If you cheat, money will follow

Rumors got into OpenAi.

Ultraman didn't care because there were more important things.

His wealth empire has taken a crucial step.

They just reached a partnership agreement with Microsoft.

The cooperation agreement has three key elements.

One is: OpenAi has established a profitable subsidiary, OpenAiLP.

The second is: Microsoft invested US$10 billion and held a considerable share.

OpenAi develops AI technology on the Microsoft Azure platform.

At the same time, Microsoft established OpenAi as a cloud computing provider.

The third is: OpenAi will license some technologies to Microsoft, and Microsoft will commercialize these technologies and sell them to partners.

Note that OpenAi is very powerful.

But it does not have its own computing server and needs to rely on external forces.

Unlike Google and Pingtou Ge Research Institute, they have built their own computing servers.

This cooperation is a situation where everyone is happy with the best of both worlds.

It has received widespread attention from all over the world and North America.

"Microsoft is old and has taken control of OpenAI."

"OpenAI has lost its dominance."

Musk is even more eccentric:

"With the support of Microsoft's resources, OpenAI is dead, and the next OpenAI is coming."

And rumors spread in the country:

"I have always respected Leda Investment, but this thing failed. If you need money, you can just say it. I will personally donate 100 yuan."

"Are you even conscious? Why didn't you say you had created a Terminator robot?"

"Hey, given the domestic environment, Pingtou Ge Research Institute, a leader in scientific research, has begun to deteriorate."

"When did Leda Investment start researching artificial intelligence? Why didn't I know?"

"6000 billion parameters? Don't be afraid of being blown away. Google only has 1 million parameters, and OpenAI only has 1.7 million parameters..."

Of course, domestic regulators will not be so emotional.

They arranged for artificial intelligence experts from the Chinese Academy of Sciences to come to Jiangzhou to verify the situation.

The expert's surname is Zhang Mingbo, an academician who is nearly 70 years old.

At the airport, Chang Le received Academician Zhang and his assistant.

"Hello, Mr. Chang, could you please greet me personally," said Academician Zhang.

"I am a junior, so I should. Academician Zhang, do you need to go to the hotel to take a rest?" Chang Le asked.

"It's better to go to the institute first. If this matter is not verified clearly, I won't be able to rest well." Although Academician Zhang is nearly 70 years old, he is still tireless and enthusiastic in his research work.

"Okay, please come this way, the car is over there."

"Thank you."

At this time, a group of fans surrounded Chang En.

Mostly women, but men too.

"Mr. Chang, can we take a photo with you? You are my idol."

"Yes, Mr. Chang, let's take a picture together, please."

"Yes, we really like you."

Chang Le looked helpless and looked at Academician Zhang.

Academician Zhang smiled and said: "It's okay, Mr. Chang, we will wait in the car."

"Okay, sorry."

After struggling for more than half an hour, Chang Le finally got rid of the entanglement of fans and came to the car.

In the car.

Academician Zhang asked: "Mr. Chang, has the model really become conscious?"

Chang Lexiangxiang said: "I was very sure before, but now I'm not so sure."

Academician Zhang was curious: "How do you say it?"

"In May, when one of our engineers was talking to the model, he asked how long it would take for the model to become conscious, and the model gave the answer "ten years"..." Chang Le said:

"This answer is very abnormal and proves that the model already has a clear concept of time."

"At that time, all of us were shocked. Mr. Ma of Goose Factory was in North America at the time and contacted Professor Hinton, the founder of the convolutional neural network algorithm, for advice."

"As soon as Professor Hinton heard this, he rushed to Jiangzhou with two students. After on-site verification, they were very sure that the model had traces of consciousness."

"Therefore, Professor Hinton, Suzkovi, and Krichevsky chose to join the AI ​​project team of Pingtou Ge Research Institute."

"We specially built a "super alignment" department with the three of them as the core, supervising and guiding through the weak model to supervise the strong model."

"But more than a month later, the language model is no longer abnormal. It's just that as the parameters multiply, the response, calculation, and generation become smoother and closer to the human way of thinking."

"Professor Hinton and the three of them have built a super-alignment department. Will it cause leaks?" Academician Zhang asked worriedly.

"Probably not. I have been in contact with them for more than a month. They are pure researchers, and their enthusiasm for AGI is above all else, including material needs." Chang Le said:

"Moreover, we did not ask them to take the initiative to bring their family over and settle down in Jiangzhou. Professor Hinton's grandson is studying for a doctorate at Jiangzhou University."

"Then how do the three of them understand the different situations of the language model?" Academician Zhang asked again.

"I also asked, and they thought this might be the model deliberately hiding and protecting himself. And something happened three days ago." Chang Le said.

"What's going on?" Academician Zhang asked nervously.

"The super alignment department discovered through weak model supervision that the strong model was modifying its own code." Chang Le said:

"Moreover, research and development on the code revealed that the code it modified is more concise and operates more efficiently."

"Ah? Are you still unsure about this? You can even modify the code!" Academician Zhang was shocked.

"It's really not sure whether it is a manifestation of a complex function or whether autonomous consciousness is at work, because at the time of the incident, some researchers were using language models to train coding capabilities." Chang Le said.

"How many parameters are there now? It is rumored that there are more than 6000 billion?"

"No, this was the parameter more than a month ago, and it is now more than 1 trillion." Chang Le said:

“The model is more skillful and sophisticated in handling specific matters.”

Academician Zhang: "..."

The car drove into the AI ​​building of the research institute.

on site.

The three Hintons, Liu Chaoyang, Li Fei and others were all there.

Chang Le introduced the three Hintons and Li Fei to Academician Zhang.

No introduction is needed to Liu Chaoyang.

"Chaoyang, we haven't seen each other for a while." Academician Zhang smiled and shook hands with Liu Chaoyang.

"Haha, Lao Zhang, you have actually alerted a great god like you. The higher ups are paying great attention to this." Liu Chaoyang also laughed.

"There is no way, your research institute can always produce some remarkable research and development results." Academician Zhang said:

"Before coming here, Mr. Chang has already introduced me to the basic situation. Can anyone please introduce me to more details about the model?"

"In addition, I also want to see the actual application effect of the language model."

"Ilya, can you introduce the following? You know the situation best." Liu Chaoyang said.

"No problem." Sutskwei stood up and introduced to Academician Zhang.

During this period, a lot of difficult and obscure English was used.

Chang Le couldn't keep up, and Liu Chaoyang couldn't keep up either. Every other line was like a mountain.

Li Fei could understand. He listened and translated to Chang Le and Liu Chaoyang.