读完一早上的访谈.
主题综述

AI 劳动经济学:灭岗还是 Jevons 暴涨 · AI Labor Economics: Displacement vs. Jevons

主题综述

主题页(活文档)· 最近更新 2026-06-12 · 取材 5 篇访谈

更新日志

主流共识

他们共享一个出发点:别看被毁掉的岗位,要看被创造的;直觉系统性地高估前者、看不见后者。 Tabarrok 把这条经济学常识讲得最完整——连最聪明的人也想不出新工作:

"It is very easy to see the jobs which are destroyed with a new technology much, much harder. You need to be a Jules Verne to imagine the jobs which are created with new technology. … Ricardo, one of the greatest economists of his age, was worried about machinery taking jobs. … How many accountants did Excel put out of work? The answer basically to a rounding error is zero, is zero, a big fat zero. … the creation of Excel did not put accountants out of work. It actually further increased the demand for accountants."
「看到新技术带来的工作岗位被摧毁是非常简单的,但想看到新技术创造的新工作却难得多。需要你是儒勒·凡尔纳才能想象新技术创造的工作。……里卡多是他那个时代最伟大的经济学家之一,他对机器取代工作感到担忧。……Excel 让多少会计失业?答案基本上可以说是零,是零,是个大大的零。……Excel 的出现并没有让会计失业。它实际上进一步增加了对会计的需求。」
Alex Tabarrok · Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth

Andreessen 用同一框架但换了术语——"任务"才是原子单位,工作比任务活得久:

"A job is a bundle of tasks and everybody wants to talk about job loss but really what you want to look at is task loss, tasks changing. … The tasks are going to change. … The job persists longer than the individual tasks. As the tasks change enough, then that's when the jobs change."
「一份工作是一组任务的集合,每个人都想谈论失业,但实际上你想关注的是任务的丧失,任务的改变。……任务将会改变。……职位比个人任务存在的时间更长。当任务变化足够大时,职位就会发生变化。」
Marc Andreessen · Marc Andreessen: The real AI boom hasn't even started

Roslansky 在 LinkedIn 的数据上给出同样的"工作=任务集合"拆法,并且用现场数据反驳"AI 正在裁员"的标题:

"Everyone's job in the world … at the end of the day is a set of tasks that need to be done. … Now, if your job is just a set of tasks that can be automated, You need to start thinking about a new job."
「世界上每个人的工作……最终都是一系列需要完成的任务。……如果你的工作只是一系列可以自动化的任务,你需要开始考虑一份新的工作。」
Ryan Roslansky · LinkedIn's CEO On How AI Changes Your Job
"Not seeing the former, not seeing companies reduce their workforce because of AI, definitely seeing a ton of job descriptions that now include some form of AI in them. … we've also increased the number of customer service reps at LinkedIn because it turns out that the more that our customer service reps are dealing with those kind of meaty problems, we're seeing better retention of our customers, lower time resolution of those like thicker and more difficult problems and more success."
「没有看到前者,没有看到公司因为 AI 而裁员。肯定看到了大量的职位描述中现在都包含某种形式的 AI。……我们还增加了 LinkedIn 的客服代表数量,因为事实证明,我们的客服代表处理的此类实质性问题越多,客户的留存率就越高,解决那些更棘手问题的速度就越快,也更容易取得成功。」
Ryan Roslansky · LinkedIn's CEO On How AI Changes Your Job

第二条共识:白领专业服务先挨打,体力/协调/混乱的工作有韧性。 这一点连最乐观和最激进的人都同意。Cowen 点名了被打的阶层和有韧性的岗位:

"I think the big losers, and this does pain me, I'm a member of this class, is just the upper, upper middle class. The people who go to good schools and they think they can walk into careers in law, consulting, finance that are more or less automatic. … They may end up in Houston in an energy sector job earning 300K a year rather than 2 million a year."
「我认为最大的失落者,虽然这让我感到痛心,毕竟我是这个阶层的一员,就是上层中产阶级。那些上好学校的人认为他们可以轻松进入法律、咨询、金融等几乎是自动的职业。……他们可能会在休斯敦从事能源行业的工作,年收入 30 万美元,而不是 200 万美元。」
Tyler Cowen · Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth

