting hit in fields where generative-AI tools such as ChatGPT can most easily automate tasks done by humans, such as software development, according to a paper released Tuesday by three Stanford University economists. They crunched anonymized data on millions of employees at tens of thousands of companies, including detailed information on workers’ ages and jobs, making this one of clearest indicators yet of AI’s disruptive impact.

斯坦福大学三位经济学家于周二发布的一篇论文指出，在ChatGPT等生成式AI工具最易实现人类任务自动化的领域（如软件开发），年轻从业者正遭受冲击。研究团队分析了数万家企业数百万员工的匿名数据，其中包含员工年龄与职业的详细信息，这一研究成为目前最能清晰体现人工智能颠覆性影响的证据之一。

“There’s a clear, evident change when you specifically look at young workers who are highly exposed to AI,” said Stanford economist Erik Brynjolfsson, who conducted the research with Bharat Chandar and Ruyu Chen.

“当你专门关注受人工智能高度影响的年轻从业者时，会发现一个显而易见的变化，”斯坦福大学经济学家埃里克・布林约尔松表示。该研究由他与巴拉特・钱达尔、陈如玉（音译）共同开展。

At the same time, the economists found evidence that in fields where AI can help people in their work, rather than replace them, employment among young people is improving.

与此同时，经济学家们还发现，在人工智能“辅助人类工作而非替代人类”的领域，年轻人的就业情况正逐步改善。

The work—which hasn’t been peer reviewed or published yet—helps answer a question that has been burning since OpenAI introduced ChatGPT in November 2022. Subsequent versions, and similar generative-AI tools from competitors such as Google-parent Alphabet, have only heightened worries the technology will make some jobs obsolete.

这项尚未经过同行评审和发表的研究，解答了自2022年11月OpenAI 推出ChatGPT以来一直备受关注的问题。此后，ChatGPT的更新版本以及谷歌母公司Alphabet等竞争对手推出的类似生成式AI工具，进一步加剧了人们对“该技术将导致部分工作岗位消失”的担忧。

While anecdotal evidence has emerged showing AI’s effects on certain professions, such as software coding, there has been little harder evidence that the technology was significantly weighing on the labor market.

尽管已有轶事证据显示AI对软件编程等特定职业产生影响，但几乎没有确凿证据表明该技术正对劳动力市场造成显著冲击。

One problem: ChatGPT rolled out during a period when the Federal Reserve was curbing economic growth by sharply raising interest rates, and job growth was moderating from the pandemic-related hiring surge. The new research helps tease out the AI impact from those other factors.

其中一个关键问题在于：ChatGPT 推出之际，美联储正通过大幅加息抑制经济增长，而就业增长也正从疫情期间的招聘热潮中放缓。这项新研究则有助于将AI的影响与其他因素区分开来。

Using records from paycheck processor ADP, the economists were able to get a granular view of how generative AI has affected the labor market. The data includes detailed information on workers’ ages and occupations, making it far more comprehensive than the survey of households the Labor Department uses for its monthly employment report.

通过分析薪酬处理公司ADP的记录，经济学家得以细致观察生成式AI对劳动力市场的影响。这些数据包含了员工的年龄和职业等详细信息，比美国劳工部月度就业报告所采用的家庭调查数据要全面得多。

The Stanford economists first looked at areas where AI can automate many of the tasks workers perform, and therefore potentially replace them. Those include jobs such as software developers, receptionists, translators and customor service representatives.

斯坦福大学的经济学家首先关注了“人工智能可自动化员工多项工作任务、进而可能替代人类”的领域，这类岗位包括软件开发者、前台接待员、翻译以及客服代表等。

Their finding: Overall employment in those categories has softened since late 2022 relative to other occupations, with the weakness concentrated among younger workers.

他们的研究发现：自2022年末以来，与其他职业相比，这些领域的整体就业情况有所疲软，且这种疲软主要集中在年轻从业者群体中。

“After late 2022 and early 2023 you start seeing that their employment has really gone in a different direction than other workers,” Brynjolfson said.

“2022年末至2023年初之后，你会发现年轻从业者的就业趋势与其他年龄段从业者开始明显分化，” 布林约尔松说。

Among software developers aged 22 to 25, for example, the head count was nearly 20% lower this July versus its late 2022 peak. These are daunting obstacles for the large number of students earning bachelor’s degrees in computer science in recent years.

