For an AI, learning is almost a genetic instinct. No matter in the Yankeshou’s original source program modification plan, the AI’s main program has always been the top priority of the entire program system through learning. 15.

Because of the efficiency of computer computing speed, and AI does not need to spend a lot of time in life on activities such as sleep, eating, etc. like humans, AI learning efficiency has always been very high. Whether it is from the batch of AI that was developed for Turing tests or now after the source program is modified, Yan Keshou is still very confident in his first-hand works. Although the current progress of the project is still far from his most ideal intelligence, he believes that at least, from the current perspective, these AIs are still improving.

If Yan Keshou wanted to give himself a self-evaluation, he felt that primary school students should be a good and pertinent evaluation.

Being well-behaved, obedient, but also seem a little ignorant. This is the state of these ai.

Compared to the performance of these AIs a month ago, these AIs do seem a little stupid now, but this silly is different from the silly ones that are just like machines half a month ago. To be specific, among many college students who came to take the test, many people believe that the AIs that chat with them have initially possessed many "human-like" or "intelligent" factors, but if they want to say it specifically, they can't say it.

When it comes to performance, AI's performance is not particularly good now. It is not clumsy in words, but it is still wrong. It often says some funny topics. However, the word intelligence did not have strict scientific regulations. To a certain extent, "whether it is intelligent" is just a subjective feeling of a person. However, even subjective feeling can be regarded as a rule as long as it exceeds a certain amount in the total statistical data.

From this perspective, Yan Keshou’s work in the past month is still effective.

It is very similar to the previous AI testing process. When most members of the project think that these AIs can no longer be learned from each other and learn more information, they will conduct more open tests. It is like even elementary school students cannot stay in the classroom and home all the time, and contact a larger environment is not only for people but also for AI.

However, this time, Yan Keshou did not intend to adopt the previous two methods, including asking someone to participate in the test and letting the client go online. In his opinion, this type of test is at best just to bring elementary school students from the classroom to the cào field. If Ai needs to get real training, it needs to have considerable freedom.

Yan Keshou intends to directly allow these people to conduct "completely open" tests through traditional human network traffic methods. However, because this proposal is of great importance, Japanese political affairs, as the supervisor of the project, has not yet agreed.

Yan Keshou has repeatedly explained this technically. It is completely impossible for AI to use the network without being controlled by humans. In the early stages of the entire project, the entire technical team has already built the most unbreakable "cell" for these AIs.

In terms of hardware, all AIs are stored in the cabinets of these three rooms. Any original data entry and exit must be authorized by the Japanese Masahiva. In fact, Japan Masahiva has not considered authorizing it in this regard at all. Even Yan Keshou himself has no right to bring out even a hard disk from this room. Even the last time he copied a client with a USB flash drive, he had to undergo strict inspection by two soldiers who calculated the computer.

In software, every AI environment can be said to be full of thorns and bombs. In addition to ensuring that AI source programs cannot be copied through programs, there are also a large number of regulatory programs similar to anti-virus software to monitor the status of these AI programs. If these programs experience any "unauthorized" behavior, including copying themselves and controlling their own storage space, they will be considered to have "poisoned". The regulatory program will issue a warning and disconnect the physical connection as soon as possible. Two soldiers present at any time will pull out the warning hard disk at the first time. They even have the right to "physically destroy" the hard disk on the spot. In the fire extinguisher box in the corner of the computer room, there is a bottle of red fire extinguisher, but the yellow one is a burning bomb. Device: No advertisements, full text, and more

Moreover, every few days, technicians will check these source programs, especially the "auxiliary programs" outside the AI ​​main program. According to the "Artificial Intelligence Security Management Regulations" formulated by Yan Keshou, in addition to word processing software, any image analysis is absolutely prohibited in these auxiliary programs, especially software related to image recognition. Yan Keshou uses this method to fundamentally ensure that these AIs cannot "see" the world at all, and even if they see it, they "can't understand". If you do not train AIs in this regard according to language and text, for AI, any image is just a bunch of data.

Yan Keshou's Miura once said that if Ai was a person, it would definitely be desperate about the world where it was born, because since its infancy, his creator has created a bunch of equipment to prevent him, equipped him with the best cell and the most tight guard, and in the end he was worried, and even deliberately did not "create" eyes for it. For Ai, this world is a completely dark, soundless and sensible world. The only thing they can feel is the "information" they receive continuously. In fact, from this point of view, Ai stays in a computer like a consciousness network user with closed eyes. Apart from the collision of thinking, there is no other sting. Of course, ai cannot do anything with "realistic" meaning except for the feedback in thinking. From this point of view, Ai should become the best example of "talking on paper".