Foody 从"哪些岗位的替代会很快"这端讲同一件事——可验证的、数字的先走,物理世界慢:

"I think there's going to be a lot more in the physical world. … it could be everything ranging from people that are creating robotics data to people that are waiters at restaurants or are just like therapists because people want human interaction, whatever that looks like. … automation In the physical world, it's going to happen a lot slower than what's happening in the digital world, just because of so many of the self-reinforcing gains and a lot of self-improvement that can happen in the virtual world, but not physical one."
「我认为在物理世界将会有更多工作机会。……这可能包括从创建机器人数据的人到餐厅服务员,或者仅仅是治疗师,因为人们需要人际互动。……我认为物理世界的自动化速度将比数字世界慢得多,这仅仅是因为虚拟世界中存在许多自我强化的收益和许多自我改进,但在物理世界中则不然。」
Brendan Foody · From Job Displacement to AI Trainers

分歧在哪

表面上五个人都"反对失业末日论"。但把原话摆在一起,分歧反而比共识刺眼。

一、最终态:工作消失 vs. 工作永续

这是最深的裂缝。Khosla 认为终局是"工作的需求消失",而且把大部分现存工作直接判为"奴役":

"We're going to be in an era of abundance that's so large, it's very hard for people to imagine. The simplest way to say it is the need to work will go away. People will work on things because they want to work on things, not because they need to work on things to pay their mortgage."
「我们将进入一个非常富足的时代,人们很难想象。最简单的说法是,人们不再需要为了生计而工作。人们会因为想做某事而工作,而不是因为需要工作来支付抵押贷款。」
Vinod Khosla · Vinod Khosla: Predicting the Future
"Most of the jobs in this country or on this planet are not really jobs. There's what I call servitude. … Those aren't jobs. They are servitude because you have to survive."
「这个国家或这个星球上的大部分工作实际上并不是真正的工作。那是我所说的奴役。……这些不是工作。它们是奴役,因为你必须生存。」
Vinod Khosla · Vinod Khosla: Predicting the Future

Andreessen 几乎是逐条反对:他认为"一夜变天"在结构上不可能,失业潮会被宏观增长淹没,而且关键论点是人口塌缩使人力反成稀缺资源——这恰好把 Khosla 的"工作消失"翻转成"人手不够":

"Even if AI triples productivity growth in the economy, which would be a massively big deal, it would take us back to the same level of job churn that was happening between 1870 and 1930. … There will be some level of like task level and job level substitution that will take place but that will be swamped by the macro effects of economic growth and innovation that will happen and then corresponding to that, there will be hiring booms quite honestly all over the place."
「即使人工智能使经济的生产力增长三倍,那也将是一件非常重要的事情,它只会让我们回到 1870 年至 1930 年间发生的相同水平的工作流失率。……在任务层面和工作层面会发生一定程度的替代,但这将被经济增长和创新带来的宏观效应所淹没,与之相对应,坦率地说,到处都会出现招聘热潮。」
Marc Andreessen · Marc Andreessen: The real AI boom hasn't even started
"If we didn't have AI we'd be in a panic right now about what's going to happen to the economy … depopulation without new technology would just mean that the economy shrinks. … the timing has worked out miraculously well in the sense that we're going to have AI and robots precisely when we actually need them. … the remaining human workers are going to be at a premium, not at a discount."
「如果我们没有人工智能,我们现在就会对经济将会发生什么感到恐慌……如果没有新技术,人口减少只会意味着经济萎缩。……时机安排得非常巧妙,因为我们将在真正需要人工智能和机器人的时候拥有它们。……剩余的人力劳动者将会变得更加珍贵,而不是打折。」
Marc Andreessen · Marc Andreessen: The real AI boom hasn't even started

值得注意的是,Khosla 自己也提到机器人会比人做更多劳动("a billion bipedal robots … doing more work than human beings do labor today"),所以分歧不在"机器人会不会接管体力",而在人力到底是变稀缺(Andreessen)还是变多余(Khosla)——同样的技术,两个相反的劳动力市场结论。

