Riding the wind of rebirth

"The story goes that in ancient China there was a man named Bo Le who was very good at judging horses. He wrote a book called 'The Classic of Judging Horses,' which said that the standard for a good horse is a prominent forehead, eyes as round as large copper coins, and hooves as piled up like large cakes used for making wine yeast."

"His son took the scriptures and went out to find a horse, but after searching for a long time, he couldn't find one. He went back and told his father, who told him to try again. This time, he returned soon after leaving and said to Bo Le, 'I just got a horse, which is almost exactly as described in your book. It has a raised forehead and round eyes like big copper coins, but its hooves are a little lacking. It's not quite good enough for making wine yeast cakes.' Bo Le then told his son to bring the horse over for a look. Upon inspection, it turned out to be a toad!"

Bo Le could only smile wryly and say, "This horse is good, but its hooves mean it likes to jump, making it impossible to control."

"Hahaha, what's this idiom called again..." Mai Mingdong laughed, but then got stuck and couldn't remember the name of the idiom.

"Follow the map to find the horse," Zhou Zhi added.

“Yes, yes, yes, follow the clues.” Mai Mingdong laughed: “Actually, the method of logical reasoning used by Bole’s son is not wrong, but the reason why he used reasonable logic to deduce an absurd conclusion is because the premise was not taken into account, and the biggest premise was ignored—that is, the toad is not a horse at all.”

"Yes, so in a sense, the study of 'elementary learning' in my country throughout history has more or less fallen into the logical trap of 'following the map' and has never broken free."

"You mean to say that their conclusions, because they lack actual sound samples for reference, are often wrong or even absurd?"

"That's fair to say, to a large extent, yes," Zhou Zhi said. "They even deduced the trilled 'r' sound, which is completely inconsistent with the traditional Chinese language style. This is because they were heavily influenced by Western phonological research tools, which led to the deviation."

"Based on the dialect samples we collected, the trilled 'r' sound is only present in certain extremely special regions of Chinese and in some extremely special characters. In the vast majority of regions and applications, this method does not exist. Therefore, we can use actual samples to show that this pronunciation method is not mainstream in Chinese-speaking regions, and it is even a foreign pronunciation method. When reconstructing Middle Chinese, it should be removed, unless it is a foreign word, such as a Sanskrit transliteration of 'Svaha'."

"The reconstruction I am about to undertake will be based on the following premises. First, the various dialects of the Chinese language will retain traces of Middle Chinese. Even though the Chinese language has undergone tremendous changes over thousands of years, it will always leave some traces that can be traced."

"The second premise is the research of famous scholars throughout history. We will not blindly follow or believe in them. We need to find corresponding factual evidence in the dialect specimen database to support the conclusions they made based on written records and logical deductions."

"Only if we can find empirical evidence in the dialect database can we consider it credible; otherwise, we can only continue to question it..."

“Well, that might still cause problems,” Mai Mingdong said. “If there are too many questions, then this research will be meaningless.”

“Yes,” Zhou Zhi said with a smile. “However, at least there is no problem in the reconstruction of Middle Chinese. Cantonese, Hakka, Wu dialect, and the palatal and velar consonants and entering consonants in Peking and Kunqu operas, etc., can now be reconstructed into a very complete Middle Chinese phonological system.”

"So what is the new idea you just mentioned?"

"What I need to do now is to take all the rhyme books, notes, proverbs, poems, and parallel prose from past dynasties—in short, all the content that can inspire the combination of sounds and rhymes—as evidence to verify the logical relationship between sounds and rhymes in language. I will then combine this evidence with the conclusions of great scholars of philology from past dynasties, as well as the results of modern research methods, to create a unified logical reconstruction. Finally, I will obtain the most complete and authoritative reconstruction scheme for Middle Chinese phonology at present."

"Is this a new idea? Isn't this your original research topic?" Mai Mingdong didn't understand.

“My new idea is to entrust this task to Xiaozhi.” “Deep learning functionality?” Mai Mingdong was startled, immediately recalling something Zhou Zhi had once mentioned: “Artificial intelligence?”

"It hasn't reached the level of artificial intelligence yet," Zhou Zhi said with a smile. "It's more appropriate to call it deep learning."

"In fact, we also follow this technical route in chess training. We have already accumulated some foundational knowledge and made pioneering contributions to the nonlinear learning unit cluster generation model and validation model. So far, we have achieved phased results."

"This achievement started with the analysis of graphic edge detection and the intelligent collection and summarization of graphics. At that time, it was applied to the oracle bone script joining project and the oracle bone script root analysis project, and achieved remarkable results."

"After Xiao Miao arrived, we ported this system to a nonlinear supercomputing matrix, and the effect was even better. We also explored a preliminary deep learning functional model."

"Next, we will use supercomputing matrices to simulate nonlinear physical phenomena. The important results are liquid cavitation simulation and meteorological and financial market prediction. The results are also very good, and several mathematical generation and calculation models have been generated as a result."

"Once the supercomputing matrix is ​​fully connected to optical communication, the matrix's overall computing power will be greatly improved. Based on this, we started cooperating with the chess academy's game software."

"This is a milestone achievement. Starting with Xiaozhi's first victory over Nie Qisheng during the internal testing, it marks a revolutionary achievement for Xiaozhi in professional-grade deep learning functions. We have also completed the functional development of computer deep learning based on nonlinear algorithms."

"Now we have entered a new stage. I plan to combine the research on Middle Chinese phonetics with this and use it as an application project after the development of the algorithm to verify it."

This is one of the things that Mai Mingdong has to admire about Zhou Zhi. Many research projects in domestic university departments are often divorced from reality, resulting in the waste of a lot of research funds, but the final results can only be shelved and cannot be well applied.

Enterprises' spontaneous scientific research often focuses only on solving practical problems in production, lacking academic foresight and developmental redundancy. As a result, the research results are often limited to solving the enterprise's own specific problems and have little further development or academic value.

Zhouzhi has always been selective in his choices, always taking a two-pronged approach, and he has consistently done very well in doing so.

From the very beginning, the use of a unified character encoding and the Hanwen large character library laid the foundation for the Clover Group's development style.

Unified character encoding is a purely academic matter, involving three levels of work.

First, at the most basic protocol layer, a standardized protocol suitable for Chinese character encoding is needed. This protocol must fit the characteristics of Chinese characters, not be constrained by the traditional ASCII code, and also leave ample room for development. In the future, in addition to accommodating Chinese characters, it should also be able to accommodate encodings for mahjong, dominoes, Taoist talismans, special variant Chinese characters, and even minority languages ​​such as Tibetan, Mongolian, Yi, and Manchu.

Zhou Zhi ultimately chose UNICODE, and it now seems indeed excellent. (End of Chapter)