Meta's AI Unit Plunges into Disarray
Meta's ambitious pivot to artificial intelligence has hit a significant snag, with its recently established Applied AI unit experiencing acute internal dissatisfaction and a near-rebellion among its 6,500 engineers and product managers. Formed in March, this unit is described by multiple current employees as a "soul-crushing gulag," where highly paid Silicon Valley talent are now "draftees" assigned to menial tasks. These tasks primarily involve generating puzzles and coding problems to train Meta's AI models.
The severity of the situation became public during a livestreamed internal meeting, open to thousands of employees, which was disrupted by an expletive-filled outburst. An attendee reportedly urged others to "tell him that he's a piece of sh_t" in reference to a senior Meta AI executive, causing one presenter to cover their face in shock. Many employees claim they were transferred to this group against their will, given an ultimatum to "join or quit."
Forced Reassignments and Demoralizing Work
Engineers within the Applied AI unit, many of whom refer to themselves as "draftees," express significant frustration over their new roles. The work, which includes creating coding challenges and puzzles for AI systems, is widely considered monotonous and demoralizing, a stark contrast to their previous software development positions. One employee explicitly called the division a "literal gulag," while others echoed sentiments of the work being "soul-crushing."
The unit, which reports to Meta CTO Andrew Bosworth and is led by Maher Saba, a former vice president in Reality Labs, initially had a flat management structure where up to 50 employees answered to a single manager, further contributing to the chaotic environment. This reorganization follows Meta's significant investment in the metaverse through Reality Labs, which reportedly spent $83 billion before the company shifted its focus to AI.
Wider Employee Discontent and Surveillance Concerns
The unrest within the Applied AI unit is not an isolated incident. More than 1,600 Meta employees across the company have signed a petition protesting a program that monitors their mouse clicks and keystrokes for AI training data. This initiative, known as the Model Capability Initiative (MCI), was designed to capture data on how employees use their computers to develop AI agents capable of performing software-based tasks.
Employees raised concerns about the extensive scope of the monitoring, its impact on their privacy, and reports that the software significantly increased home internet usage, with some exhausting a month's data allowance in days. In response to the backlash, Meta has scaled back the program, introducing new controls that allow employees to pause data collection for up to 30 minutes and request exemptions. Meta's Chief Product Officer Chris Cox acknowledged the "brutal" atmosphere during an internal call, and CEO Mark Zuckerberg addressed the situation in an internal memo, admitting that recent changes had "caused distress" and that the company had made "mistakes" it intends to fix.