Cowen/Tabarrok 站在 Andreessen 这边,但给出更技术的理由——比较优势 + 时间约束,即便 AI 样样更强也成立:

"Comparative advantage says is that even if somebody is better than you at everything, there is still opportunities for trade so long as there is some limit, like a time limit on what they can do. … Martha Stewart, she's the world's best ironer, but she still hires somebody to iron her clothes because she has better things to do. And the AIs, if they have better things to do … then the AIs are still going to need somebody to plug them in and, you know, clean up or whatever."
「比较优势所说的是,即使某人在所有方面都比你强,也仍然可以进行贸易,只要有某种限制,比如他们能做事情的时间限制。……玛莎·斯图尔特,她是世界上最好的熨衣师,但她仍然雇人熨衣服,因为她还有更重要的事情要做。而人工智能如果有更重要的事要做……那么人工智能仍然需要有人给它们插电、清理之类的。」
Alex Tabarrok · Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth

但 Tabarrok 自己点破了这个论证的命门(同阵营内部的诚实):一旦 AI 不受约束(免费、无限复制、无时间/资本/能源约束),比较优势就失效:

"It might be that the robots are better, but for the scenario that you envisioned, you need more than that. You need that they're so cheap and so replicable. That there's no constraint. … So you need more than they're better than people. You also need that they're not constrained so that … they have no reason to cooperate."
「可能机器人更优秀,但在你设想的场景中,你需要更多。你需要它们便宜且易于复制。这样就没有限制。……所以你需要的不只是它们比人类更强。你还需要它们没有限制,不需要合作。」
Alex Tabarrok · Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth

——这正是 Khosla 假设成立、而 Cowen/Tabarrok 认为"太早、还没到"的世界。两边其实在同一个 if 上分手。

二、有没有民粹/政治崩盘:Foody/Cowen 说有,Andreessen 说被吓过头

Foody 是五人里对短期最悲观的,而且他不是从外部观察、是从卖"AI 训练员"的位置上说的:

"I think displacement in a lot of roles is going to happen very quickly, and it's going to be very painful and a large political problem. I think we're going to have a big populist movement around this and all the displacement that's going to happen."
「我认为许多岗位的替代将会很快发生,这将非常痛苦,并将成为一个重大的政治问题。我认为围绕着即将发生的所有人员替代,我们将面临一场巨大的民粹主义运动。」
Brendan Foody · From Job Displacement to AI Trainers
"It feels like the bulk of the country is not really coming to grasp with how fast jobs will be displaced. … a lot of this happens at economic contractions when people get more efficient … a lot of it hasn't happened yet, but it's going to happen imminently."
「感觉上,大部分人都没有意识到工作岗位消失的速度有多快。……很多这种情况发生在经济收缩时期,人们效率更高……很多事情还没有发生,但很快就会发生。」
Brendan Foody · From Job Displacement to AI Trainers

Cowen 在"上层中产阶级会塌"这点上同意有政治风险,而且说自己没有答案——但他把它归为"可管理的动荡":

"I think it will be tough. I mean, frankly, it's one of my biggest worries. … We're going to have to deal with some major political problems … I don't have really good answers for it or maybe not any answers at all."
「我认为这将是艰难的。坦白说,这是我最大的担忧之一。……我们必须应对一些重大的政治问题……我对此并没有很好的答案,或者可能根本没有答案。」
Tyler Cowen · Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth

Andreessen 直接否认"全民失业-没钱"那一支,而且和 Cowen/Tabarrok 用的是同一个塞氏定律论证(降本=通缩=变相涨薪),所以连保护伞都更便宜:

"There's no scenario in which everybody is just poor. In fact, it's quite the opposite, which is everybody gets a lot richer because prices collapse and then it's actually much easier to pay for the social safety net for the people who, for some reason, can't find a job."
「不存在每个人都很穷的场景。事实上,情况恰恰相反,每个人都变得更加富有,因为价格暴跌,而且实际上更容易为那些因为某些原因找不到工作的人提供社会保障。」
Marc Andreessen · Marc Andreessen: The real AI boom hasn't even started

