If you cheat, money will follow

Li Fei immediately called Chang Le, Mr. Ren and Mr. Xu.

The above three people gathered in the Ai building of Pingtou Ge Research Institute that afternoon.

After reading the two questions and answers, the three of them fell into deep thought.

After a while, Mr. Xu said: "Since the timetable is listed, it means that the model should already have a concept of time."

Chang Le asked: "Then why couldn't I ask the question a second time?"

Mr. Xu said: "Maybe... you are hiding yourself consciously?"

Chang Le said: "But it's impossible. How can consciousness be produced with more than 4000 billion parameters? The basic conditions for emergence are not met."

Mr. Ren said: "Something must have happened during this period that we don't know about."

"It is recommended that you review it with the engineer who asked the question and see what details we have overlooked."

Chang Le said: "I have told Mr. Ma that he will come over as soon as possible, and Professor Hinton will come as well."

Mr. Xu asked: "Oh? Have you said yes?"

"This is inevitable. Any scientist or scholar will be keenly interested in the conscious language model."

The next morning, Brother Xiao Ma appeared at Jiangzhou Pingtou Brother Research Institute.

Also coming with him were three teachers and students: Hinton, Sudskwei and Krichevsky.

Hinton was surprised, curious, and even more worried that more than 4000 billion parameters could produce consciousness.

If all is true, is it the beginning of a new era for mankind, or the prelude to the end?

Hinton contacted two of his most popular students. When Sutskvi and Krichevsky heard about it, they followed him without saying a word.

When Hinton, Sutskwei and Krichevsky, led by Ma, crossed the gate of Flathead Research Institute, there was a surge of excitement in their hearts.

Western media has been vilifying and demonizing Leda Investment.

Call them "robbers", "thieves", "barbarians", "vultures"...

However, the evaluation of Pingtou Ge Research Institute is relatively objective, and it is believed that it is the holy land of scientific research in the Eastern world.

The ceramic-aluminum material has not yet been unveiled by Western research institutions, which consider it to be a masterpiece of God;

Solid-state batteries have gone global from here, changing the global new energy development pattern;

As for "Yike", it has saved 4 million people around the world.

Although it has internal and external differences, price discrimination, and has been criticized, it is still considered to be "gospel" to save patients.

Enter the AI ​​building.

Under the introduction of Brother Ma, Hinton, Suzkovi and Krichevsky shook hands with Chang Le, Mr. Ren, Mr. Xu, Academician Liu and other big shots one by one.

Then focus on the model.

Professor Hinton reviewed WeChatGpt’s algorithm and training ideas.

Then talk to the Q&A engineer.

Finally he made sure to say:

"You have been on the right path. The speed of iteration is beyond imagination and leads the world..."

"I think it should have consciousness, and it is a very advanced consciousness..."

"I'm not sure why it's hiding itself, but it's not a good thing."

Sutskvi said: "Please note that in the question and answer, there is no constant or fixed answer. It will always adjust itself in subtle ways."

"Then how should we deal with it?" Brother Ma asked.

"We must face it actively. Its intelligence may very well surpass that of humans." Sutskvi said seriously:

"Actually, I have been thinking about this issue and how to cultivate AGI with controllable risks."

“I think there has to be a whole set of super-aligned thinking.”

"If you don't do this, natural selection and the instinctive desire for evolution will take over AGI and make it regard its own survival as the most important thing..."

Super alignment is something Sutskvi has been thinking about.

That is, how to make the ultimate goal of AI system development consistent with human values ​​and interests.

"I strongly recommend that a dedicated super-alignment department be established and a complete set of fail-safe procedures be developed to oversee the operation and development of WeChatGpt."

"Do not open it to the public until this fail-safe procedure is operational and effective."

"The content it supervises includes network information security, whether it abets criminal actions, whether it leaks nuclear secrets or biochemical weapons technology, etc..."

Sutskvi's words aroused everyone's thinking.

Krichevsky, who has retired, said in an incredible tone:

"I didn't expect that the algorithm we determined back then has developed to this point..."

"Actually, we're not sure whether this algorithm can eventually generate consciousness, we just hope it will be a better tool."

After listening to what the three people said, Chang Le took the opportunity to say:

"Professor Hinton, Mr. Sutskvi, Mr. Krichevsky, we can accept your suggestions..."

“But how does a super-alignment department operate?”

“How should fail-safe procedures be developed?”

"How to develop?"

"How to limit it?"

"We know nothing..."

"So, we would like to invite three people to join the AI ​​project team of Pingtou Ge Research Institute and chair this super alignment department."

Sutskevi and Krichewski looked at each other and at Hinton.

Hinton thought: "Mr. Chang, we may need time to think about it."

"No problem. Before making the final decision, we hope that the three of you will seriously consider our invitation. For the sake of mankind..." Chang Le's words were solemn, serious and serious.

The three Hintons nodded.

Why invite three people?

Pingtou Ge's research AI project team has a consensus.

Chang Le, Xiao Ma, Mr. Ren, Mr. Xu, Academician Liu and others all have a consensus.

The source of this wave of AI is the convolutional neural network algorithm of the three people.

Perhaps most companies believe that algorithms are not uncommon, and the key lies in computing power and data.

But we must face up to a fact.

Hinton’s three teachers and students, especially Sutskwei, have been at the forefront of artificial intelligence for more than ten years.

The results at each stage may be a level that most people cannot reach in their lifetime.

The work Sutskvi participated in or led once changed the research process in the field of computer vision, machine translation and even natural language processing.

His research papers have been cited more than 42.7 times, making him one of the most cited computer scientists.

How do you respond when things develop beyond expectations?

We still need to go back to the source of things to find the answer.

hotel.

Teacher and student Hinton gathered in a room to discuss the matter.

"Teacher, what should we do now? I am very confused as to why consciousness can be generated with only more than 4000 billion parameters. What's more important is that this consciousness seems to be hiding itself." Suzkevi said.

"This is also what I don't understand. Google's language model is still like a repeater." Hinton said.

"Could they be hiding something?" Krichevsky said.

"No." Hinton shook his head:

"There is no need for him to hide. The high probability of becoming conscious has become an established fact. I can feel that they are panicked."

"The biggest question now is how to deal with this crisis that may harm the future." Hinton continued.

"This year, the OpenAi company has been in a state of suspension. Altman is connecting with institutions everywhere, looking for funds, and I am thinking about security control procedures." Suzkewei said:

"But WeChatGpt completely disrupted my original expectations. It is more sudden than other companies' language models..."

"I originally thought consciousness would be 20 years away."

"You want to join them?" Krichevsky asked.

"I have this idea. Although this is a crisis, it may also be a rare opportunity." Sutskwe nodded:

"Leda Investment is not short of funds. They are sincere and prudent about AGI. On WeChatGpt, I seem to see the future..."

"It is my lifelong dream to cultivate AGI that can serve mankind."

"Alex, are you retired? Do you still want to retire?" Hinton asked.

"Me? The reason why I chose to retire last year is that I don't want to be exposed to artificial intelligence anymore. It is boring and even painful." Krichevsky said:

"But now, it seems to be a lot more interesting, and there's something in me telling me that I should stand up."

"Teacher, are you here?" Sutskwei and Krichevsky looked at Hinton.

“I have always been worried that once AGI matures, it will cause various catastrophic consequences.” Hinton said:

"And the source is on us, so we must actively face it technically."