Summary of Key Points
The profession of data annotators, who "teach AI to understand the world," is now spreading to smaller cities in regions like Huishui in Guizhou and Binzhou in Shandong, becoming a new employment option for local college graduates, mothers at home, and unemployed individuals. However, workers generally face issues such as low salaries (around 3,000 yuan on average), monotonous and repetitive work, and high psychological stress. They are also trapped in the lowest-profit part of the supply chain due to the practice of "layered subcontracting." Large companies pass orders through multiple hands, and by the time they reach the county-level companies, the profit has been significantly reduced, significantly impacting the annotators' earnings. These workers hope for better treatment and policy support to help them escape their current difficult situation.
1. New Jobs in Small Cities: Who Is Doing Data Annotation?
Most annotators in small cities are people looking for work nearby:
- Recent college graduates: Like Zhou Xundi, a post-2000s graduate who found a data annotation job at a local job fair after struggling to find employment elsewhere, avoiding the need to leave home.
- Mothers at home and unemployed individuals: Zhang Shengyue's team includes mothers taking care of children at home and those looking for a new job; the work hours are relatively flexible (some can work from home), making it suitable for those who need to balance family and work.
- Local entrepreneurs: Zhang Shengyue returned to his hometown from Jinan to start a company, seeing the potential in the AI data market, but the entrepreneurial path has not been smooth.
For small cities, this is a rare new type of job that can attract young labor, addressing the issue of college graduates staying in their hometowns.
2. What Exactly Does the Work Involve?
In simple terms, annotators' job is to "teach AI to recognize things":
- Basic tasks: They use a mouse to create 3D frames from 3D point cloud data (a collection of points representing a three-dimensional world), enclosing objects like cars, pedestrians, roadblocks, and traffic lights, and aligning them with 2D photos taken by cameras to ensure accurate recognition by AI.
- Advanced requirements: Different projects have specific rules—some require smooth movement trajectories (e.g., pedestrians should not walk in a crooked line), while others have strict requirements for object coverage (e.g., a car partially obscured by a tree must be framed precisely).
- The most frustrating part: The work must be reviewed by quality control officers, and any mistakes require revision. Zhou Xundi once spent hours adjusting frames only to be told they did not fit the requirements, which nearly cost him his job.
Essentially, annotators provide the "nutrition" for AI, as AI cannot recognize the world on its own; human effort is needed to organize the raw data in a form that AI can learn from.
3. Low Pay, Heavy Workload, and High Stress: How Much Pressure Do They Face?
The low income and high stress of annotators are closely linked to their low salaries and job instability:
- Meager salaries: Zhou Xundi earned 2,300 yuan during his internship and expects a raise after becoming a full-time employee, but after deductions for attendance and performance, it barely covers his daily expenses. Employees in Zhang Shengyue's team earn between 2,500 and 3,000 yuan per month, with some earning up to 4,000 yuan, which is considered "okay" in small cities but not affluent.
- Challenges for employers: Zhang Shengyue has encountered situations where project clients disappeared, leaving him with unfinished work and no payment; the instability of orders means constant retraining for new projects, which is time-consuming.
- High psychological stress: The repetitive nature of the work and the need for frequent revisions can be very stressful. Zhou Xundi has even lost his temper due to these issues.
This job appears simple but actually requires a combination of physical, mental, and emotional effort.
4. The Dilemma of Layered Subcontracting: Why Can't They Make More Money?
The root of the low salaries lies in the thin profits at the lower end of the supply chain:
- Order distribution like peeling an onion: A large company (e.g., Baidu or Chery) may sell one frame of data for 10 yuan. After being subcontracted multiple times, the price drops significantly by the time it reaches the county-level companies, leaving only a few cents per frame.
- Limited access to direct orders: Large companies have strict requirements for company qualifications (e.g., a certain number of employees and experience), making it difficult for small county-level companies to obtain direct orders. They can only work with "secondhand" or "multi-level" orders.
- Unstable orders: Projects can be interrupted mid-way, resulting in wasted time and the need to start over with new requirements.
This is similar to the food chain, where only the top players benefit, leaving the lower levels with the scraps.
5. What Do They Hope for in the Future?
The workers' expectations are practical:
- Salary increases: They hope for a raise to improve their living standards.
- Policy support: They want the government to provide assistance to county-level annotation companies (such as subsidies and qualification support) to help them secure higher-level orders.
- Recognition of their role: Zhou Xundi hopes that the value of their work will be recognized—after all, their efforts are essential for the advancement of autonomous driving technology. He also hopes that the psychological pressure they face will be taken seriously, so they are not overlooked as "invisible" workers.
Data annotation is a fundamental part of the AI industry, but the living conditions of those who do this work deserve more attention.
(The entire article is written in plain language to make it understandable to non-financial professionals, revealing the real challenges faced by data annotators in small cities.)