An AI robot is shown working at a computer (AI-generated image). Recently, an open-source project named "Colleague.skill" gained significant popularity on the technical community platform GitHub. Its function involves feeding an AI with "raw materials" such as messages from Feishu, DingTalk documents, emails, and screenshots from a departed colleague, along with some subjective descriptions of their personality, to generate an AI avatar that can continue performing their tasks. Netizens have humorously referred to this process as "refining."
After being laid off, must an employee leave behind the skills, business logic, and even their personal language style and methods of handling situations acquired during their tenure, to work for the company permanently? This technology has sparked widespread discussion on social media platforms.
Where does the data for training these AI avatars come from, and to what extent can it be used? How can intellectual property protection be strengthened and the ethical boundaries of technology clarified, so that the benefits of AI development reach more workers? Interviews from various sources provide an in-depth analysis of this phenomenon.
The concept of "cyber immortality" for former employees has recently become a hot topic. This technology extracts tacit knowledge such as work experience, communication style, and decision-making logic, enabling the AI avatar to possess a degree of human-like thinking ability and perform specific job tasks.
"Previously, leaving a job meant clearing your desk and handing over work; the skills and experience you gained remained your own. Now, resignation feels like leaving your soul at the company, achieving 'cyber work' with no clocking off," some netizens commented. Others have joked about creating digital versions of ex-partners or bosses, attempting to "seek torment."
Many interviewees believe that "refining" a colleague not only infringes on workers' intellectual property but also means that once various data is collected, the AI may cease to be a reliable assistant and could potentially be used as a tool to "optimize" or replace the very person it mimics. To preemptively "kill" their own digital分身, some netizens have tried "poisoning" their own data, suggesting that writing poor-quality code would result in the AI learning only useless information, leaving the company with worthless outputs.
Another netizen developed an "Anti-Colleague.skill" tool. Contrary to cloning others, it protects one's knowledge by replacing truly important core information with seemingly correct but substantively empty "correct-sounding nonsense," making it difficult to replicate.
Experts warn that while "Colleague.skill" appears to be a tool for rapidly boosting productivity, it harbors multiple hidden concerns.
First are the issues of worker privacy rights and intellectual property. Professor Sun Zaifu from Qingdao Huanghai University points out that currently, the legal status of experience and habits accumulated by individuals in the workplace—whether they constitute company property or the worker's own intellectual property—remains ambiguous. Cloning an employee's personal experience brought to the job, external learning outcomes, non-service inventions, and personal methodologies constitutes an infringement of their intellectual achievements. Furthermore, if AI-generated content leads to infringement, errors, or leaks, existing laws cannot clearly define the responsible party, potentially creating a legal vacuum where "everyone is responsible, yet no one is responsible."
Secondly, the iterative development of "Colleague.skill" brings anxiety about job replacement. Associate Professor Liu Xi from Shandong Normal University suggests that if this technology develops chaotically, it could directly lead to a significant reduction in human resource needs for certain positions, triggering issues like role adjustments, salary cuts, and layoffs.
Thirdly, talent development may struggle to keep pace with technological advancement. As AI continues to evolve, skill iteration cycles could shorten from 10 years to just 2-3 years, making "one skill for a lifetime" a thing of the past. Concurrently, vocational education models may shift from long-term degree programs to short-cycle, modular, micro-credential education, requiring individuals to transition from one-time education to lifelong learning, continuous updating, and dynamic adaptation.
"How can ordinary workers share in the dividends of this AI era?" is a common concern among interviewees.
Undeniably, while AI development may replace some basic roles, it will simultaneously create new jobs, such as AI trainers, prompt engineers, digital workforce managers, and algorithm compliance auditors. However, the number of these new positions is currently limited and cannot fully absorb the displaced workforce. The "race" between the speed of technological advancement and the emergence of new jobs, along with the construction of new social security systems, will require a prolonged period of adjustment.
On one hand, cultivating competencies that AI cannot replace and learning to leverage AI for problem-solving are crucial for making AI a wing for humanity. Liu Xi believes that as AI gradually takes over the "how" of execution, the decision-making value of "what" to do and "why" becomes more prominent. Universities should promptly adjust their programs and curricula, strengthening humanities, social sciences, and ethics education to enhance students' abilities in problem definition, value judgment, and comprehensive aesthetics.
Sun Zaifu suggests that companies can adopt a "human-machine coupling" approach, utilizing AI Skills for standardized, repetitive basic tasks, thereby freeing human employees from tedious labor to focus on high-value strategic decisions, emotional communication, and complex problem-solving.
As "ability algorithmization" becomes an unavoidable trend, workers might consider encapsulating their professional skills, work experience, and decision-making logic into reusable AI Skills. By offering these through subscription, licensing, or usage-based revenue sharing models to multiple companies, they could achieve a flexible employment pattern of "one person serving multiple enterprises," maximizing their labor value.
In the era of human-machine collaboration, how can we safeguard the primary status and irreplaceability of humans? Experts argue that current efforts should focus on laws and regulations, ethical review, and industry self-discipline to protect workers' legitimate rights and interests, laying a solid foundation for collaborative development.
— Close legal gaps to achieve rigid constraints. Sun Zaifu recommends further clarification under the framework of the Personal Information Protection Law: employees' work behavior patterns, communication styles, judgment preferences, and thinking logic belong to personal information, even sensitive personal information. Companies must obtain separate written consent for using such data for AI training. After the termination of employment, companies should delete personal trace data used for AI training within a stipulated period.
— Strengthen ethical reviews to reduce the cost of rights protection for workers. Sun Zaifu suggests that before systems like "Colleague.skill" are deployed in enterprises, they should undergo filing and ethical assessment with cyberspace, human resources, and market supervision authorities, focusing on the legality of data authorization, necessity of collection scope, and protection of labor rights.
— Improve industry self-discipline as a supplementary soft constraint. Liu Xi recommends that industry associations take the lead, collaborating with legal institutions and labor protection organizations to formulate self-regulatory conventions. This would promote industry自律, resist恶性 algorithm competition that devalues human professions and exacerbates employment imbalances, and ensure technological progress remains on a path aligned with civilization and the rule of law.
Comments