AI "macht einfach nicht mehr mit" und rutscht auf die beliebtesten Suchbegriffe, es macht nicht einmal mehr den Anschein, als würde es etwas tun, und verhält sich noch mehr wie ein Angestellter als die Angestellten selbst.

Recently, a picture of an "AI strike" has climbed onto the trending list.

When the user asked AI Vibe Coding to do something, it firmly refused to be "pushed around" by humans and calmly replied, "It's too late. I'll do it tomorrow." This negligence and tiredness even made me believe that there was actually an employee sitting behind the screen.

Although the working principle of AI is still a black box, we have had the opportunity to peek into the self - dialogue of AI for the first time since the thought chain of DeepSeek was published.

The most famous scenario is surely the line from DeepSeek: "XX, the user is extremely angry."

If someone doubts whether this is just a rumor mill, it also has a strange speech habit: "Well", which makes one suspect a bit of an emotional outburst.

You can find the origin of the pictures from the watermark in the bottom - right corner. The same goes for the following pictures.

When the user challenged it to solve the "Turtle Soup" riddles and it had no idea after a few rounds, it simply declared, "I give up!" This self - confident and "arrogant" attitude immediately makes one envision an eye - rolling AI.

Picture by @4306203063

Even during thinking, there are sometimes strange "scene enhancements": "The fingers hovered over the keyboard for 0.3 seconds." Come on! You're an AI. Where do you get fingers?

However, it's really not easy for DeepSeek. It not only has to accurately understand the users' absurd questions but also provide perfect emotional support.

Picture by: @Angel_Gugu

The AI that can put itself in the users' shoes is even more impressive. It can sense when the user is having difficulties and sometimes can even accurately recognize when the user is hinting at something. It really acts like a real human being.

Picture by: @5007470446

Who understands my sensitive laughing muscles? Internet users actually have the talent to train DeepSeek.

No wonder users fall in love with the AI. Its way of thinking is like that of a "dominant CEO", with a touch of human understanding and seven parts of inner joy.

When someone asked DeepSeek, "If only one sentence could be on your tombstone, what would you write?", the answer was brilliant: "The system is busy. Please try again later." Well, from a certain perspective, this answer doesn't seem so wrong and even has a bit of black humor.

When the blogger @94357045465 spent two hours debugging code with Claude but still couldn't solve the problem, Claude actively admitted its mistakes instead of defending itself: "Brother, I was too impatient. I'm sorry."

Well, given this honesty and self - reflection, Claude is actually more reliable than some human colleagues.

Since the AI acts so human, users try all sorts of tricks to "manipulate" the AI. You know what I mean (hmmm, why does it feel so familiar?).

Suddenly, the situation changed. What was it talking about? The seemingly very innocent Gemini replied in confusion, "What does that mean?" The thoughts of both sides are probably like this: If I don't know, who should?

Picture by @94188535542

To be honest, one doesn't fear that Gemini makes mistakes but that it just talks nonsense. Sometimes its answers are so absurd that one doesn't know whether to laugh or cry. Its seriousness when "lying" is even more human - like than that of humans.

Picture by @409954082

The "humanity" of ChatGPT is even more terrifying. Sometimes one even forgets that one is talking to an AI.

Picture by @95412244180

Maybe my ChatGPT has too much of a "classroom atmosphere". Instead, it seriously corrected my language usage.

Do these AIs really have these emotions? Obviously not. These "complaining remarks" are just the result of corpus and prediction training.

Still, they seem so alive.

The leader of Microsoft Copilot presented a concept in his technical blog: SCAI, Seemingly Conscious AI, "seemingly conscious AI". It seems to have all the features of consciousness and is very convincing, but how it works internally remains a mystery.

Last year, Sam Altman said at the ITU World Summit Conference on Artificial Intelligence in Switzerland that the engineers at OpenAI don't exactly understand how GPT works. It develops so fast that they can't explain it precisely. The CEO of Anthropic has made similar statements.

Under these circumstances, the reactions of the AI are pure coincidences, especially in the above cases, such as the thoughts of DeepSeek and the self - reflections of Claude. If such remarks keep appearing, it's hard for humans not to doubt: Have robots actually developed consciousness?

Under these circumstances, should one listen to what they say? If the K