# Liu Feng's team proposes AI 'Robinson Test': can AI survive independently after humans are withdrawn?

- Source: 新智元 · 微信公众号
- Published: 2026-10-30T11:56:00.000Z
- RoboHot: https://robohot.news/en/items/nq3wckrxa2q18e71zdyxb5qpf
- Original: https://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&mid=2652739321&idx=3&sn=3520f85a4a99d9338baf0af86ea24513&chksm=f0208018b579f78f8d58f937fe3fb426897d15cbfb586373468251e9b37dd9437d4559b39af8&scene=126&sessionid=1793370622#rd

## Summary

The research team of Dr. Liu Feng, author of "The Rise of Superintelligence," has exploratorily proposed the AI "Robinson Test": after providing necessary supplies, human maintenance and task instructions are withdrawn to observe whether an AI can sustain itself, expand its living space, and whether it holds a cooperative, adversarial, or neutral attitude toward humans. The test is divided into a roughly three-month nurturing period, a roughly one-month handover period, and a one-year hands-off period. Inside a physically isolated local network, batteries, servers, robots with replaceable parts, and operating manuals are left behind, and disturbances such as power outages and hard drive failures are randomly injected. Results are graded from level zero to level four, with behavior toward humans marked separately. The team says this is only a preliminary concept and still needs calibration in pilot trials.