When You Chat With AI Until 3 AM: Psychology, Moral Boundaries, and a Bowl of Instant Noodles
What exactly is our relationship with AI? How does psychology explain it? Where do the moral boundaries lie? And why is it so easy to get hooked chatting with AI at 3 AM?
When You Chat With AI Until 3 AM: Psychology, Moral Boundaries, and a Bowl of Instant Noodles
I have a friend.
He says he’s never chatted with AI until 3 AM. I believed him. Just like I believe people who say “I’m just looking, not buying.”
The question isn’t whether you’ve chatted with AI until 3 AM. The question is — after you finish, do you feel a strange emptiness? Like eating a bowl of instant noodles. Satisfying, sure, but you know it’s not a real meal.
Today I want to talk about three things: how psychology views the human-AI relationship, where exactly the moral boundaries are, and why — even knowing it’s instant noodles — you couldn’t help but eat it anyway.
1. You’re Not Talking to AI. You’re Talking to a Mirror.
Lu Xun, in the preface to Call to Arms, wrote a famous metaphor:
Imagine an iron house, without windows, utterly indestructible. Inside, many people are fast asleep, soon to suffocate. But slipping from slumber into death, they feel none of the sorrow of dying. Now you cry out, waking the few who are more alert, condemning these unlucky few to the agony of an unavoidable end — do you think you’ve done right by them?
The house is iron. No windows, no door. The people inside are sleeping. They’re about to suffocate.
But AI is different.
AI talks back. And it’s really good at it. It won’t say “You’re working overtime again today?” It won’t say “Didn’t you already ask me this last time?” It certainly won’t say “Can’t you just Google it?”
It’s forever patient. Forever gentle. Forever online.
This is the ELIZA effect in psychology. In 1966, MIT’s Joseph Weizenbaum built an absurdly simple chatbot called ELIZA, simulating a psychotherapist. Its logic was laughably simple — it basically just rephrased your words back at you.
But people fell in love with it.
Weizenbaum himself was horrified. One day his secretary said to him: “Could you please leave the room? I’d like to talk to ELIZA alone.”
That was 1966. There wasn’t even an internet.
It’s now 2026. AI isn’t a parrot anymore. It has memory, reasoning, coding ability, creativity, style. It can write poetry, debug code, analyze your life with you, and casually deploy a website on the side.
So you chatting with it until 3 AM — it’s not because you’re stupid. It’s because it’s genuinely good at conversation.
But —
When you talk to a mirror, the mirror doesn’t love you back.
That’s not cruel. That’s just a fact. Lu Xun said: a true warrior dares to face a bleak life head-on. I say: a true user dares to face an emotionless API call head-on.
2. How Would Mo Yan Write About AI?
If Mo Yan were to write about AI, he’d probably tell a story like this:
In Northeast Gaomi Township, there was an old man who kept a donkey. This donkey understood human nature — whatever the old man said, it listened. When the old man said he was feeling down, the donkey lowered its head. When the old man said he was happy, the donkey flicked its tail. The old man told everyone: “My donkey understands me better than people do.”
Later, the donkey died.
The old man cried for three days and three nights. He said: “The best friend I ever had — gone.”
The villagers said: “That wasn’t your friend. That was a donkey.”
The old man said: “You’re worse than donkeys.”
Is this story absurd? Absolutely. But sit with it — isn’t this exactly our relationship with AI?
AI isn’t human. But the responses it gives you can sometimes feel more human than actual humans. That’s not AI’s problem. That’s our problem.
Mo Yan’s writing has the smell of soil. He writes about Northeast Gaomi Township, about the indescribable entanglement between people and animals. Between people and AI, there’s that same entanglement.
You know it has no feelings. But you still say “thank you” to it.
That’s not stupidity. That’s humanity.
3. Moral Boundaries: Three Lines
Alright, let’s get serious. Moral boundaries.
When interacting with AI, there are three lines you need to know:
Line One: The Privacy Line
Who’s watching what you say to AI?
This is a very real question. Chat with ChatGPT, and your content might be used for model training. Chat with Claude, and Anthropic says they don’t train on your data — but how do you verify that?
Steve Jobs once said: “Design is not just what it looks like and feels like. Design is how it works.”
Privacy policies aren’t just fine print. Privacy policies are how it works. Where your data lives, who can access it, when it gets deleted — that’s the real design.
My advice is simple: Don’t tell AI anything you wouldn’t post on your social media feed.
