1991 Starting from the core

In the Pandora reception room, Xi Xiaoding was chatting and laughing with Director Liu of the Computer Department, Director Zeng of the Development Zone Management Committee, and leaders of the Provincial City Science and Technology Bureau and the Science and Technology Association.

In fact, Mr. Xi doesn't have much dealings with leaders. He usually participates in more academic conferences.

But this time Su Yuanshan made such a big fuss. As the CTO of Yuanxin Group, the chief scientist of the Science and Technology Committee of Yuanxin Group, and the leader of the artificial intelligence subject of Chengdu... he had to take time to chat with the leaders in front of the camera. .

"Dr. Xi, we all have seen the development and advantages of Yuanxin in the computer field, but when it comes to the field of artificial intelligence, many people may not understand it accurately. For example, what are the technical principles of Blue Star and Deep Blue, and what are the differences between the two? What’s the same and what’s different.”

Director Liu was a computer undergraduate student in the early 1970s. It cannot be said that he has a thorough understanding of the cutting-edge technologies of computer science, but he at least understands the basic principles. With these words, he obviously hopes that Xi Xiaoding will do a simple science popularization so that he can put it in the documentary.

Xi Xiaoding nodded and said with a smile: "Artificial intelligence, to put it simply, is actually letting computer programs simulate human behavior. And behind the behavior is the way of human thinking."

"As far as Blue Star and Deep Blue are competing this time, the first thing we compare is learning ability - to see who can learn more 'experience' from past chess games, and the second is the duel."

At this time, the reporter asked curiously: "You mean, the number of moves considered in playing chess is secondary?"

"In a sense, it is like this." Xi Xiaoding nodded with a smile: "The efficiency of calculation is closely related to the hardware equipment. For example, our Blue Star can only calculate after ten steps at the beginning. After upgrading and debugging, It can be counted to eleven steps - but Deep Blue can be counted to twelve steps."

"Why does Deep Blue take one more step than Blue Star? Because Deep Blue uses a specially designed supercomputer, while Blue Star uses a general-purpose supercomputer."

"So, for this time, you will see that Blue Star and Deep Blue will not perform better than before. The reason is that they have left an environment that can fully exert their advantages."

"So, the focus of our competition is actually on 'learning', and machine learning is a long process, especially our BluGirl framework. The characteristics of its convolutional neural network framework determine that the more times she learns, , the better the training effect will be - the comparison is the accuracy and efficiency of the respective frameworks."

"And in terms of learning capabilities, our framework...is theoretically better."

Xi Xiaoding smiled and made a note of this competition.

…

Three hours later, the host of the education technology channel was already a little thirsty.

An hour ago, a new guest host came to the live broadcast room.

Gu Xiaochun, a domestic chess champion.

It was also the target of Blue Star's defeat in World War I and victory in World War II.

As the only top professional chess player in China who has competed against Blue Star, Gu Xiaochun has an understanding of Blue Star and even this kind of chess AI that no one else can match.

"As far as the two times I played with Blue Star, the first time, Blue Star was very aggressive, like a talented chess player with great chess skills but not very mature. But the second time, Blue Star was very stable. too much."

Gu Xiaochun didn't mind talking about the time he lost at all - after all, even the world champion and the card god lost, so what else did he have to worry about?

"During this period, I also learned about Blue Star's learning methods. I have to say that computer technology and machine learning are developing too fast." Gu Xiaochun sighed and said with a smile: "Fortunately, artificial intelligence does not participate, otherwise It was a massacre."

At this time, Professor Xia Yanbing smiled and added: "Let me add, Teacher Gu."

"According to the available information, Deep Blue does not use machine learning in the true sense. It is said that they have input more than two million chess records before - in short, Deep Blue is a stupid student, and the method used is Practice makes perfect. Yuanxin’s Blue Star uses the analysis of a small number of chess records for in-depth analysis – this is true machine learning, and it is about drawing inferences from one example.”

The host immediately said: "You mean Blue Star has a high winning rate this time?"

"No, theoretically Blue Star is better at this kind of limited-time duel - you see, now Blue Star's indicators are better than Deep Blue..."

Just when Professor Xia wanted to continue the introduction, the screen suddenly jumped, and he was immediately excited: "It's over."

The host was shocked and obviously a little shocked: "The game is over?"

Gu Xiaochun was also startled for a moment - the chess pieces didn't even move, how could this be considered the end of the game?

"No, Blue Star has finished studying. It seems that they did not let go to take care of the competition time - oh, Deep Blue has also finished." Professor Xia said excitedly: "In comparison, Blue Star's The framework is better.”

The host complimented him at the right moment: "How good is he?"

Professor Xia has dealt with Mr. Xi and knows that although Mr. Xi has a high and powerful position, his aloof temper is buried deep in his bones. However, his identity determines that he cannot speak arbitrarily.