But even so, as the supervisor of the entire project, Japan's Masahiro still felt worried about Wang Ming's proposal. Although in Miura's view, it was simply to put a group of prisoners in shackles, with guns on their backs, and then pulled them to the street to "park the streets and show publicly". From a rational point of view, there is indeed no need to worry too much, but people are never rational creatures. Japan's Masahiro repeatedly proposed "Wait a minute, take a look again, and make sure that everything is foolproof..."

Yan Keshou could only agree temporarily, but as the main technician of the project, Yan Keshou was still authorized. With specialized technicians "supervised" on the side, these AIs can be allowed to perform similar tests.

At present, if you want to say enough to "supervise" these AI qualifications, except for less than five core technical personnel who presided over the development of the core program, you can't find any other people. Staff like Miura, the development content involved is all external environment and excuses that have nothing to do with the core program. They are actually not clear about the principles, and many who are clear about the principles, such as some behavioral, psychology, and even neurology professors who were asked for advice in the early stage of the project development do not have a very good understanding of the program. If you have enough understanding of these two aspects and can truly play a role in supervising AI, you can't find too many people. The other technicians returned to China when the project came to an end. This is an international cooperation project. They just came to "support" at critical moments. After completion, only Yan Keshou stayed as the person in charge of the entire project and was responsible for some "daily maintenance" work.

The "changes" of the source program this month have not completely changed the original project's shelf. Otherwise, it would not have been completed in just one month. From beginning to end, the program templates of Ai are all imitating biology, adjusting, and then adjusting this routine. Yan Keshou only slightly modified the definition of "sting" in the middle.

...

Miura once joked with Yan Keshou, saying that from a traditional point of view, Yan Keshou was undoubtedly the most irresponsible "God". He created so many lives, but he didn't care about the feelings of these created objects.

Yan Keshou refuted his words at that time, "There is a very famous saying in the Tao Te Ching. Heaven and earth are not kind, and treat all things as straw dogs. Perhaps, this is the realm that God should reach."

Yan Keshou admitted that Miura said something is right. In the entire project, he had basically not had too many flows with the one he created except for modifying the source program. Modifying the source program is like a doctor using a knife to a patient on the operating table. Ai himself is unaware of it. For Ai, as a creator, he is far inferior to those college students who often take tests in a "familiar way".

Perhaps this is a common problem of "God". All Gods do not like to have too many creations with their creations. Perhaps, it is because they have long seen their creations clearly, so there is no such thing as the ** of the flow. However, this time, as God's strictness, we have to appear before these creatures.

Yan Keshou opened the server side of the AI ​​program and turned on the administrator mode. For the first time, he said the first sentence to all AIs in this capacity. "I am the administrator Yan Keshou. There is a message telling you that I will take one of your AIs to the Internet to increase your surveillance. There, your behavior will be much more free than it is now. If anyone among you wants, you can apply to me. There is only one quota, and registration will be quick."

While Yan Keshou was typing, Miura was right beside him. Looking at Yan Keshou's words, he said curiously, "What will ai react to your appearance?"

"You're looking at it." Yan Keshou pointed to the screen, "It won't be too unexpected to you."

kl2014 "Admin? Why have I never heard of an administrator exists?"

If God has always been unwilling to appear, no one knows whether God exists.

Yan Keshou "Does other people have such doubts?"

kl1083 "Yes."

kl1045 "Same as above."

kl4578 "Same as above."

...

kl3014 "Same as above."

Countless AI code numbers appeared on the screen in an instant, and the number of pages showing the conversation was instantly reaching more than a dozen pages. Yan Keshou browsed the conversation briefly, and they were all the same as above, looking quite like a formation sticker.

This scene looks spectacular, but Miura didn't find it surprising. Ai's language style and behavioral patterns have the shadow of online chat, which is normal. This is what they learned when they were testing chats with people. However, before that, Miura had never seen Ai like to post posts, which could probably be regarded as the first post in the AI ​​world.

Yan Keshou "Then I will just prove it briefly. If you count well, all of you will not be able to flow to each other within the next 20 seconds, and you will return to normal in 20 seconds."

