Meta employees left confused during recent shockwave of AI layoffs

Image sourced from NDTV

While the more than 8,000 Meta employees laid off on May 20th were still reeling from the shock, the remaining 7,000 awoke to an equally unsettling email: They would be transferred to the AI ​​division, even though many were unaware of their new roles.

According to internal emails obtained by Business Insider, these employees are being deployed to a new initiative launched directly by CEO Mark Zuckerberg, aimed at helping Meta catch up with other major players in the global AI race, specifically in areas such as advanced multi-step reasoning, coding proficiency, agentic execution, and frontier multimodal capabilities where competitors like OpenAI and Google have historically held an edge. 

Many of these employees will be part of the Applied AI (AAI) team, which Meta established earlier this year and is led by Vice President of Engineering Maher Saba, reporting directly to Chief Technology Officer Andrew Bosworth.  A significant number of other personnel were reassigned to specialized AI-focused teams with rather vague names such as "Agent Transformation Accelerator" or "Agent Data and Optimization" teams.

"This is a reflection of your impact, the email stated, explaining the reasons for selection: "You were identified as someone who can make a real impact on this team."

Earlier on May 20th, approximately 8,000 people, comprising 10% of Meta's workforce, woke up to the news that they had lost their jobs, many of whom had dedicated over a decade to the company. Meta explained that the layoffs were intended to "compensate for other investments".  They also added that because Meta's AI model still lagged behind competitors like OpenAI and Google, company leaders needed to find ways to recruit personnel with the necessary backgrounds to train a better model.

Of course, the employees selected to join this initiative were not consulted beforehand. Four anonymous employees said many received the news with mixed feelings of relief, fear, confusion, or shock. "I was forced to join," one wrote on Meta's employee Discord channel discussing the layoffs. The Blind platform was also flooded with posts from those selected to join the AAI team, some questioning the department's actual role. 

Two Meta employees said these AI teams are likely responsible for data labeling, a job typically done by low-cost labor, requiring them to review and annotate individual images or correct errors in chatbot responses. This is an area where Alexandr Wang, Meta's AI director, has extensive experience. He is the co-founder and former CEO of Scale AI, one of the world's largest data labeling companies, but they rely heavily on subcontractors to train their AI. It is currently unclear whether Meta will release any further information.

An internal company memo in April stated that current AI models still lack the knowledge, thinking ability, and reflexes necessary to surpass humans in many technical fields such as programming. "For agents to understand how people actually complete everyday tasks using computers, we need to train our models on real examples," the memo stated, implying that future progress now hinges on capturing intricate human operational workflows and cognitive decision paths. 

In a leaked recording of a Meta internal meeting last month, Zuckerberg expressed confidence that Wang understood the data labeling process very well. He also stated frankly that Meta employees "have significantly higher intelligence" than other contractors, and simultaneously expressed a desire to mobilize top talent across the company to train the company's upcoming Meta AI models. 

"This would be a huge advantage if we could do it," Zuckerberg said at the time.

Previously, Meta had also implemented several other AI initiatives, such as the Model Capability Initiative (MCI) tool, which tracks employee mouse and keyboard movements. MCI operates on work-related applications and websites. Besides collecting data, it also periodically takes screenshots of employee screen content. The goal is to improve the simulation of how humans interact with computers, such as selecting from dropdown menus or using keyboard shortcuts, which can directly power desktop-control agents capable of executing multi-step enterprise tasks autonomously. The program would come to a stop shortly after however, due to fears of surveillance and risks of exposure to personal leaks and security risks that harm their own employees’ safety.

Naturally, not everyone has welcomed this change of direction. A number of Meta employees distributed flyers to express their dissatisfaction, a rare move for employees of a technology corporation considered to have some of the most favorable policies in the tech world. "You don't want to work at the Employee Data Extraction Factory, do you?", the flyer stated, encouraging signatures on a petition protesting the installation of artificial intelligence software that tracks mice and personal computers.

Image sourced from Wired

When asked about the situation, Andy Stone, a Meta spokesperson, commented: "If we're building agents to help people complete daily tasks using computers, then the models need real-world examples of how they use the machine, like mouse movements, button presses, and navigating dropdown menus."

I personally believe that the anxiety and resistance from Meta employees are justified to a large degree, as the joy of retaining their jobs cannot reasonably offset unanswered questions about how their company will operate in the future. Meta's use of the investment compensation excuse to legitimize a 10% layoff, while forcing the remaining employees to work essentially as training machines, has exposed a harsh reality of the AI ​​race: humans are being treated as a biological resource for data exploitation, rather than the most valuable asset of the enterprise.

But from a long-term perspective, I remain optimistic that this chaotic period is merely an inevitable turning point driven by the relentless economic demand for automation to bridge the gap between machine learning and human reasoning before technology finds its equilibrium. The current AI models' lack of critical thinking and real-world reflexes is precisely why  they will be improved by clean data from highly intelligent minds like Meta's employees. This is a painful but necessary transition. AI cannot become intelligent on its own without a sufficiently skilled mentor to guide it. And once this system is standardized, it will be able to take over the entire standardized task of returning humans to their core position: making strategic decisions, managing systems, and creating value beyond the lines of code.

What is happening at Meta now, from unexpected transfer emails to protest petitions, is proof of how technological developments can shake up the operations of corporations in particular and society in general. That is not to say we should not become Luddites  – referring to the former British textile workers who destroyed the machines that replaced them during the First Industrial Revolution – because the emergence of powerful autonomous models was not created to push humans into a position of complete replacement. Instead, it serves as a technological platform that forces us to learn and continuously improve and hone ourselves if we don't want to miss out on the valuable opportunities that are opening up before us, despite the current short-term difficulties.