Title: 你在用 AI 写文章吗？ | BestBlogs.dev
URL Source: https://www.bestblogs.dev/article/e610376d
Published Time: 2026-03-25 10:14:00
Markdown Content: Skip to main content Toggle navigation menu Toggle navigation menuArticlesPodcastsVideosTweetsSourcesNewsletters
⌘K
Change language Switch ThemeSign In
Narrow Mode
你在用 AI 写文章吗？
阑 阑夕 @阑夕
One Sentence Summary
This article explores the phenomenon of over-reliance on AI-generated content and cognitive degradation in the AI era, emphasizing the importance of maintaining human 'taste' and discernment in AI-assisted creation.
Summary
Written by veteran tech commentator Lan Xi, this article explores the alienation of human cognition and reading habits following the widespread adoption of AI. The author points out that the public's instinctive rejection of long-form content and hypersensitivity to 'AI-generated' labels reflect a degradation of human discernment in the AI era. The article emphasizes that while AI can provide basic material, humans must possess high-level judgment and taste to master the tool, rather than being 'dumbed down' by it. Furthermore, the author reflects on the reverse conditioning of human expression by AI training data, calling for the preservation of uniqueness and depth in human creation during the AI era.
Main Points
* 1. The core capability in the AI era is discernment, not mere generation.AI can only provide 70-point material; users need 90-point cognition and taste to filter, integrate, and imbue content with a soul. Otherwise, they will fall into a passive state of total acceptance. * 2. AI is causing a 'dumbing down' of human cognitive abilities.The public's rejection of long-form content and hypersensitivity to AI-generated content reflect a loss of patience for deep reading and independent thinking under AI assistance, cheapening everything that cannot be easily replicated. * 3. Beware of the reverse conditioning of human expression by AI.When AI overuses certain sentence structures (e.g., 'not X, but Y'), humans abandon these expressions to avoid suspicion. This phenomenon of being reverse-shaped by algorithms is unreasonable; we should insist on the autonomy of human creation.
Metadata
AI Score
84
Website mp.weixin.qq.com
Published At Today
Length 1470 words (about 6 min)
Sign in to use highlight and note-taking features for a better reading experience. Sign in now
原创 →2026-03-25 18:14 北京
Slop和Taste，是AI时代被频繁提及的两个概念。
文 ｜ 阑夕
一个观察，不一定对：
我发现被AI驯化的人有个特点，看到稍微长一点的文本，就会怀疑这是不是AI写的，并本能地想让AI来做总结。
在我的社交媒体评论区里，新粉丝也多了起来，张口闭口就是「这是不是AI写的？」
Slop和Taste，是AI时代被频繁提及的两个概念。
翻译过来，就是泔水和品味，它们看似针锋相对，但很多时候其实是一体两面。
就像我经常看到，有人刚刚大赞Taste在AI时代有多重要，转头就把一条Slop搬到群里，浑然不觉其中矛盾。
在我看来，AI带来的最大错觉，是让没脑子的人，以为自己也有脑子，而无法分辨区别，是用不用得好AI的分水岭。
什么意思呢？
就 用写作来举例吧，AI可以生产70分的内容，但你必须要有90分的能力，才能知道怎样去使用那些70分的材料，如果你只有50分，那就意味着你只能全盘接受，而无从分辨对错。
罗素他老人家说，「这个世界的问题就是，智者充满疑虑，蠢货坚信不疑」，要我说，疑虑的未必都是智者，坚信的一定是蠢货。
AI几乎无限的生产能力带来的另外一个副产品，就是在发的长文底下总有大聪明说这是不是AI写的，不能说完全不糟心吧，就是非常倒反天罡：
AI正是靠着我这样的人贡献的语料完成训练——而且我很确信自己的占比——才有了今天的本事，现在倒好，被强夺者被当成了索取者。
此处应有斯内普看到哈利波特对自己使用神锋无影的表情包：
你竟敢用我发明的魔法对付我？
我其实很能理解这类评论的用户画像，他们大概率是从来就没有消费长文的能力，在没有AI的时代会说「太长不看」，有了AI之后更会如此廉价化的对待一切无法复现的行为。
一边被漫山遍野的Slop灌成泡芙，一边将所视之物都当作Slop予以简化，这就是我说的，没有分辨能力，就无从判断好坏。
更重要的是，在很多人看来，写作是一件负担，是很辛苦的生产工作，因为自己自小如此，所以便不可能有人把写作视为乐趣，怎么可能会有人主动投入那么多的时间、写这么长的文章，要么是AI写的，要么是有人付钱的商单。
和菜头前段时间也写过一篇文章「机器永不眠」，意思是在写作上去和机器卷效率是毫无意义的，AI可以让写东西变得简单、高效和自动化，但真正让他感到愉悦的，是在写作中遇到的麻烦和困难，以及如何塑造了自己。
「我手里拿着来这个世界的单程票，我想我在最早一定有个什么目的，这个目的一定不是来这个世界上轻松躺着，等机器帮我完成一切。」
是的，机器写作即便成功，大概也留不住读者，这是非常功利主义的实话，如果我真要兜售AI生产的商品，那无异于自断财路的自杀行为，我的每一个客户，都可以自己用AI生产内容，要我做甚。
认知的边界，就是世界的边界，我谈科幻文艺，谈电影和宇宙，就是想分享这些，然后就有人忍不住自证，自己所处的方寸之间，才是天地，你的那些表达，王八念经，不听不听。
