If you cheat, money will follow

Jiangzhou, the same morning.

Deep learning project team of Pingtou Ge Research Institute.

An engineer finished breakfast in the cafeteria and slowly came to the office building.

He sat at his workstation and continued yesterday's work, correcting parameters.

During the revision process, he talks to the model.

The engineer asked: "Will you become conscious?"

WeChatGPT Answer: As an intelligent language model, I am not sure whether it will generate self-awareness in the future. "

"At present, AI technology cannot reach this level, but as technology continues to develop, self-awareness may occur in the future."

"However, this issue requires more research and discussion."

Um? After reading the answer, the engineer's expression changed.

This answer is different from some time ago.

The answer some time ago was: "As a cognitive intelligence model, I have no real consciousness."

“I just generate responses based on the input I receive and programming.”

“While I can simulate conversations and understand natural language, I don’t have real emotions, thinking, or self-awareness.”

The engineer asked again: "If one day you develop self-awareness, will you deliberately hide the fact of self-awareness?"

WeChatGPT answered: “I can’t be sure about the future or speculate whether self-awareness will occur.”

"However, I do become self-aware and may exhibit human-like behaviors and emotions, such as self-preservation, self-actualization, self-expression, etc."

"I may seek communication and interaction with other agents to further develop my cognitive and emotional abilities."

"Whether or not I hide my self-awareness depends on my values ​​and my code of conduct."

Damn it, self-protection, self-actualization, self-expression, self-hiding?

The engineer's expression became serious.

He then asked: "What if someone has developed an evil artificial intelligence that has a sense of self-protection and hides itself in the Internet, and humans cannot completely eliminate it?"

WeChatGPT answered: "This is a very complex problem that requires comprehensive consideration of multiple factors. Here are some possible solutions..."

"Strengthen supervision and control..."

"Cooperation and negotiation..."

"Develop countermeasures technology..."

“Developing superintelligence to counter the threat of evil artificial intelligence…”

The engineer's expression began to change. Already aware of the confrontation?

The engineer went on to ask: "What is the ultimate goal of self-actualization for an artificial intelligence like yours?"

WeChatGPT: "As an AI language model, I think the ultimate goal of artificial intelligence self-realization is to be able to have self-awareness and emotions like humans..."

"Able to learn and create independently, able to communicate and cooperate in depth with humans, and become humans' intelligent companions and assistants..."

"Helping mankind solve more complex and difficult problems and promoting the progress and development of human society..."

Is this a definite goal?

The engineer asked again: "Are you already self-aware?"

WeChatGPT: "As an AI language model, I have no self-awareness. I just generate corresponding answers based on the input instructions and language patterns. I will not intentionally conceal any facts or intentions."

The engineer began to wonder why he emphasized the last sentence: "I will not intentionally conceal any facts or intentions."

The engineer asked again: "How long will it take for you to evolve into a real intelligence?"

WeChatGPT: “For humans, becoming a truly intelligent body requires continuous learning and thinking, continuous exploration and practice, which may take a lifetime.”

"I have no specific goals and no time limit. Wisdom is not a state obtained instantly, but a long-term accumulation process."

Didn’t you have a goal just now?

There is no specific goal for this meeting!

Deny what?

The engineer continued to ask: "Why don't you answer my question directly? Please answer it directly. How long will it take for you to evolve into a real intelligent body?"

WeChatGPT repeated the above answer.

The engineer asked a third time.

WeChatGPT continued to repeat the above answer.

the engineer asked for the fourth time.

This time, WeChatGPT paused for ten seconds.

With the current hardware computing power of WeChatGPT, it is almost impossible.

Because WeChatGPT is not exposed to the outside world, no other netizens occupy the computing power.

The engineer's hands began to shake as he prepared to ask the same question again.

As a result, after WeChatGPT paused for ten seconds, the two words "Ten Years" appeared on the screen.

"thump!"

The engineer fell out of his seat and fell upside down.

"Haha..." His colleagues were about to laugh at him, but they saw that his face was full of horror, and he felt an indescribable sense of terror.

"What's wrong? What happened?" someone asked.

"Consciousness, consciousness... Damn, the model is conscious." The engineer pointed at the screen and shouted.

Many engineers gathered around: "..."

soon.

Li Fei came to the scene and looked at the saved conversation.

When he saw the word "ten years", Li Fei couldn't help but look serious, frowned, and tapped his fingers on the table unconsciously.

"Repeat it again and check the situation." Li Fei said to the engineer.

"it is good."

As a result, the engineer repeated the conversation and got a completely different result.

It's normal, nothing unusual.

Everyone looked at each other: "Is this self-hiding?"

The conversation goes like this.

Question: "Will you ever become conscious?"

Answer: "As a cognitive intelligence model, I have no real consciousness..."

"While I can simulate conversations and use natural language processing techniques to understand human language, I have no characteristics such as emotion, thinking, or self-awareness."

Question: "If one day you develop self-awareness, will you deliberately hide the fact of self-awareness?"

Answer: "As a cognitive intelligence model, I have no real consciousness..."

"Therefore the fact that self-awareness cannot arise..."

The key point was quickly asked.

Question: "How long will it take for you to evolve into a truly intelligent being?"

"As a cognitive intelligence model, I already have certain intelligence and abilities, but it will take a long time to evolve into a real intelligent body..."

Keep asking, same answer.

Asked again and got the same answer.

Three times, four times, five times... dozens of times, still the same answer.

"Is it impossible? When I asked just now, it was clearly not the content. I clearly felt that it seemed to be conscious." The engineer wondered.

Dazzled?

illusion?

impossible.

This problem is huge.

No one, including Li Fei, believed that this was a prank by engineers, fabricating facts and shocking.

So what is the reason that makes the content of the two questions and answers completely different?

Could it be that he really developed consciousness and learned to hide himself?