Riding the wind of rebirth

“How could we possibly bypass the Chinese Chess Academy when developing Go software, especially Go software capable of challenging top-level Go players?” Zhou Zhi countered. “The leaders of the Chess Academy strongly support our ideas, although they believe that it is basically unrealistic for Xiao Zhi to have the ability to play against them.”

"What was the answer the chess master gave you? He wouldn't say that he plays chess based on intuition, would he?" Yang Honghui laughed.

“Hey! You’re right, it’s all based on intuition,” Zhou Zhi said. “However, this kind of intuition is an intuitive behavior that comes after systematic training.”

"And it's not just him who said that; several ninth-dan players said they acted based on intuition, and they couldn't explain the specific reasons." Zhou Zhi became proud as he said this: "However, our Xiao Miao used mathematical language to roughly explain the formation process of this way of thinking."

"You mean, Xiao Miao can use a computer to simulate human thought?" Liang Hong was startled: "Your Xiao Zhi can already think like a human?"

"Holy crap! Terminator?!" Yang Honghui was immediately reminded of that wildly popular blockbuster from a few years ago.

“It’s not that exaggerated,” Zhou Zhi said with a smile. “To be precise, it’s about training Xiao Zhi’s neural network to a certain extent, and then learning from a large number of Go games to predict the next decision that humans will make, and determining the probability of the positions that humans might make in the first five or ten moves.”

"This involves two neural network systems, one of which is a decision network system. Using millions of online Go games, through supervised learning, we enabled Xiaozhi to develop decision-making abilities, allowing it to simulate human Go-playing behavior."

“Choose any point on the chessboard and train the system to predict the next decision a human will make; the system’s input is the top five or top ten most likely moves at that particular position; in this way, you only need to look at the most recent five to ten possibilities instead of analyzing all two hundred possibilities, which greatly reduces the amount of computation.”

"Once Ash learns this, we can introduce error reinforcement learning and conduct millions of real-world training sessions on the system. When it wins, it will make Ash realize that it can make similar decisions in similar situations next time. Conversely, if it loses, it will not choose this move next time a similar situation arises."

"After forming the decision-making neural network system, we can begin to build the second neural network system - the numerical system."

"If the decision-making system is to form 'intuition,' then the numerical system is to quantify 'intuition.' Through millions of games, the system is trained, allowing Xiaozhi to learn about different placement points through confidence intervals, defining the winning probability between 0 and 1, where 0 means it is impossible to win and 1 means it is 100% guaranteed to win. Then Xiaozhi calculates the probability."

"The final result is that by combining these two neural networks, we can roughly predict the current situation. This is actually very similar to the process of a master learning Go - through rigorous training, such as studying game records and analyzing local moves, we thoroughly study the optimal moves in the next few steps, and then form a kind of 'intuitive decision-making' in the human brain."

"However, the human brain's ability to process information is exaggerated, and the situations it faces are also very diverse. Even if Xiao Zhi eventually reaches the level of a ninth-dan Go player, that is only one of the millions of abilities that humans possess. The exaggerated information processing ability of the robots in Terminator, which is similar to that of humans, can only be a figment of imagination."

"However, we also have to say it the other way around. Although Xiaozhi's abilities cannot be compared with the human brain, his hardware architecture, software system, and algorithmic thinking represent a major advancement in human technology. They are epoch-making technological ideas that far surpass traditional computing solutions like Deep Blue and Galaxy. I think their importance is roughly equivalent to the division in physics before and after Newton."

"Wow? So this little seedling is a living Newton?" Liang Hong was dumbfounded.

"No, no," Mai Xiaomiao waved her hands repeatedly, startled. "Elbow was talking about the importance of the neuron-like technology solution, but I didn't invent this technology, nor did I invent the mathematical model. We just implemented them in the system design." "That's still impressive," Yang Honghui nodded. "You've talked for so long without mentioning any achievements. Have you made any achievements yet?"

"We do have a record, but the people at the chess academy are still debating whether it counts as a record." Zhou Zhi and Mai Xiaomiao couldn't help but laugh when they talked about this.

"What's wrong? There must be something strange, right?" Liang Hong could tell from Zhou Zhi's expression that this was not a simple matter.

"If we judge solely by results, we can now conclude that Xiaozhi, the Chess Saint, and several top-tier players such as Chang Hao, Ma Xiaochun, and Yu Bin have a higher win rate than loss."

"Then why is it said that it doesn't count as a result? Is it really that difficult to determine the outcome of a Go game?"

“Our system doesn’t solely consider winning or losing.” Zhou Zhi shook his head: “The main problem is that when playing against skilled players, Xiao Zhi often has a ‘sudden inspiration’ and will place a piece on the fifth or even sixth rank early in the game.”

"Is there anything strange about this?" Liang Hong didn't know much about Go.

"In fact, there is an important dividing line in Go, which is the third line from the edge to the center. Placing a stone on this line means using the terrain of the edges and corners to gain territory in the fewest moves. It is like fighting for the world. First, you need to conquer a piece of territory with the least cost. After you have formed a stable 'base', you can then send troops to fight for the central plains."

"The way of Go from ancient times to the present cannot escape this principle. If you try to break out of the central region before you have secured the border areas, you are very likely to become an isolated force and fall into a situation where you are surrounded by enemies."

"Therefore, in the numerous chess games over the past three thousand years, people have come to the conclusion that placing a piece on the third or fourth line means that you can comfortably 'occupy the side,' which is the optimal and safest solution."

"If you cut into the center too early, it means that you are giving up the initiative in the process of 'occupying the sides'. Although it will bring more changes to the game, the center is a battleground and you need to deal with the 'encirclement' from all directions. This allows the opponent to continue to strengthen the four sides while encircling you, and expand the initiative advantage more and more."

"Xiao Zhi's moves often involve placing stones too early on the fifth or sixth rank, advancing into the central area of ​​the board. This goes against the logic of playing chess that has been around for thousands of years. In the eyes of top-level players, this is undoubtedly a suicidal move, an 'unreasonable move'."

“What does ‘unreasonable hand’ mean?” Liang Hong asked.

“It’s just bad play,” Yang Honghui said. Although he didn’t know much about Go, he was familiar with some of the professional terms thanks to the popularity of the game and the broadcast of various tournaments. “It’s like the saying in martial arts, ‘A chaotic punch can kill a master.’ A chaotic punch is a irrational move.”

"But that doesn't make sense either. If Xiaozhi loses because of this, then it can be called an 'unreasonable move,' right?" Yang Honghui asked. "But if he wins, it's like two people fighting, one a novice and the other a pro. Then the pro gets beaten black and blue by the novice and then blames the opponent for not following the rules and for being bad at fighting. That doesn't make sense either, does it?" (End of Chapter)