而且计量单位也是一个让我很难有同理心的地方，为什么那么多人会对千字篇幅的文本喊长，这是要把大脑寄存到多么贫瘠的荒野里，才会出现的面露难色？
比AI降智可怕一万倍的，是它让人类的降智。
前几天看到祝佳音说，他看到一篇文章用了「不是、而是」的句式，下意识的以为这又是AI写的，然后看到发表日期是在2017年，这才重新陷入沉思。
这也是我说的倒反天罡，我自己有时候都会特别注意，尽量避开「不是、而是」的表达，否则就会和AI「同流合污」了，但这实际上是不合理的，AI把一种经典的文本技巧用烂了用吐了，最后是要人类反过来迁就AI，这对吗？
需要我帮你出一版更加犀利、简洁的文章吗？ 跳转微信打开
阑 阑夕 @阑夕
One Sentence Summary
This article explores the phenomenon of over-reliance on AI-generated content and cognitive degradation in the AI era, emphasizing the importance of maintaining human 'taste' and discernment in AI-assisted creation.
Summary
Written by veteran tech commentator Lan Xi, this article explores the alienation of human cognition and reading habits following the widespread adoption of AI. The author points out that the public's instinctive rejection of long-form content and hypersensitivity to 'AI-generated' labels reflect a degradation of human discernment in the AI era. The article emphasizes that while AI can provide basic material, humans must possess high-level judgment and taste to master the tool, rather than being 'dumbed down' by it. Furthermore, the author reflects on the reverse conditioning of human expression by AI training data, calling for the preservation of uniqueness and depth in human creation during the AI era.
Main Points
* 1. The core capability in the AI era is discernment, not mere generation.
AI can only provide 70-point material; users need 90-point cognition and taste to filter, integrate, and imbue content with a soul. Otherwise, they will fall into a passive state of total acceptance.
* 2. AI is causing a 'dumbing down' of human cognitive abilities.
The public's rejection of long-form content and hypersensitivity to AI-generated content reflect a loss of patience for deep reading and independent thinking under AI assistance, cheapening everything that cannot be easily replicated.
* 3. Beware of the reverse conditioning of human expression by AI.
When AI overuses certain sentence structures (e.g., 'not X, but Y'), humans abandon these expressions to avoid suspicion. This phenomenon of being reverse-shaped by algorithms is unreasonable; we should insist on the autonomy of human creation.
Key Quotes
* In my view, the greatest illusion brought by AI is making mindless people think they have a brain. The inability to distinguish this difference is the watershed between using AI well and using it poorly. * What is ten thousand times scarier than AI 'dumbing down' is it causing humans to dumb themselves down. * AI was trained precisely on the corpus contributed by people like me... Now, the exploited have been turned into the exploiters.
AI Score
84
Website mp.weixin.qq.com
Published At Today
Length 1470 words (about 6 min)
Tags
AI
Cognitive Upgrade
Writing
Slop
Taste
Related Articles
* #410. Marc Andreessen: The Real AI Boom Hasn't Started Yet, We Are at a Historic Turning Point * Qwen Grows Multiple Arms and Legs * New Species PixVerse R1: Breaking into the No-Man's Land of LLMs * Vol.89 AI Industry 2025 Annual Summary Supplement (V4 "Can't Wait" Edition) --- 70-page PPT Solo * #434. Survival Rules in the AI Era: Naval on Vibe Coding, Personal Leverage, and the Future of Creativity * Is the Browser Worth Rebuilding Again? * Vol.86 Same Generation Technology, Two Systems: A 181-Page PPT Record of the AI Industry in 2025 * Anxiety-Peddling in the AI Circle Has Gone Too Far * #437. A Conversation with Elad Gil: SaaS Shifts, Growth Myths, and Defensive Strategies in the AI Era * #423. Silicon Valley Godfather Ben Horowitz: American Ambition, Management Philosophy, and Culture Building in the AI Era HomeArticlesPodcastsVideosTweets