Riding the wind of rebirth

Big Blue Horse is the lead horse and is very experienced in leading the team. The horses that Old Chi Ri found for everyone were also carefully selected. Under the leadership of Big Blue Horse and Chestnut Horse, they are quite docile.

Yang Honghui and Liang Hong are both people with excellent motor skills, especially Yang Honghui. To put it more mysteriously, he has a kind of "strength" in his body, which is the kind of strength he showed when Zhou Zhi and Mai Xiaomiao climbed the wooden stairs.

Although he had never ridden a horse before, Yang Honghui knew a lot about how to coordinate and utilize the power transmitted by the horse, and in no time he was as skilled as a veteran who had been riding for many years.

Liang Hong was almost there, but with just a little guidance, she was able to grasp the knack and master it quickly. Even Zhou Zhi couldn't help but give her a thumbs up: "Sister Hong is amazing. I've finally seen what Yan Xiao meant when he said you can pick up any sport as soon as you start."

"It's not that exaggerated," Liang Hong laughed. "I don't know how to play Go or chess. And fishing too."

"Fishing is one thing, but you call Go and chess 'sports'?"

"Before the Chess Academy was established, these two sports were also under the jurisdiction of the Sports Commission for a period of time. Since they were under the jurisdiction of the Sports Commission, weren't they sports?"

"Uh……"

"Hey, it seems like you guys are working on something about getting machines to play chess, right? How's it going? Is it as good as Deep Blue?"

This was a major event in the computer industry this year. IBM's Deep Blue computer played its first match against world champion Garry Kasparov in February, and the battle ended in a human victory, with Kasparov winning the match 4 to 2.

"Deep Blue is like chess, with 32 processors performing mixed decision-making and evaluation, capable of calculating 200 million moves per second."

"In other words, Deep Blue itself doesn't have learning capabilities. It uses calculations far exceeding human reaction speed to exhaustively search for subsequent moves in the game, and then makes the optimal decision based on the weights of the pieces and moves. It's like navigating a maze, dividing all paths into stages, then referring to the paths others have already taken, and finding the most reliable one." Mai Xiaomiao explained this without caring whether Liang Hong understood: "This method is actually learned from the 'parallel processing method' or 'parallel construction method' in the construction industry. In the construction industry, the construction period can be shortened by organizing multiple construction sections to work simultaneously at the same time."

Seeing that Liang Hong and Yang Honghui were confused, Zhou Zhi explained: "Generally, building a house follows a linear process. First, dig the foundation, then build the structural columns, then the floors, walls, stairs, pipes, and finally the water and electricity systems, followed by the exterior and interior decoration."

“But parallel construction is different. You can think of it as not starting the plumbing and electrical work only after the entire building is completed. Instead, you lay and decorate one floor at a time as you build the other. Once the structural columns of the top floor are completed, all the floors below can be handed over for use. This saves a lot of time,” Zhou Zhi said. “Deep Blue simulated this method, greatly compressing decision-making time and achieving an exhaustive calculation capability of 200 million steps per second. Then it could compete with top human masters.”

"Is your chess-playing computer like that too?" Yang Honghui fully understood Zhou Zhi's example and then became concerned about this question.

"Although Go only has two rules, the complexity of its moves is unmatched by chess. You can't exhaust all the possibilities in Go and find the best decision using brute force. By the way, do you know how many atoms there are in the entire universe?"

“The number of atoms in the entire universe is only 10 to the power of 80,” Zhou Zhi said. “But in Go, there are theoretically 10 to the power of 170 possible moves. That is to say, the total number of moves in Go is much greater than the number of atoms in the universe. Therefore, we need a smarter approach to design Go software.”

“To achieve this, we need to solve two problems first. The first is the problem of the huge number of branching factors,” Mai Xiaomiao said. “Branching factors refer to the huge search space of the moves,” Zhou Zhi translated again. “On the Go board, each piece has an average of two hundred possible positions, while in chess there are only twenty positions on average. Each possible move is a branching factor, and these branching factors need to be calculated. Therefore, the search space in Go is much larger than that in chess.”

"But what's even more difficult is the formation of the evaluation function," Mai Xiaomiao said, referring to the second point.

“This is equivalent to scoring the probability of each move,” Zhou Zhi said. “Chess is a relatively simple game, and it is a physical game. By simply counting the pieces on both sides and adding the mobility of each piece, this evaluation function can be constructed relatively easily.”

“Moreover, as the game progresses in chess, the number of pieces decreases, and the evaluation function becomes simpler and simpler,” Mai Xiaomiao said. “In contrast, Go is a constructive game. At the beginning, the board is empty, and gradually the players fill it up.”

“Therefore, if you are preparing to assess the current situation in the middle of the board, in chess, you only need to look at the current board to tell you the general situation;” Zhou Zhi added, “while in Go, you must assess what might happen in the future in order to assess the current situation.”

"To put it another way, chess focuses more on tactics, while Go focuses more on strategy. Through a huge database, we have deduced that as long as there are fewer than nine pieces on the chessboard, the final outcome can be clearly calculated using mathematical algorithms."

"As long as the computer is fast enough and the iterative algorithm is improved enough, the world's top chess programs, such as Deep Blue, will no longer make 'technical' mistakes. Therefore, when the chess algorithm is close to its limit, we will have no way to improve it further."

"Humans are bound to make mistakes. So when there are fewer than nine pieces on the board, if the program has the advantage, or if it's a draw, humans no longer have a chance to beat the computer."

"Even if a human has the advantage, the computer can still turn the tables if a human makes a mistake."

"At this point, computer simulations of chess no longer represent any significant technological advancement."

"But if DeepBlue's technology is applied to Go, the result would be that even a professional Go player would not be able to beat it, let alone a world champion."

“This data is also in our Go game system,” Zhou Zhi said with a smile. “Our Go game in QQ Games actually uses a similar calculation method to Deep Blue, which can satisfy the needs of playing against enthusiasts, but it can’t even beat amateur first dan players.”

"Xiao Zhi is much more powerful. It uses a neuron-like algorithm to decide how to play. This method is actually somewhat similar to how humans play."

"How do humans play Go?" Liang Hong asked. "Could the Go master Nie Weiping calculate so many moves?"

"We did ask the Go master this question. When we asked him how he decides his next move, or how he makes the next move after that, guess what his answer was?"

"You still have such an opportunity?" Liang Hong's gossip-loving nature was immediately ignited: "He also participated in your project?" (End of this chapter)