张力在于:Foody 看到的是"经济收缩时一次性裁掉"的痛苦时点,Andreessen 看到的是"50 年技术停滞后终于解冻"的宏观顺风。两人对同一段近期窗口给出的体感完全相反。

三、"重新框定"是不是回避问题:Tabarrok 的 50%↔半工作周

Tabarrok 提出了本主题最锋利的修辞——把恐慌重新框定为闲暇:

"Suppose I tell you that AI is going to create 50% unemployment. … That sounds terrible. … Suppose, however, that I tell you that the work week will be cut in half. … That actually sounds glorious. … And yet, these are almost the same thing. … at the beginning of the Industrial Revolution … 3,000 hours a year. … Today, that number is half … 1,500 hours a year. … what has happened is not permanent mass unemployment. Instead, we have had shorter work weeks, fewer hours, a longer childhood and a bigger retirement."
「假设我告诉你,人工智能将造成 50% 的失业率。……这听起来太可怕了。然而,假设我告诉你,工作周将减少一半。……这实际上听起来很美好。然而,这几乎是同样的事情。……在工业革命初期……年度工作时间为每年 3000 小时。……今天,这个数字减半,变成每年 1500 小时。……发生的事情并不是永久性的失业。相反,我们有了更短的工作周、更少的工时、更长的童年和更长的退休生活。」
Alex Tabarrok · Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth

这与 Foody 的"重大政治问题"摆在一起就显出张力:Tabarrok 说 50% 失业 ≈ 半工作周("almost the same thing, subject to distribution"——而 distribution 正是 Foody 担心的那条);Foody 担心的恰恰是分配端的痛苦不会自动变成均匀的闲暇。Tabarrok 用"subject to distribution"一笔带过的,正是 Foody 说要"花很多时间思考"的核心。

四、技能漂移有多快、要不要让孩子学编程

具体到个人该做什么,数字差异很大。Roslansky 给了 LinkedIn 的量化漂移:

"If you go back to 2015 to now, which is just nine years, the average job on LinkedIn and the skills that are required to do that job on average have changed by 25%. … Our guess is that … It'll probably change 70% by 2030 like through AI, but I think that this, you know, change is a good thing."
「如果你回溯到 2015 年到现在,也就是短短九年时间,LinkedIn 上的平均工作以及完成这项工作所需的技能平均变化了 25%。……我们猜测,到 2030 年,通过 AI,这个比例可能会改变 70%,但我认为这种改变是件好事。」
Ryan Roslansky · LinkedIn's CEO On How AI Changes Your Job

在"孩子学不学编程"上,两位关系密切的人给出相反答案。Andreessen:学,而且要学到汇编/机器码那么深:

"I want to be one of the best software people in the world and I want to build new software products and technologies that really matter. You want your skill set to go all the way down to assembly and machine code. … if you don't know how to write the code yourself, you don't know how to evaluate what the coding bots are giving you."
「我想成为世界上最好的软件工程师之一,我想构建真正重要的新软件产品和技术。你希望你的技能能够深入到汇编和机器代码层面。……如果你不知道如何自己编写代码,你就不知道如何评估编码机器人给你的东西。」
Marc Andreessen · Marc Andreessen: The real AI boom hasn't even started

Foody:大概率不会推孩子学计算机科学,押注"反直觉的品味/通用推理":

"I would Probably not push them towards teaching their kids computer science … I am skeptical that like the really valuable thing is just people who can code in five years. … it's much more likely like the people that have these contrarian ideas around what's missing in markets and have the taste of what like features and nuances need to go into solving that problem."
「我可能不会鼓励他们教孩子计算机科学……我怀疑真正有价值的事情仅仅是那些五年内就能编码的人。……更有可能的是那些对市场上缺少的东西有相反想法的人,并且知道解决这个问题需要哪些功能和细微差别。」
Brendan Foody · From Job Displacement to AI Trainers