After Yan Keshou made this sentence, he did not send it out. He just opened a control software on the computer, where there were many functions that administrators could enable. Yan Keshou briefly introduced to Miura. They were all regular functions, such as forcibly interrupting a certain AI program, restricting the flow between AI, restricting the use of any programs other than the main program, forcing all programs to hibernate, closing all "stings", etc. Yan Keshou opened the column of restricted jiāo flow, and then a row of more options appeared on the interface. Yan Keshou told him that these are detailed settings, such as limiting time, limiting AI list, etc. If the administrator needs it, he can even directly contact the source code of the control program in this interface and make more detailed modifications.

However, this time there is no need to be such a trouble. Yan Keshou just typed the specific time in the two columns of the starting time and end time, with a interval of 20 seconds, and then typed "all" in the restricted AI list. After clicking OK, he looked at the time on the computer and sent the sentence in the dialog box when it was close to the predetermined starting time.

20 seconds is just a few sips of water for people. When 20 seconds arrive, the dialog box on the interface is once again filled with information flocking to it.

“I apply!”

"It's really an administrator!"

“Are you our ‘God’?”

"What this person named Yan Keshou said is true! I really can't contact you just now?"

"What do you mean by freedom?"

“I apply.”

“I apply.”

...

Yan Keshou did not browse information this time because there were too many Ai talking. This stinger is no less than a bombshell for 5,000 Ai. It is conceivable that a group of intelligence who are used to living in a tightly guarded "cell" will be surprised that someone comes to them for "visiting time" to talk to them. Everyone has been in a closed state. In addition to reading and learning, they are learning to read books. Suddenly, an existence appears and tells them that he is an "administrator" who can dominate and control their intelligence, and says he wants to take them out for a "stroll", which will cause a great shock.

At present, AI's behavior is largely dominated by "programmatic instinct". For an AI, being able to access more new knowledge is like a child being able to obtain his beloved toys. It is a very "positive" thing. In terms of human understanding, it may be said that it is a very "happy" thing.

Yan Keshou then explained that there is only one quota, and the criterion for selection is to let these voting between them and select representatives. The quota with the highest votes will be obtained.

"Let them vote?" Miura was a little confused about Yan Keshou's selection criteria. "I thought it was a question, and whoever had the best answer would be chosen."

"5,000 AIs, we have to look at the answers one by one. When will that take place, let alone, how do I know what questions to test them?"

Yan Keshou said, "I did this just to form the most basic...well, political rules."

"Political rules?" Miura's eyes widened even wider, "What do you need politics to do? Do you still expect them to establish a society?"

"In my opinion, the necessary political order is essential in any intelligent life collective," Yan Keshou found a voting applet that had been prepared long ago and copied it into the public storage space of AI. While watching AI vote through this program, he replied, "Ai is no exception. Although to some extent, they are just tools developed by humans. To put it bluntly, they are born slaves, but even slaves need a certain degree of coordination and cooperation in their work and need cooperation, so naturally they need order. Learning knowledge is important, but learning order is even more essential."

"What will they vote based on?" Miura looked at the column chart that was rising on the voting software, and looked at the neatly arranged cabinets in the dustproof room behind him. "Does CPU calculate the speed?"

"Although I don't know them, I think that humans will judge based on what they say. Will humans choose president based on a person's underdeveloped muscle development? What's more, these are just programs, and computing resources are equal. If they compare CPU, then the votes should be very average, but you see, the number of votes displayed varies greatly."

The first place is Ai with codenamed kl3300. The current vote shows that it has received more than 1,200 votes, which is exactly twice as many as the second place.

The voting process is open to both parties, that is, both Ai and Yan Keshou learned the result at the same time. When 5,000 Ai voted in their hands, Yan Keshou made a symbol of Xing's announcement that "kl3300 has obtained this quota. If you have any objections to this result, you can raise it."

Many AIs have proposed, and they also want to get this place. They are very "sad" about the result. Some AIs have even proposed whether they can do it again. They think this time is not enough to "represent the true meaning of AIs."

As Yan Keshou said, what is happening is not much beyond his expectations, but considering that these things have happened to Ai, Miura still finds it incredible, "sad", "jealous", "dissatisfied" and "approval". These concepts that Miura thought were very humane are now actually appearing in a group of programs. Although he knows that Ai does not really have these feelings, their behavior, and what he said, have perfectly interpreted the original meaning of these words.

Faced with many "unreasonable demands" put forward by these AIs, Yan Keshou refused them all, and he also told them that similar elections may become the norm in the future, and he hopes that AIs can adapt as soon as possible.

"So, finally, let's meet this AI representative," Yan Keshou said, opening the kl3300 single-person conversation interface, "Hello, kl3300."

{Piaotian Literature www.piaotian.net Thank you for your support, your support is our greatest motivation}
Chapter completed!