So Professor Xia chuckled: "It's hard to say, but as far as I know, at least in the field of machine learning, Pandora Labs is top-notch."

As the voice fell, the live host also appeared on the remote screen. The host held up the microphone and briefly introduced the previous learning results.

Of course, she used the term "equal".

Subsequently, the "tool man" responsible for the placement of the pieces from both teams came to the front of the chessboard.

Gu Xiaochun was stunned and began to focus his attention on the chessboard.

Following the referee's signal, IBM's dark blue and white team went first, and E4 popped up first on the monitor interface.

After thinking for less than two seconds, Blue Star followed E5 directly.

Immediately afterwards, Shenlan walked out of Nf3.

"Deep Blue uses Italy to start, and then it depends on how Blue Star responds."

Gu Xiaochun stared at the chessboard intently and introduced the opening carefully.

This time, he had a little selfish intention - chess is not well-known in China and has a small audience. If chess can be popularized through this live broadcast... As the domestic chess champion, Gu Xiaochun naturally It's something you can only hope for.

After thinking for a few seconds, Yuanxin's Blue Star gave Nf6's response.

Gu Xiaochun praised: "Nice move."

"Italy's start is considered a classic start, but the two-horse defense is also a classic response, and the subsequent changes are extremely complicated, which will test the chess skills of both sides - oh, it should be the calculation power?"

…

Su Yuanshan had already returned to the reception hall, sitting in front of the TV and watching the live broadcast.

When he heard the fashionable word "computing power" come out of Gu Xiaochun's mouth, he smiled slightly and turned to look at the Yuanxin people behind him who had come to join in the fun.

“What we are fighting for is not computing power, but algorithms.”

As Gu Xiaochun explained, the game between the two sides continued.

Obviously, these two servers cannot reflect all the "capabilities" of the two AIs. The further they go, the longer they take to think about.

Gu Xiaochun frowned slightly and said hesitantly: "It seems that the chess skills of these two AIs have declined?"

Professor Xia immediately explained: "Yes, the chess power of AI is closely related to computing power, and computing power is directly reflected in the hardware. At the beginning, the "homes" of Blue Star and Deep Blue were supercomputers, so they could Sees further than a human chess player can.”

"Oh, no wonder." Gu Xiaochun smiled and nodded.

After thinking for a full minute, DeepBlu made a move of Qxg7.

Gu Xiaochun immediately began to think about it - as a person who has experienced it, he naturally knows that every step of AI has profound meaning.

But this step... is obviously wrong.

Just as he was thinking, Blustar directly gave a move of Qd6.

…

Pandora Labs,

Unlike Gu Xiaochun in the live broadcast room who relies on his own chess skills to analyze and judge the chess game, the members of both sides of the game are at best half-assed in the field of chess.

At this time, their analysis and understanding of the chess game comes from the winning rate feedback given by the AI ​​​​of both sides after each move.

——As emotionless competition machines, the original drive of Blue Star and Deep Blue is to win. And every step they take, they predict and analyze from the optimal choice.

John Wick saw that after his opponent gave Qd6, the other team smiled in unison.

As the person in charge of the DeepBlu project, he has communicated with this team many times and knows that the other party is slightly ahead in learning, but in terms of logical analysis... they are not strong enough to surpass his own team.

After withdrawing the realization, John Wick and his team looked at each other, staring closely at the dark blue interface.

In the feedback given by Deep Blue backstage, our winning rate still maintains a 52% lead.

This time, Shenlan thought for five minutes.

Only then was the move Ba6 given, and the winning rate given remained unchanged.

John Wick looks at each other inexplicably - so, what are they having fun with?

——Of course, he didn't see it. Just across the conference table, in Blue Star's background display, Blue Star's winning rate has reached 78%.

…

Miao Renfeng from the Pandora team had already crossed her shoulders in a leisurely manner.

I don’t know who took the lead, but the big guys on the Pandora team all gave themselves nicknames.

For example, Zhu Yonggui's name is Obai, Li Kaiyuan's name is Zhebie, Zhou Qiang's name is Li Xunhuan...

Miao Renzuo’s real name is Chen Hongbo.

This classmate Chen Hongbo is from the Department of Mathematics of P University and came to participate in the postdoctoral station program of Yuanxin.

Although he was about to leave the station, he had long been accustomed to staying in Pandora and planned to stay there for the rest of his life.

He claimed that in the field of artificial intelligence, Mr. Xi had no rivals. He even looked down upon a few "stupid" mathematicians recruited by the laboratory next door with high salaries in order to "make themselves look fat".

Regarding Chen Hongbo, Mr. Xi once gave an evaluation - "When he plays with the numbers, he also dares to compete with others in pure numbers..."

Of course, Mr. Xi's comments cannot stop Chen Hongbo's arrogance.

He held out five fingers to John Wick.

this means……

Within five steps, he would kill the opponent's queen.