这篇帖子从信任角度剖析了AI写作的致命伤，还拿爱马仕包包打比方，观点很犀利，值得一看。
Anthropic的Claude开始为生成文本添加水印，帮助识别AI内容。作者认为即使没有水印，AI写作也无法成功，因为写作本质是信息传递或娱乐，而AI生成的内容缺乏真实性。读者一旦发现内容由AI生成，信任度会下降100%，而非20%。作者将AI内容比作奢侈品，强调手工艺和真实性才是价值所在，建议仅将AI用于研究和事实核查。
Claude is starting to watermark its generated text, which will help others identify that it was made...
Claude is starting to watermark its generated text, which will help others identify that it was made with AI. But even if Anthropic didn't make this decision, AI writing was never going to work... No matter how good the model gets. No matter how much money the AI labs spend to make it sound human. Here's why I think it won't work: Writing is just a medium. The actual outcome is information transfer or entertainment. Agreed? Think about it. Anything you write on X (could be about your company, yourself, whatever). The end goal is either to entertain someone or inform someone. Writing and words are just a medium to reach that outcome. Now when you use AI, you know yourself that you are not being true to yourself. It's another system speaking. Choosing what words to use. Even what to say, what ideas to say. It's not 100% authentic. It's not you. That's true, right? It is NOT truly you. It cannot be. Because you wouldn't have said the same thing. What are the odds that you would have said the same thing in the same order? Zero! So now that we've established that AI writing is not truly you, that's step one. Step two is the audience realizing that you have used AI. And once the reader finds out that this wasn't posted by Brian authentically, the desire to read it doesn't go down by 20%. It goes down by 100%. Because the reader will feel: "oh this is written by AI. It's slop. It's garbage. It's not even what Brian truly thinks." And that loss in trust lasts. Because your face and brand get associated with using AI to communicate your ideas. Note that views are not the same as trust. You can farm views with AI and still train people that your name means mass produced text. It's kind of like luxury goods. Brands like Burberry would rather destroy inventory or eat the cost than flood the market with discounts. The very fact that there is more supply doesn't reduce demand by 20%. It reduces demand by 100% for their bags. Nobody wants something that is easily replicable or mass produced. And that is what AI slop is. Mass produced. Nobody has really put in the time to create that piece of content beautifully and painstakingly. It's not handcrafted like a Hermes Birkin. Everyone knows a Birkin is made by a single artisan from start to finish. That time and craft IS the reason people want it. Now if Brian took 5 minutes to draft that AI slop thread, people know it's not even worth 10 minutes of their time to read it. But if people can tell Brian took an hour to write that thread, they will spare 10 minutes for his real thoughts. And probably millions of people will spend those 10 minutes. I hope this makes sense. My favorite way to use AI when it comes to writing is research and fact checking. Just use it to make sure you know what you're talking about and not making a stupid mistake with a specific fact or something. But stay away from it when it comes to generating ideas. And please for god's sake stay away when it comes to writing. It's very easy to tell. It's been a good year for companies that benefit from more content. X, Substack, platforms, Instagram. AI means content becomes easier to produce. But that doesn't mean it's good content. If you're reading this, please just run in the other direction when it comes to using AI for writing. Basically: assume that if you're using AI, the other person knows you are using AI. Are you fine with that? If you're posting on Twitter... then no. You're probably not fine with your followers knowing you have used AI. If you're vibecoding a presentation for a colleague at work... You might be fine with them knowing it was created with AI. But if instead of a colleague it's your boss, you might not be fine with that now. Differs by situation! But just assume that if you're using AI, the person on the other end knows it. Polymarket @Polymarket JUST IN: Claude will now invisibly watermark AI-generated text so it can be detected after being copied & pasted. 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 1 👀 2834 ⚡