Back to home

Compare

Comparing: AI Hasn’t Raised Overall Unemployment, but It’s Narrowing Entry-Level Jobs & AI 尚未拉低总体失业率,却在收窄初级岗位入口

AEN
OpenAIAghionStanford University·

AI Hasn’t Raised Overall Unemployment, but It’s Narrowing Entry-Level Jobs

U.S. Census Bureau data shows that 95.7% of businesses report no change in headcount after adopting AI. Yet Stanford research has also found a relative 13% decline in employment among workers aged 22 to 25 in occupations with high AI exposure. Our assessment is that AI’s clearest impact right now is not a collapse in total employment, but a narrowing of the door into a career.

What this is

The original article breaks down “Will AI replace people?” into three dimensions: aggregate employment, structural shifts, and distribution. At the aggregate level, the U.S. unemployment rate and total employment have not yet seen a cliff-like change. Structurally, AI is first reducing the entry points for junior roles that involve hiring, training, and promotion. In terms of distribution, labor’s share of U.S. income has fallen to 53.8%, meaning new output does not necessarily flow to workers in equal proportion.

Generative AI—artificial intelligence that can generate content such as text, code, and images—differs from earlier automation tools. Earlier waves primarily replaced physical labor and repetitive processes; this one is beginning to enter cognitive work such as writing and analysis. What is really changing may not be that jobs disappear overnight, but that companies hire fewer people while raising output expectations for the roles that remain.

Industry view

The main case for optimism comes from the macroeconomy. Nobel Prize in Economics laureate Philippe Aghion’s research based on French factory data found that for every one-percentage-point increase in automation, employment grows by roughly 0.4% over ten years. This suggests that technology diffusion can create new demand, provided consumers buy more as prices fall. Once demand approaches saturation, however, the historical pattern may no longer hold.

The risks are equally clear: reduced junior hiring may cut off the pipeline through which people enter an industry. Task automation may also weaken workers’ bargaining power, causing labor income to grow more slowly than economic output. Policy calls for “job-transition buffers” and a prudent, measured rollout also indicate that regulators are cushioning transition costs rather than assuming adjustment will happen naturally.

“Total employment hasn’t collapsed” can describe the macro outcome, but it cannot explain why entry points into a particular industry are narrowing. We favor a sector-by-sector, role-by-role assessment: Can the task be decomposed? Must responsibility still be handled by a human? Can demand for the output expand as costs fall?

Impact on regular people

For enterprise IT: Budget priorities will shift from deploying individual tools toward process redesign, data governance, and access control. When evaluating AI projects, teams should track new jobs created, hours saved, and income distribution alongside model capabilities.

For individual careers: The supply of executable work may increase, placing pressure on career entry points and intermediate roles. More valuable accumulation now comes from domain expertise, complex judgment, customer relationships, and the ability to take final responsibility.

For consumer markets: Lower delivery costs may produce cheaper or more plentiful services, but whether those savings translate into additional demand still depends on price elasticity and market capacity.

Source: juejin.cn
BZH
OpenAIAghion斯坦福大学·

AI 尚未拉低总体失业率,却在收窄初级岗位入口

美国普查局数据显示,95.7% 的企业称采用 AI 后用工人数没有变化;但斯坦福研究同时发现,22 至 25 岁劳动者在高 AI 暴露职业中的就业相对下降 13%。我们的判断是:AI 眼下更明显的影响不是让就业总量坍塌,而是收窄职业入口。

这是什么

原文把“AI 会不会取代人”拆成三个层面:总量、结构和分配。总量层面,美国失业率与就业总量暂未出现断崖式变化;结构层面,AI 先减少招聘、培养和晋升初级岗位的入口;分配层面,美国劳动收入份额降至 53.8%,意味着新增产出未必等比例流向劳动者。

生成式 AI(能生成文字、代码和图像等内容的人工智能)与以往自动化工具不同:过去多替代体力劳动和固定流程,这次开始进入写作、分析等认知工作。真正发生变化的,不一定是岗位当天消失,而是企业减少招人,并把原有岗位拆成更少的人手与更高的产出要求。

行业怎么看

乐观依据主要来自经济总量。诺贝尔经济学奖得主 Philippe Aghion 基于法国工厂数据的研究发现,自动化程度每提高 1 个百分点,十年后就业增长约 0.4%。这说明技术扩散可能催生新需求,但前提是消费者因价格下降而增加购买;一旦需求趋于饱和,历史规律就未必继续成立。

风险也很明确:企业减少初级招聘,可能切断人才进入行业的通道;任务自动化还可能削弱议价能力,使劳动收入增长慢于经济产出。政策层面的“岗位转型缓冲期”和稳慎推进要求,也说明监管者已为转型成本预留缓冲,而非默认调整会自然消化。

因此,“总就业没崩”可以描述宏观结果,却不能解释某个行业入口变窄的问题。我们更倾向于逐行业、逐岗位判断:任务是否可被拆解、责任是否仍需人工承担、产出需求能否随成本下降而扩大。

对普通人的影响

对企业 IT:预算重点会从部署单个工具转向流程改造、数据治理和权限控制。评估 AI 项目时,应同时记录新增岗位、节省工时与收入分配,而不只看模型能力。

对个人职场:可执行工作的供给可能增加,职业起点和中间环节承压。更有价值的积累来自领域经验、复杂判断、客户关系以及承担最终责任的能力。

对消费市场:更低的交付成本可能带来更便宜或更 plentiful 的服务,但节省的成本能否转化为新增需求,仍取决于价格弹性与市场容量。

Source: juejin.cn