文章探讨了对人工智能（AI）的过度依赖问题，特别是作者在使用ChatGPT后的体验。作者发现自己逐渐依赖AI解决问题，类似于过去依赖导航软件的情况。文章强调了这种依赖可能导致独立思考和解决问题能力的退化。为了应对这种依赖，作者尝试减少对AI的依赖，回归传统方式，将AI视为辅助工具而非万能答案，并重新重视人际关系和独立思考的重要性。

🤔作者分享了对ChatGPT的过度依赖体验。起初，ChatGPT在编码和情感支持方面提供了巨大帮助，但随后作者开始过度依赖它，甚至在面临问题时首先想到的是如何向AI提问。

🗺️作者将这种依赖与过去对Google Maps的依赖进行了类比。过度依赖导航工具导致作者丧失了基本的方向感和独立解决问题的能力，就像依赖ChatGPT一样，可能会削弱自身的解决问题的能力。

💡作者尝试通过减少对AI的依赖来解决这个问题。他将AI视为一个知识渊博的同事，而不是解决所有问题的捷径。作者也重新开始依赖朋友，而不是ChatGPT来解决情感问题。

💻作者指出了AI在某些方面的局限性。例如，ChatGPT在处理训练数据之外的新技术时会遇到困难。作者认为，过度依赖AI会阻碍个人成长，并可能导致社交技能的退化。

Published on April 27, 2025 8:18 AM GMT

 ChatGPT recently introduced improved cross-chat memory, and I've found myself using it a lot more sense then. Everything from helping write components with unfamiliar JavaScript libraries, to counseling me through a breakup. So much so that when I face a new problem, my mind turns not to breaking it down myself, but how to structure it as a prompt for o3 to help me solve. As useful as these models are, I think I'm experiencing first hand a new type of dependency. 

This parallels a similar problem I had before LLMs got large enough to be useful. I've always been more of a word-cell than a shape-rotator, so when navigating unfamiliar areas, it's easier to pull out Google Maps than actually understand where things are, and how to get there. So ever since I learned to drive, I would follow Google's unwavering blue line to my destination. And if I ever faced a roadblock, it would seamlessly reroute me along the next fastest path. However like ChatGPT, I found myself completely helpless without this tool. To the point where if my phone died, I would be forced to charge it before I could find my way home.

I'm now finding that instead of just having trouble navigating from point A to point B, I struggle to navigate life without the advice of my pocket oracle. I've thought about embracing it, and purchasing Meta glasses to fill in the gaps where ChatGPT doesn't fit, but I've seen where that leads. Instead I'm trying an experiment: doing things the old fashioned way. Instead of generating most of my code, I'm treating AI like a knowledgeable colleague who I don't want to bother too frequently. Rather than using o3 like a cheatcode to my love life, I'm leaning more on my friends.

o3 has no problem one-shotting a boiler-plate frontend, but repeatedly failed to help me code a particle editor using a newer library than existed in it's training data. Also, sitting in my basement staring at ChatGPT.com is not (yet) the best way to maintain friendships and find a new boyfriend.

Discuss