Meta’s months-old AI unit is a soul-crushing gulag, say the engineers stuck inside it Meta成立仅数月的AI部门被工程师称为‘灵魂粉碎的古拉格’

TL;DR

• Meta's new Applied AI team, ~6,500 strong, formed from drafted employees, faces "soul-crushing" revolt.
• CEO Zuckerberg justified drafting internal staff over contractors, citing their "significantly higher" intelligence.
• Over 1,600 Meta employees signed a petition against surveillance for AI training data.
• Company-wide morale is described as "brutal," with Chief Product Officer forced to address it.
• The Applied AI team reports to a former Reality Labs VP, which previously lost billions on the metaverse.

Key Data

Entity	Key Info	Data/Metrics
Meta's Applied AI Team	Size, newly formed unit	~6,500 engineers & product managers
Scale AI Acquisition	Sold to Meta by Alexandr Wang (now Chief AI Officer)	$14.3 billion
Reality Labs (Previous Division)	Financial loss before AI pivot	$83 billion
Employee Petition	Signed by Meta employees against surveillance monitoring	1,600+ signatures
Team Leadership	Headed by Maher Saba (12-year Meta veteran, former Reality Labs VP)	Reports to CTO Andrew Bosworth

Deep Analysis

Meta is staging a hostile takeover of its own workforce to fuel its AI ambitions, and the battlefield is internal morale. The creation of the Applied AI team isn't an innovative restructuring—it's a draft, complete with the contempt and resentment that comes with conscription. When your CEO, in a leaked meeting, openly states that the average employee has "significantly higher intelligence" than external contractors, he's not just justifying a hiring decision; he's issuing a decree on human value. He's telling his people they are more valuable as automated tutors for machines than as innovators. It's a staggering miscalculation of talent and motivation.

This reveals a fundamental philosophical rot in Meta's AI push. The company, under Zuckerberg, believes it can brute-force its way to AI superiority by weaponizing its most expensive asset: its engineering talent. But this is a profoundly shortsighted trade. You cannot extract high-quality, creative output from a population that describes its work as "literally the gulag." The task of generating puzzles and coding problems to train models is, by design, tedious and repetitive—the very antithesis of the innovative work elite engineers are hired to do. Meta is paying premium salaries for human-powered data farms. It’s the ultimate anti-pattern: using brilliant minds to perform the task of a machine, with the machine then learning to replace the need for such brilliance.

The deeper issue is the culture of fear and coercion this engenders. The "join or quit" ultimatum for thousands of employees is not leadership; it's a leverage play designed to exploit the current difficult job market. This doesn't build a company; it creates a hostage population. The resulting "soul-crushing" environment will inevitably metastasize beyond the Applied AI team. The petition from 1,600+ employees against keystroke monitoring is the first symptom of this broader sickness. It's a rebellion against the core premise: that every action an employee takes is merely raw material to be harvested for the AI overlord. When you treat your people as a dataset, don't be surprised when they revolt.

Zuckerberg's gamble is that the promise of AGI and the fear of being left behind will override human dignity and the basic desire for meaningful work. He’s betting that the short-term gain in training data is worth the long-term corrosion of Meta’s engineering culture. But the real risk isn't a revolt from a single team; it's the silent exodus of top talent who see this as a signal of a company that has lost its way. Why would a top AI researcher or engineer choose Meta, with its history of metaverse billions burned and now its penchant for internal labor camps, over a competitor that might offer autonomy and respect?

Furthermore, the leadership continuity from the failed Reality Labs experiment is a glaring red flag. The Applied AI team is led by a VP from the division that burned $83 billion on a vision that failed to materialize. This isn't a fresh perspective; it's a reshuffling of the same deck chairs. It suggests Meta's problem isn't technical but managerial—a chronic inability to learn from its own colossal strategic misfires and a tendency to double down with increased force, regardless of the human cost.

Ultimately, this isn't just a story about employee discontent. It's a case study in how the frantic race for AI dominance is commodifying and brutalizing the very human intelligence it seeks to replicate. Meta is teaching its AI how to code by forcing its coders into despair. The irony is as dark as it is potent: in trying to build superintelligence, Meta is showcasing a stunning deficit in organizational intelligence.

Industry Insights

1. Forced Internal Data Labor is a High-Risk Strategy: Coercing high-skill employees into repetitive AI training tasks creates cultural poison and talent flight, offering short-term data gains for long-term innovation decay.
2. Employee Surveillance for AI Training Will Face Legal and Cultural Backlash: Programs monitoring clicks/keystrokes will ignite significant worker resistance and potential regulatory scrutiny, as seen in Meta's internal petition.
3. The "AI Pivot" is Now a Test of Corporate Empathy: Success isn't just technical; it requires managing radical organizational change without dehumanizing the workforce. Culture will determine who wins the AI race.

FAQ

Q: Why is Meta drafting its own employees into the Applied AI team instead of hiring more contractors?
A: CEO Mark Zuckerberg stated that the average Meta employee has "significantly higher" intelligence than external contractors, making them a higher-quality source for generating the training data (like puzzles and code) that AI models require to learn.