例如，在22至25岁的软件开发人员中，今年7月的人数较2022年末的峰值下降了近20%。对于近年来大批获得计算机科学学士学位的学生而言，这无疑是巨大的就业障碍。

For workers aged 26 to 30, head count was close to flat, but among older workers, head count continued to grow.

26至30岁从业者的人数基本持平，而30岁以上从业者的人数则持续增长。

Other factors could be hitting those computer-science jobs, including a general slump in employment at technology companies or pandemic-related education disruptions. But the data suggest such possibilities can’t explain away the AI effect on other types of jobs. Head counts among customer service representatives—a category that, unlike software development, generally doesn’t require a college education—followed a similar pattern.

计算机科学相关岗位的疲软或许还受到其他因素影响，包括科技公司整体招聘放缓，或是疫情导致的教育中断。但数据显示，这些因素无法解释人工智能对其他类型岗位的影响。以客服代表为例——该岗位与软件开发不同，通常不需要大学学历——其就业人数变化也呈现出类似（年轻从业者减少）的趋势。

The economists were also able to rule out other factors that might skew the data, such as the interest-rate sensitivity of different businesses, or some occupations being more susceptible to remote work and outsourcing.

经济学家还排除了其他可能干扰数据的因素，例如不同企业对利率的敏感度不同，或是某些职业更易受远程办公和外包这两种趋势的影响。

Harder-to-automate skills older workers picked up during their careers might be insulating them from the same kind of AI hit. A senior software developer, for example, might have learned how to work collaboratively with non-coders and deliver the product the company needs. Such skills remain highly valued to employers and may never be automated away.

年长从业者在职业生涯中积累的“难被自动化”的技能，可能使他们免受AI的冲击。例如，一位资深软件开发者可能掌握了与非技术人员协作、并交付公司所需产品的能力。这类技能仍深受雇主重视，且可能永远无法被自动化替代。

This raises a potential labor-market paradox: If the only way to develop that knowledge is to put in time doing work that AI has largely automated away, who will replace today’s experts when they retire? Addressing that might require rethinking how young workers learn on the job.

这引发了一个潜在的劳动力市场悖论：若积累这类专业知识的唯一途径，是投入时间从事那些已被人工智能大幅自动化的工作，那么当如今的专家退休后，谁来接替他们？要解决这一问题，或许需要重新思考年轻从业者的在职学习方式。

“I think that we’ll have to more explicitly train people, as opposed to just hoping that they will figure these things out on their own,” said Brynjolfsson.

“我认为我们必须更有针对性地培训人才，而不是单纯期望他们能自行摸索掌握这些技能，”布林约尔松说。

While much of the focus on generative AI and the labor market has been on which jobs the technology might replace, AI could also help some workers do their jobs better. AI might help medical professionals make accurate diagnoses more quickly, for example.

尽管目前人们对生成式AI与劳动力市场的关注，多集中在 “该技术可能替代哪些岗位”上，但人工智能也能帮助部分从业者提升工作效率。例如，人工智能可能帮助医疗专业人员更快地做出准确诊断。

Young workers in occupations where researchers have found AI could act as a helper, rather than a replacement, actually saw employment growth that exceeded overall employment, the economists found. That wasn’t enough to offset the weakness in occupations where AI automates tasks workers do, but it does raise the hope that AI could be harnessed to augment workers’ skills, ultimately helping them do their jobs better and making them better off as a result.

经济学家发现，在那些被研究者认定人工智能可充当辅助工具而非替代人力的职业中，年轻员工的就业增长率实际上超过了整体水平。虽然这一增长尚不足以完全抵消人工智能自动化岗位的就业颓势，但确实带来了新的希望：人工智能有望被用于提升从业者的技能，最终帮助他们更好地完成工作，并改善其就业状况。

Simply automating tasks that people do can save money, but it doesn’t really create anything new, said Brynjolfsson. What’s more valuable is doing new things that extend people’s capabilities, leading to gains that encourage businesses to hire more, rather than fewer, people.

布林约尔松表示，单纯实现人类任务的自动化虽然可以节省成本，但并不能真正创造新价值。真正可贵的是通过人工智能拓展人类能力边界，开创全新可能——由此带来的收益增长将促使企业增加而非缩减用人规模。

“I was delighted to see in the data that indeed, this augmentation approach could benefit people and lead to more employment,” he said.

“我很高兴能从数据中看到，这种‘人工智能辅助人类’的模式确实能让人们受益，并带来更多就业机会，” 他说。

来源：AI观察室