Chinese Tech Workers Begin to Push Back Against Training AI That Replicates Them

Chinese tech workers are expressing apprehension about the process of training artificial intelligence (AI) that could replace them. As projects to create 'AI doubles' that replicate colleagues' skills and personalities proliferate across workplaces, laborers are raising concerns about their jobs being taken away.

Tianyi Zhou, an engineer at the Shanghai Artificial Intelligence Laboratory, created the 'Colleague Skill' GitHub project as a reaction to AI-related layoffs and companies' trend of demanding employee automation. This project gained traction and spread on Chinese social media. The project aims to 'distill' a colleague's work methods and personality so an AI agent can replicate them, and it offers functionality to automatically generate work manuals using records from Chinese internal messengers like Lark and DingTalk.

Amber Li, a 27-year-old tech worker residing in Shanghai, created AI replicas of all her former colleagues as a personal experiment after encountering 'Colleague Skill' on social media. Within minutes, the tool generated a file detailing how that employee performed their work. Li assessed the tool as accurate, noting it captured even the other person's minor habits or quirks at the end of sentences. Using these replicas, AI agents could assist with code debugging or provide immediate answers as new 'colleagues.' Li described the experience as peculiar yet unsettling.

While some Chinese tech workers are undertaking AI replica experiments of former colleagues, there is a nationwide trend of increasing demands from superiors for AI experiments, with tools like OpenClaw becoming popular. As AI agent tools such as 'OpenClaw' gain popularity across China, companies are encouraging tech workers to experiment with AI agents. These AI agents can perform various tasks including computer control, news summarization, email replies, and restaurant reservations.

However, frontline tech workers state that the practical utility of these agents is limited in a business context. Requiring employees to create manuals detailing their daily work, like with 'Colleague Skill,' serves as one way to bridge this gap. For employees, the process of creating AI agents or writing blueprints for them can be an alienating and disengaging experience.

Hanqing Chao, an assistant professor at Emory University who conducts research on AI and labor, explained that companies have valid reasons for pushing employees to create these work blueprints, beyond just following trends. He believes companies can gain richer data on employees' know-how, workflows, and decision-making patterns, in addition to building internal experience with these tools. This helps companies understand which parts of work can be standardized or systemized, and which parts still rely on human judgment.