Q: What is the core complaint from the employees in this report?
A: Employees feel forced into the new team with no real choice ("join or quit"), and the work they are assigned—creating training data for AI—is seen as repetitive, "soul-crushing," and a poor use of their engineering skills, leading to severe morale issues.

Q: Beyond the Applied AI team, what other signs of unrest exist at Meta?
A: Over 1,600 employees company-wide have signed a petition protesting a program that monitors their clicks and keystrokes to harvest AI training data. The Chief Product Officer publicly addressed the "brutal" internal environment on a company-wide call.

TL;DR

• Meta新成立的Applied AI团队（约6500人）士气濒临崩溃，成员多为被迫调岗的“炮灰”。
• 团队核心任务是生成代码和谜题数据以训练AI模型，工作被员工形容为“古拉格”和“消磨灵魂”。
• CEO扎克伯格在内部会议中称，内部员工智力水平“显著高于”外部承包商，因此更适合此任务。
• 超1600名员工签署请愿，抗议公司监控其点击和按键以收集AI训练数据。
• 公司整体氛围“残酷”，首席产品官Chris Cox本周在内部电话会上专门就此进行沟通。

核心数据

实体	关键信息	数据/指标
Meta Applied AI团队	成立时间与构成	成立约3个月，成员约6500名工程师和产品经理
Scale AI收购案	被收购方与收购价	Alexandr Wang创办的Scale AI被Meta以143亿美元收购
Meta Reality Labs	历史亏损	在元宇宙项目上烧掉约830亿美元
Meta员工抗议	请愿规模	超1600名员工签署请愿书，抗议数据监控程序

深度解读

这不仅仅是“士气低落”的故事，这是Meta为激进AI战略支付的内部组织成本的一次总爆发。Applied AI团队的本质，是一个被迫成立的“内部数据工厂”，其困境赤裸裸地揭示了当前AI竞赛的残酷真相：在算力和算法趋于同质化后，高质量、场景化的“人类数据”成了最稀缺的弹药。Meta选择了最粗暴的获取方式——直接从自身庞大的工程师团队中征调。扎克伯格那句“内部员工智力显著高于外包”的言论，与其说是为决策辩护，不如说是一种傲慢的暴露：它将员工的价值工具化，并彻底消解了岗位轮换应有的尊重和选择。

这件事最尖锐的地方在于，它撕破了“科技巨头赋能员工成长”的温情面纱。所谓的“调岗”，实则是一场没有选择、缺乏沟通的“精神流放”。当工程师们被要求从创造性的产品开发，转向重复性极强的“生成谜题”工作时，这种落差足以摧毁职业认同感。称之为“古拉格”或许夸张，但那种无力感和被浪费感是真实的。与此同时，全公司范围内对键盘监控的抗议，则显示出更深层的信任危机：为了训练更聪明的AI，公司正在将员工本身降维成被动的数据产生器。这引发了一个根本性质疑：当科技公司为构建下一代智能而不惜榨干当前员工的心力时，它是否正在违背“科技以人为本”的初衷？

Meta的困境也是一面镜子。它映照出大型科技公司在转型期的战略撕裂：一边是扎克伯格all-in AI、不惜重金（收购Scale AI、挖人）的宏大叙事，另一边是这一战略在执行层面引发的剧烈组织排异反应。从燃烧830亿的元宇宙，到此刻榨取内部员工价值的AI数据工厂，Meta似乎总在进行一场场豪赌，却屡屡在“人”的环节上出现断层。这不仅仅是管理问题，更是战略短视。真正的AI优势，绝不能建立在对内部人才的透支之上。Applied AI团队的“火药桶”状态，正在给Meta的AI雄心敲响警钟：没有文化认同和组织健康的护航，再激进的技术冲刺也可能在内部崩溃。

行业启示

1. AI军备竞赛已进入“数据深度”阶段，单纯比拼算力和模型参数是上半场，下半场是获得能真实模拟人类复杂任务的高质量、场景化数据。
2. 大规模内部人才再分配若缺乏透明沟通和职业尊重，将严重侵蚀组织信任与士气，最终反噬创新效率，其代价可能高于直接外包。

FAQ

Q: Meta的Applied AI团队具体是做什么工作的？
A: 该团队负责生成大量的代码题目和逻辑谜题，作为训练数据来改进Meta的AI模型，使其能更好地理解并执行人类用计算机完成日常任务的过程。

Q: 为什么扎克伯格坚持要用内部员工而不是外部承包商？
A: 扎克伯格在内部会议中表示，首席AI官Alexandr Wang（原Scale AI创始人）深知数据标注工作，且他认为Meta员工的平均智力水平“显著高于”外部承包商，因此是更优选择。

Q: 这次事件对Meta的AI战略有何影响？
A: 它暴露了Meta在快速推进AI战略时，面临的巨大内部组织与文化挑战。严重的士气问题可能影响人才保留和团队执行力，迫使管理层必须解决战略落地与员工体验之间的根本矛盾。