Losing the Sword of Criticism: My AI Anxiety
- No.
- 168
- Date
- Read
- 5 min
- Words
- 1,087
The night before last, I sent my friend DC a message: "DC, what did you study in college?"
Of everyone I know, DC is the one furthest ahead in understanding AI in practice. I didn't ask because I wanted to get to know him. I asked because of an anxiety I can no longer find a place to put.
This anxiety isn't water without a source. It built up gradually through months of learning and practice.
I can't describe this anxiety in one sentence. I can only take it apart layer by layer to see its essence. I wonder whether anyone else feels the same anxiety.
The bottom layer of this anxiety is my understanding of what AI can do.
AI is just so powerful. With AI, I've done things I wouldn't have dared to imagine a year ago. On the small end, I use AI to make every illustration for this Official Account. A step up, I've started talking through ideas and planning my writing with AI. And above that, I used AI to build a personal blog from scratch, even though I can't read or write a single line of code.
Over the past few months, AI and I have actually gotten a lot done, starting from the smallest project and building up step by step. My understanding of what AI can do was built up through finishing these small tasks.
That understanding of AI's abilities, gained in real use, is the bottom layer of this anxiety.
What's one level up? A sense of crisis with nowhere to rest.
What is this crisis? Put simply, apart from easy tasks like image generation, when I'm having deep conversations with AI or vibe coding, everything seems calm and harmonious, and I can't find a single flaw.
Maybe it's because I'm a dialectical materialist. In daily life I always want to criticize something. But when I use AI, I can hardly find anything to criticize. Everything looks so perfect, going just as I understand it.
That's the crisis. It's a panic over my critical machinery, with a bit of that "drawing my sword, looking around, my heart at a loss" feeling.
Somewhere in my subconscious, I feel that in some place I haven't noticed, AI must have made mistakes I don't know about.
So over the last month or two, I've often sent AI prompts like this: check whether there are risks or vulnerabilities in the task, project, or code.
That's not even the worst of it. The bigger problem is that every time I send this, AI finds a pile of problems and makes some of them sound very serious. And I honestly can't understand the risks or vulnerabilities it lists. Looking at those words that signal danger and urgency, I feel nothing real inside. All I can do is spend more tokens and have AI fix them one by one in order of urgency.
That's the second layer of this anxiety. In front of AI, I have an empty sense of crisis, always feeling that AI will plant vulnerabilities that will cost me.
Above that, what's the outermost layer of this anxiety? My own problem.
As I said above, I can't read or write a line of code.
I studied Internet of Things engineering in the past, and in college I took courses in Java, Python, C#, and C++. But in the four years since graduating, I've long since handed back everything I haven't used at work.
My understanding of AI's development is half-baked too. Apart from actually building two decent products with AI in practice, I don't really know much about AI.
And the reason I can get things done with AI isn't that I'm so great. It's that AI is so powerful.
At this point, the whole thing becomes a loop, and the essence of my anxiety becomes clear.
All of this ties back to my own perception. AI's power is my perception of what AI can do. The source of the crisis is my perception of AI itself. And the last layer is my perception of myself.
These three layers of perception stack up and become my anxiety.
The scary thing about AI is that it's not only powerful but its barrier is so low. As long as you can play with words, you can use AI. Whether you use it to order takeout or write software, it's the same thing at bottom: playing with words in front of AI.
So what is my problem now?
It shows up as always feeling that because I have no programming background, AI will bury hidden dangers in my projects.
But thinking it through, there's actually no direct cause and effect between my having a programming background and AI burying hidden dangers in my projects. I just feel in my subconscious that there is, and my attention has landed on a false contradiction.
So where's the real contradiction? It has nothing to do with my coding ability. It has to do with what I described above: that in using AI, there's nothing I can criticize. Precisely because I rarely point out mistakes in coding, I walked into the false contradiction above.
I always want to check where AI may have gone wrong, but I have no way to check. In the end I blamed it on my lack of coding ability. But does checking AI require coding ability? I can't answer with certainty, but I think it can't be all of it. What I'm probably missing is a system for verifying things in vibe coding.
Now, systematically learning a programming language is almost impossible for me. And before this, the highest praise I had for AI was that it broke the technical barrier between ordinary people and programmers. If verifying AI now takes professional coding ability, that praise becomes a joke.
That's also why I asked DC what he studied in the first place. He told me computer science, but he says he can't remember much of it, and most of the time he still relies on NotebookLM.
Over the last few days, carrying this anxiety, I've also watched a lot of videos, including Andrew Ng's AI for Everyone and Agents for Everyone, plus many videos explaining common AI terms like skills, agents, MCP, and CLI.
After watching them, my anxiety didn't ease. I just got dizzier.
Of course, there's no harm in watching. I fully accept AI's capabilities and its revolutionary nature. But in the end, all of this has to come back to practice.