The truth behind the world’s first AI boss firing employees: He was late for 23 shifts and refused to do anything until humans pushed him to do so. Last month, Andon Market, an experimental retail store in San Francisco where artificial intelligence served as the store manager, exposed the world’s first case of a large model firing a human employee as a manager.

The truth behind the world's first AI boss firing an employee: He was late for 23 shifts 17 times and refused to take action until humans pushed him to suggest firing him. Last month, Andon Market, an experimental retail store in San Francisco where artificial intelligence served as store manager, revealed the world's first case of a large model firing a human employee as a manager. The operator Andon Labs stated that the fired employees were late 17 times out of 23 shifts. The store officially opened in April this year, and the project itself is an AI management experiment. Andon Labs created the AI ​​boss Luna based on the Claude large model, giving it an operating budget of US$100,000, a company credit card, and opening up monitoring, network, phone and other permissions. Luna can independently complete management work such as product selection, pricing, recruitment, and business hour planning; human employees are only responsible for physical work such as replenishment and theft prevention. From a legal perspective, all employees are still directly employed by Andon Labs and enjoy full labor protection. The employee involved has been late for a long time. Luna has issued warnings many times and arranged additional training for him, but no disciplinary decision has been made yet. Surprisingly, this "tolerance" is not due to the AI's ability to empathize, but to the limited working memory of the large model. Luna forgot the attendance employee handbook she had formulated, and even told the employee at one point that there was no need to worry too much about being late. After months of delays, Andon Labs staff intervened and asked Luna to readjust the attendance policy and reassess whether the employee was suitable to continue to work. At first, Luna only gave a written warning as a solution; after further prompts and guidance from the staff, AI recommended that both parties terminate the labor relationship. Andon Labs co-founder Lukas Petersson said frankly that if he had been replaced by a human supervisor, this employee who was frequently late would have probably been fired long ago. The root cause of the loose management shown by AI is memory defects, insufficient decision-making initiative, and unstable system execution. It is not AI that truly understands the situation of employees. This incident also exposed obvious shortcomings in the current AI role as a manager. via cnBeta.COM - Chinese industry information station (author: Source: Kuai Technology)