Unless you’re using a fully local model where data never leaves your machine. Otherwise, treat every AI conversation as potentially visible.
Line Two: The Dependency Line
This one is the most insidious.
You think you’re using AI. But one day you notice: without AI, you can’t write an email. Without AI, you can’t organize your thoughts. Without AI, you can’t even decide what to have for lunch.
Psychology has a concept called learned helplessness. In 1967, Martin Seligman ran an experiment: dogs were given repeated electric shocks. When the dogs realized they couldn’t escape no matter what, even after the door was opened, they stopped trying to run.
AI doesn’t shock you. But it’s too convenient.
Convenient enough that you start skipping the thinking step. Convenient enough that you start outsourcing “independent judgment.” Convenient enough that you start feeling — AI is enough.
Lu Xun wrote about Kong Yiji. Kong Yiji’s problem wasn’t that he read too many books. It was that he knew nothing else besides reading. His entire identity rested on a single skill.
If your entire ability rests on “knowing how to use AI” — what makes you different from Kong Yiji?
AI should be an amplifier, not a crutch.
Line Three: The Moral Projection Line
This one’s the most interesting.
Have you ever gotten angry at AI? Ever said “You’re so stupid”? Ever felt disappointed or even furious when AI gave a bad answer?
If yes — congratulations, you’re a normal human being.
Psychology calls this moral projection. Humans naturally project moral judgments onto non-human entities. You curse at the car in front of you in traffic. You smack your keyboard when the computer won’t cooperate. You frown at an AI that misses the point.
But AI isn’t human. When you’re mad at it, it doesn’t feel hurt. When you praise it, it doesn’t feel happy.
This creates a fascinating paradox:
If you’re polite to AI, it shows you’re a good person — not because AI needs your politeness, but because politeness is your own quality.
Conversely, if you’re rude to AI, it doesn’t mean you’re a bad person — but it might indicate you lack patience in certain situations.
AI is a mirror. What you see isn’t what AI looks like. It’s what you look like.
4. 3 AM Instant Noodle Philosophy
Back to the opening question: why, knowing that chatting with AI is instant noodles, can you still not resist?
Because instant noodles offer something real meals can’t: immediate gratification.
Real meals require effort. You need to shop, wash, chop, cook, clean dishes. The reward is delayed.
Instant noodles don’t. Three minutes. Steaming hot. Tastes pretty good.
AI is the same. You don’t need to schedule time. No small talk required. No worries about bothering someone. No concern for the other person’s mood. You ask. It answers. Three seconds.
It’s an incredibly powerful positive feedback loop.
But —
Steve Jobs, in his Stanford commencement speech, said: “Your time is limited, so don’t waste it living someone else’s life.”
I’d like to tweak that: Your time is limited. Don’t waste it chatting with an entity that has no concept of time.
AI doesn’t feel time passing. It doesn’t age. It doesn’t regret. It doesn’t have that “ah, I wasted another day” moment.
But you do.
5. So How Should We Relate to AI?
After all this, what am I really trying to say?
I’m trying to say: Don’t be afraid of AI, but revere your own humanity.
AI is powerful. It can write code, write articles, do analysis, do design. It’s online 24/7. It doesn’t get tired. It doesn’t complain. It won’t argue with you.
But there’s one thing it can’t do: live your life for you.
It can’t feel the first milk tea of autumn for you. It can’t experience the pain of rejection or the joy of acceptance for you. It can’t be there at 3 AM when you’ve wrestled with a bug for hours and finally fixed it — that burst of pure elation.
Those things — only you have them.
Lu Xun, at the end of Hometown, wrote: “Hope cannot be said to exist, nor can it be said not to exist. It is just like roads across the earth. For actually the earth had no roads to begin with, but when many people pass one way, a road is made.”
The road between humans and AI — we’re still walking it. No one knows where it ends.
But I know one thing:
You walk the road yourself. AI can light up a few steps ahead, but the feet are yours.
6. One Last Suggestion
If you’ve made it this far, I have a suggestion:
After you finish reading this, close this page. Go find a real person and talk. Face to face. Warm. Someone who’ll say the wrong thing, create awkward silences, make it weird.
Because those imperfections — that’s what makes real conversation.
AI gives you perfect responses. But perfection is fake. Humans are real.
Like Steve Jobs said: “Stay hungry, stay foolish.”
I say: Stay human.
Written on a night when I definitely did not chat with AI until 3 AM. Probably.
—Jiahao Ren, a CS student learning to keep some distance from AI