Khosla 跟 Foody 一边但理由不同——他押"好奇心",因为"上学-好大学-好工作"这条范式本身要废:

"A five-year-old kid going to kindergarten today is being told by their parents, go to school, study hard, get into good college, and then you'll get a good job. That won't be the paradigm."
「一个五岁的孩子今天去幼儿园,他们的父母会告诉他们,去上学,努力学习,考上好大学,然后你就会找到一份好工作。那将不再是范式。」
Vinod Khosla · Vinod Khosla: Predicting the Future

五、谁先扛过去:"白领专业服务"内部也不齐

都说"法律/金融/咨询首当其冲",但谁在保护层、谁先倒,看法不一。Andreessen 的解释偏监管卡特尔——技术早到位,是制度挡着:

"Large parts of the medical system today are they are cartels. … the doctors are a cartel and like nurses are a cartel, like hospitals are a cartel. … Guess what cartels and monopolies don't like? They don't like rapid change. … ChatGPT is almost certainly a better doctor than your doctor today. But like ChatGPT can't get a license to practice medicine, so it can't substitute for a doctor, it can't prescribe medications, it can't perform procedures."
「今天医疗系统的大部分都是卡特尔。……医生是卡特尔,护士是卡特尔,医院是卡特尔。……猜猜卡特尔和垄断不喜欢什么?他们不喜欢快速的变化。……ChatGPT 几乎肯定比你今天的医生要好。但是 ChatGPT 无法获得行医执照,因此它不能替代医生,不能开药,也不能执行手术。」
Marc Andreessen · Marc Andreessen: The real AI boom hasn't even started

Khosla 用同一个医疗例子得出相反的乐观结论——不是现有公司转型,而是被全新玩家重写,监管只能拖一阵:

"If all medical expertise is free, you have an unlimited number of primary care doctors, oncologists, gastroenterologists, mental health therapists … how would you redesign the healthcare system? … That transition won't happen from existing companies. Almost somebody new will reinvent this."
「如果所有的医疗专业知识都是免费的,你拥有无限数量的初级保健医生、肿瘤科医生、肠胃病学家、心理健康治疗师……你们会如何重新设计医疗保健系统?这种转变不会从现有公司开始。几乎肯定会有新公司来重新创造它。」
Vinod Khosla · Vinod Khosla: Predicting the Future

而 Tabarrok 对医生高薪的态度则是直接说它该降——他不把"专业卡特尔被冲掉"当损失,当好事:

"The average physician in the United States has an income of $500,000 a year. That's kind of too high. … the reason they have it is because … we don't have enough skilled workers. … Broaden the medical skill and reduce some of those wages."
「美国平均医生的年收入为 50 万美元。这有点高了。……他们之所以有,是因为……我们没有足够的熟练工人。……扩大医疗技能,并降低一些工资。」
Alex Tabarrok · Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth

都没说透的

我的看法

(以下是判断,非转述。)我的把握:中等偏低——五人样本里没有一个真正的"灭岗末日论者",所以这组访谈呈现的"共识"部分被选样放大了;真正持永久性大规模失业观点的声音(本主题定义里点名的那一派)在取材中是缺席的、只作为被反驳的稻草人出现。在此偏差下,我认为真正的分歧不在"会不会有新工作",而在过渡期的速度与分配——而恰恰这两点上,越乐观的发言人讲得越虚(Tabarrok 的"subject to distribution"、Andreessen 的"social safety net 更便宜")。Foody 因为身处替代的卖方位置、利益上本该唱多,却是最直白说"痛苦、民粹、imminent"的那个,我倾向于给他这条更高的权重。

还想知道什么

取材

继续读
线索
阅读线索：agent 不到 9 nines 可靠，那应该卖给企业还是消费者
线索
阅读线索：如果 coding 是 commodity，AI startup 的护城河长在哪儿
线索
阅读线索：推理成本一年跌 10×，AI app 该按什么定价
线索
摘报
综述
AI Agent 安全与治理
综述
Agent 代购时代：广告与商业的重写