Riding the wind of rebirth

Zhou Zhi doesn't care about this. Whoever doesn't use it will suffer. If someone is willing to live as an information island and be a frog in a well who doesn't embrace advanced things, then Zhou Zhi will just laugh it off.

After discussing these matters, Zhou Zhi reported some of his academic ideas to Professor Gu. The most important of these was to use Xiao Zhi for deep learning to complete the project of Middle Chinese phonological reconstruction.

Explaining this to Mr. Gu is much more difficult than explaining it to Mr. Mai Mingdong.

Fortunately, there is a relatively simple example to explain this: Go.

"The rules of Go are actually very simple. There are 19 lines in total, 361 points. You can only place a stone on a point. If a point is occupied by another player's stone, it will be removed. The empty points around a stone are called 'liberties'. A stone must have liberties to survive. That's how you play."

"Determining the winner is even simpler; it is determined by calculating the size of the territory occupied by each side."

"The rest is a statistical problem. Xiaozhi needs to learn from thousands of chess games to determine the winning probability of each possible move in the corresponding situation, and then place the piece in the position with the highest winning probability."

"If several locations are similar, then we need to calculate the future development and select the location that is most conducive to future development."

"If they are all about the same, then we need to find the method with the highest winning rate from our past learning experience."

"Of course, it sounds very simple, but it involves very profound mathematical knowledge. The number of possible moves in Go is enormous, theoretically greater than the total number of atoms in the universe. No matter how advanced a computer is, it is impossible to go through all the moves in a short time. Therefore, we need to use mathematical tools and statistical methods to analyze and interpret such a huge amount of data, considering its volume, diversity, speed, and value, to understand the patterns and information behind the massive amount of data, and ultimately extract valuable information and insights from it."

"The general working principle is to first conduct data analysis, using statistical methods to analyze big data, extract useful information, and reveal the relationships between data; then establish inferences and decisions, using statistical inference methods to infer overall characteristics from sample data to support decision-making; finally, combine inferences and decisions with complex data structures and data diversity to build efficient statistical models to cope with complex challenges."

"Compared to the goal of comprehensive artificial intelligence, the computer's self-learning of Go moves is basically the simplest kind of intelligent model. So far, the results we have achieved are both phased and revolutionary." Zhou Zhi smiled and said, "Just this morning, Xiaozhi followed the entire game between Chang Shentong and Dazhu Yingxiong in the viewing mode. Its analysis of the key turning points was recognized by the Chinese Chess Academy, represented by the Chess Saint, who believes that it can enter the acceptance and application stage."

"What is the connection between this and your research?" Although Professor Gu is still learning about computers at the age of ninety, which shows his dedication to research, he is still far from mastering such high-end applications, after all, academics have their own specializations.

“For Xiaozhi, the logic of winning and losing in chess is actually the same as the grammatical logic of language,” Zhou Zhi said. “As long as it is a logical world, whether it is mathematical logic or grammatical logic, Xiaozhi can learn and master it.”

"Actually, we started experimenting with the application of classical Chinese grammar and logic in the information industry quite early on, which involved developing a data retrieval engine based on a metadata database," Zhou Zhi said with a smile. "This engine later branched into two: one became the Hanwen search engine that now accompanies digital libraries, which played a significant role in the third collation of the *Jiyun Shuzheng*." "Yes, it's indeed very useful; I'm still using it now," Elder Gu nodded. "It saves a lot of time searching for information, and I heard it even led to the creation of things like 'biographies of famous people'?"

"Yes, the main function of this branch is in academia, providing data retrieval for everyone. At the same time, it can also teach the program our data retrieval thinking, allowing it to search through a vast amount of classic texts, and finally find the series of knowledge points we want, establish logical relationships, and ultimately form a retrieval result."

"My classmate Zhang Xinyi is working on the script analysis of the retrieval program for the anecdotes of the three Sus in the 'Collection of Anecdotes of Su Dongpo,' which is his graduation thesis topic."

"Send it to me when it's finished." Old Gu was naturally very interested: "Your high school classmate, right? Time flies, you're almost graduating with your bachelor's degree. Are you planning to take the postgraduate entrance exam?"

"Yes, it seems quite confident." Elbow laughed, "But let's not talk about that yet. We also have a general-purpose engine, which is a web yellow pages tag search engine that is compatible with the Clover Browser and makes it easier for everyone to search for information online."

"The point of these two examples is that these kinds of grammatical logic analysis tools already exist. As long as Xiaozhi masters how to use these tools, theoretically he can interpret and retrieve valid information based on logical relationships."

"For example, in the biography of Su Dongpo, we can first label Su Shi with a bunch of tags such as Su Shi, Zizhan, He Zhong, Tieguan Daoren, Dongpo Jushi, Meishan Su family, the Three Su, Jinshi of the Jiayou Dragon and Tiger List, third-class imperial examination, Dali Ping Shi, Fengxiang Prefecture Judge, prefect of Hangzhou, Mizhou, Xuzhou, and Huzhou, Wutai Poetry Case, deputy commander of Huangzhou militia, old party, literary giant, bold and unrestrained lyricist, Su Huang, Su Xin, Ou Su, the Eight Great Masters of Tang and Song, the Four Masters of Song, the Six Masters of Song, Dongpo's Seven Collections, Dongpo's Commentary on the Book of Changes, Dongpo's Music Collection, Cold Food Festival Post, Xiaoxiang Bamboo and Rock Painting, withered tree and strange rock painting, etc. We can use these tags to search the database, get the relevant information, and then extract more and more complex tags from this information. Finally, we can complete the biography of Su Shi, extracting all the deeds of Su Shi that have appeared in historical records, ancient notes, and official documents of the dynasty, forming a complete biography of Su Dongpo."

"We have succeeded with this method. Now, we are doing something similar to letting Xiaozhi master this method. We only need to give him an instruction to 'collect Ouyang Xiu's biography,' and he can use the same method we used to create 'Dongpo's Biography' to create 'Liuyi Jushi's Biography,' and then apply it to other historical figures."

"As for phonetic and rhyme retrieval, although it is more complex than celebrity biographical retrieval, and involves selecting samples from dialect specimen databases for assembly and verification, the underlying research logic is similar: using big data statistical analysis methods to retrieve and compare massive amounts of data, and finally finding the optimal solution."

"The manifestation of phonetics in writing is quite extensive, including poems and prose left from previous dynasties, proverbs recorded in ancient people's notes, lyrics, and songs. More importantly, there are phonetic documents officially promulgated by successive dynasties, as well as the research results of scholars on phonetics from successive dynasties."

“We should try our best to find all these things, summarize the patterns according to different eras, select the best solution from dialect samples, determine it as the closest to the pronunciation at that time, and finally integrate them into a set of Middle Chinese phonetic reconstruction tables. I think there is nothing wrong with the idea.”

"The main point is to leverage the advantages of modern information technology." (End of Chapter)