AI淘汰得最快的是普通人：既不追求卓越也不追求出错的人 昨天下午去听了一场AI沙龙，主题是“你能打败AI吗？—人类与大语言模型的创造力大比拼”，教授主要是展示他去年的一篇论文，通过定题，实地采样几万人（包括山区留守儿童）的数据，和现有的十个最新的语言模型（包括去年1月新出的deepseek），由算法来评分，结论是创造力一致——即ai和普通人创造力水平差不多 AI吃过苹果吗？ 这个问题很有趣，“语言模型比较容易生成苹果这个词，但是AI绝对没吃过苹果。” 这个教授的背景是艺术转心理学转AI方向研究，他在这里提出一个“以人为本”的理念，引出了这个问题。虽然很温和地劝大家别焦虑，但有些事情不以人的意志为转移，综合应用下来，AI淘汰得最快的就是“平均水平”的人，有个听众提了个问题：以后社会怎么处理“无用阶层”？ 嗯，这非常东亚视角。离我们最近的未来大概率是人机协同工作，就业市场不会再提供那么多普通的岗位，但我觉得“担责”的岗位是基本需求，就是出了问题找谁，不可能找机器吧，所以基于这个层面考虑还是无法失去人的节点。 AI eliminates “average people” the fastest — those who neither pursue excellence nor dare to fail. Yesterday afternoon I attended an AI salon titled “Can You Beat AI? — A Creativity Showdown Between Humans and LLMs.” The professor presented a paper from last year: through fixed prompts and large-scale field sampling (tens of thousands of participants, including left-behind children in rural areas), human responses were compared with outputs from ten state-of-the-art language models (including DeepSeek released last January), all scored algorithmically. The conclusion: creativity is roughly the same — AI performs on par with ordinary people. “Has AI ever eaten an apple?” This question stuck with me. Language models can easily generate the word apple, but AI has never actually eaten one. The professor, who transitioned from art to psychology to AI research, used this to introduce a human-centered perspective. He gently told the audience not to be too anxious — but some things don’t change based on reassurance alone. When applied at scale, AI replaces average labor the fastest. One audience member asked: How will society deal with the ‘useless class’ in the future? That’s a very East Asian way of framing it. In the near future, human-AI collaboration is far more likely than total replacement. The job market won’t offer as many “generic” roles anymore, but roles that carry responsibility will remain essential. When something goes wrong, accountability has to land somewhere — and it won’t be on a machine. From that perspective, the human node is still irreplaceable.