"Tell him, he's just an asshole." 6,500 engineers are forced to work for AI, and Meta's internal meeting suddenly spirals out of control: Someone bursts out swearing in public, angrily yelling "This place is a Gulag"

While Mark Zuckerberg is still spending tens of billions of dollars to recruit talents and build a "super-intelligent empire," increasingly harsh voices are emerging within Meta.

Last week, during an internal live-streamed meeting at Meta for thousands of employees, an employee suddenly interrupted the speech and uncontrollably blurted out a string of expletives. He complained that he had become a "tool for the company". Then he publicly named a Meta AI executive and asked the meeting host to convey a message to the person:

"Tell him he's a piece of shit."

According to Wired, a speaker on-site was so embarrassed that he covered his face with both hands. Although the host then asked everyone to mute and continued with the technical sharing, many employees were still discussing this "spicy" start in the chat area.

The new AI department of 6,500 people, some call it a "modern gulag"

The center of the storm is the new department, Applied AI (AAI), established by Meta in March this year. This department is large, with about 6,500 engineers and product managers. Its main task is to support Meta's newly established Meta Superintelligence Labs (MSL, Superintelligence Laboratory).

According to Meta's official statement, this team undertakes the important mission of helping AI models evolve. However, in the eyes of many employees, they are more like "forced laborers" conscripted temporarily.

According to several current employees, Meta formed the Applied AI department in a very crude way: many people did not apply to join voluntarily but suddenly received an email one day, being informed that they had been assigned to the new team. A Meta employee also posted on Reddit before, saying that the whole transfer process was "quite random."

And they basically have only two choices: either join Applied AI or leave Meta.

"This place is just like a gulag (a term for the Soviet-era labor camp system, now often used to describe a highly oppressive and autonomous-lacking work environment)," an employee described to the media. "Suddenly, your life loses its meaning. You hardly communicate with anyone and just keep completing those tasks every week."

Some people directly commented, "Almost everyone is unhappy and finds this job mentally exhausting."

People who used to write code are now setting questions for AI

So, what exactly do these engineers transferred to Applied AI do every day?

According to Meta's internal documents and information disclosed by employees, their main work is not to develop products or study model architectures but to produce training data. Specifically, they need to design programming questions, write complex software development scenarios, construct logical puzzles, evaluate model performance, create test samples, and label data.

Normally, each employee needs to complete at least two of the above tasks every week. These contents will ultimately be used by AI scientists to train and evaluate the latest generation of large models and help future AI Agents learn how to write code, operate software, and complete complex tasks.

Meta's explanation in an internal announcement is that current AI models still can't truly surpass humans in technical tasks such as programming, so a large number of real cases are needed for training: "In order for AI Agents to understand how humans use computers to complete daily work, we need to train the models with real cases."

In terms of technical difficulty, these tasks are simpler than previous development work. But the problem lies precisely in the fact that these jobs are too mechanical, repetitive, and lack creativity. An employee said, "This doesn't utilize my abilities and knowledge at all." He joined Meta originally to develop products for billions of users, but now he feels like he's "producing fodder" for AI models every day.

Actually, Meta had other options. Building training data has always been the work of a large number of outsourcing teams. One of the most well-known companies in the industry before was Scale AI. Last year, Meta completed a deal that shocked the industry:

It spent $14.3 billion to acquire the core business of Scale AI and recruited its founder, Alexandr Wang, who was only 28 years old at that time, to serve as Meta's Chief AI Officer and be in charge of Meta Superintelligence Labs.

In a previously leaked recording of an internal meeting, Zuckerberg explained why they didn't adopt the traditional outsourcing solution: Alexandr Wang knows the data annotation industry very well, and the average intelligence level of Meta employees is "significantly higher" than that of third-party contractors.

Therefore, Meta believes that it will be more effective for internal engineers to complete this work.

Not only employees, but even executives can't help but curse

Applied AI is not the only team with problems. The dissatisfaction within Meta has spread throughout the company.

Since this year, Zuckerberg has almost bet all of Meta's strategic focus on AI. To free up resources, the company has carried out a large-scale organizational restructuring. Just last month, Meta laid off about 8,000 employees, accounting for about 10% of the total number of employees.

Meanwhile, the company has also shifted a large amount of resources from the metaverse business to AI. Maher Saba, the head of Applied AI, previously came from Reality Labs - the department that has burned more than $83 billion in the past few years. In addition, several employees also revealed that the data center engineering team, the Instagram team, and the infrastructure department are all facing unprecedented work intensity.

To obtain more training data, Meta has also launched a controversial new plan: starting to monitor the mouse clicks and keyboard inputs of its US employees and using this data for AI model training. This practice quickly triggered a backlash. Currently, more than 1,600 employees have signed a joint petition, demanding that the project be stopped.

Under pressure, Meta then made some concessions: allowing employees to pause data collection for up to 30 minutes and apply for individual exemptions. However, these measures clearly failed to completely quell the controversy.

Not only employees, this anxiety has even spread to the management. According to a recording of an internal meeting obtained by Wired, at last week's all-hands meeting of Instagram, Meta's Chief Product Officer Chris Cox rarely publicly talked about the current state of the company.

He described the past few months as "difficult and cruel" and used a vivid metaphor to describe the employees' situation: "It's like running a marathon in a hailstorm. Suddenly, your teammates are replaced, and the company is monitoring you the whole time." At this point, Cox couldn't help but curse: "What the f**k."

In his opinion, Meta needs to rebuild the connection with its employees and avoid deifying AI: "AI is neither a god nor a devil. It's far from as powerful as you think, and it's also far from as bad as you think."

Zuckerberg admits: the company did make mistakes, but won't stop

Facing the continuously escalating dissatisfaction, Zuckerberg finally had to step in to comfort the employees.

In a recently issued internal memo, he admitted for the first time: "The recent organizational changes have indeed brought troubles to the employees." He also candidly said, "Due to the extreme complexity of these adjustments, we have made some mistakes, and we will probably continue to make mistakes in the future."

However, to stabilize the morale of the troops, Zuckerberg also promised:

• There will be no more large-scale layoffs this year;
• Limit the number of subordinates for managers;
• Increase the budget for team activities;
• Restore the fixed workstation system;
• Hold a large-scale Hackathon event.

The most concerned adjustment is optimizing the management structure. Previously, there was an extreme situation in the Applied AI team where "one manager managed 50 employees." In the future, the company will limit the number of subordinate employees.

Additionally, Zuckerberg also specifically mentioned Applied AI in the memo. He explained to the employees that this is not the end of their careers but a transitional stage: "Jobs like AAI are crucial for promoting the development of models and allow many very talented people to participate. In the next few months, we will create more new positions suitable for them."

However, at least for now, Zuckerberg's statement has not completely eliminated the employees' anxiety. After all, everyone knows that under the squeeze of competitors such as OpenAI, Google, and Anthropic, Meta is making a high-stakes AI bet on its future.

And those engineers transferred to Applied AI are at the forefront of this high-stakes bet - perhaps more accurately, they have been dragged into a war they never voluntarily signed up for.

Reference link:

https://www.wired.com/story/mark-zuckerberg-meta-employee-meeting-interrupt-ai/

This article is from the WeChat official account "CSDN", author: Zheng Liyuan. Republished by 36Kr with permission.