A serious internal conflict is brewing within Meta's Applied AI division. Employees are complaining about extremely harsh working conditions caused by monotonous tasks for generating technical data to train neural networks and strict monitoring of their activity.
What Happened
Conflict has escalated between staff and management within Meta's Applied AI division, which employs approximately 6,500 people. Engineers have been faced with the necessity of performing routine tasks to generate technical data, which they characterize as a "soul-crushing gulag." The situation was further complicated by an incident involving the hijacking of a corporate stream and protests against monitoring systems that track employee clicks and keystrokes. The division is led by Maher Saba under CTO Andrew Bosworth.
Context
The problem is linked to the scaling of data preparation for training Large Language Models (LLMs). Instead of using specialized contractors like Scale AI, Meta is utilizing highly qualified in-house ML engineers to perform tasks comparable to low-skilled data labeling.
Why It Matters for the Industry
The situation exposes a systemic industry problem: the growing need for high-quality synthetic data is leading to the degradation of working conditions for high-end specialists. Using expensive engineers for mechanical work creates an inefficient cost structure, risks talent burnout, and raises questions about the ethics of data preparation processes in major AI labs.
Why It Matters for Users
For readers and specialists, this serves as an example of how even in top tech companies, the AI development process can turn into mechanical routine, undermining internal culture and team stability and creating risks of losing key competencies.
Sources
Author
Look at AI, Editorial Team