Shin Jinseo achieved 16 consecutive wins in the Nongshim Cup, surpassing the legendary record of 14 consecutive wins by Lee Changho. As the main player, he single-handedly defeated all participating Chinese team players, winning the Three Kingdoms Tournament for his fourth consecutive time, once again proving that he is the undisputed world number one in Go. Here, I want to talk about my impressions of Shin Jinseo, including his Go skills and character. In terms of Go skills, I mainly want to discuss where exactly Shin's Go is leading, especially with a brief analysis from the perspective of AI training methods and the theoretical level of Go technology. This part, based on the analysis of knowledge theory, may benefit our frontline players and even young talents. The reason why I want to talk about character beyond Go skills is that many people still believe in the saying "as the person, so the game", believing that the personal character of a top player must align with their Go skills—these two aspects are almost inseparable. This belief, in fact, faced serious doubts in recent years. Not only from societal questions (people asking: "Is being good at Go any different from being good at esports?"), but also doubts among players themselves (including children learning Go and their parents). Fortunately, Shin Jinseo seems to have, at least temporarily, restored this belief. Many Go players like to use folding fans inscribed by Master Go Seigen, with the most common inscription being "Yàn Rán Ér Rì Zhāng" (闇然而日章). Perhaps it's more out of admiration for Go Seigen's exceptional achievements that they haven't deeply considered why Master Go Seigen inscribed these five characters on the fan. "Yàn Rán Ér Rì Zhāng" comes from "Zhong Yong" (Confucian Doctrine of the Golden Mean): "The way of the gentleman is hidden and yet becomes more visible every day; the way of the petty person is flashy but gradually fades away." Here, "Yàn" shares a meaning with "darkness," suggesting that a gentleman's brilliant virtues are cautiously hidden within, initially seeming unremarkable, yet their moral integrity brightly reveals itself over time; conversely, the petty person focuses outward, vainly flaunting without self-reflection, succumbing to material desires without turning back, initially appearing bright but internally hollow, "exposed outwardly without substance to sustain it," thus gradually declining. In his later years, Master Go Seigen envisioned Go in the 21st century, proposing Go as the way of "moderation" and esteemed "Zhong Yong," inscribing "Yàn Rán Ér Rì Zhāng" on the fan to pass down his life philosophy and view on Go. Shin Jinseo, a player born in the first year of the 21st century, has shown the gentleman's way of "Yàn Rán Ér Rì Zhāng" in his journey to the pinnacle of Go, becoming a grandmaster and a role model for many young learners—truly a blessing for the entire Go community. Firstly, Shin Jinseo's phenomenal dominance has at least disproved one misconception. After the widespread adoption of Go AI, there was a notion that AI had narrowed the gap between players' skills, as everyone was learning from AI, hence levels became similar. However, Shin Jinseo's consistently near 90% win rate, rivaling Lee Changho's dominance, clearly refutes this idea. Obviously, not everyone studying AI reaches similar levels. Our domestic players especially should not be deceived by this argument anymore. AI indeed brought about great equality, but this equality is mainly in terms of training conditions (after everyone has access to strong AI for daily training), not in training outcomes. The advent of AI actually raises higher demands on players' abilities, a point that frontline players and those dedicated to competitive Go must take seriously, which is the main part I want to discuss. It's undisputed that Shin Jinseo has reached the highest level of human Go so far (of course, the overall level of human Go is continuously improving, and undoubtedly, players of even higher caliber will emerge in the future). So, how did Shin Jinseo achieve this height? What lows and difficulties did he encounter in his climb to the pinnacle of Go? Where does his technique primarily lead? Many have noticed that in recent years, Shin Jinseo's gameplay has transitioned from the aggressive style of his early career to a controlling style. In the Nongshim Cup, he maintained control from start to finish in all but one game against China's main player Gu Zihao, which experienced significant turbulence (proving that Shin also has vulnerabilities that could lead to defeat), leaving his opponents with no chance to win. Here, I want to propose a simple thesis: in modern professional Go competition, the most dominant players are likely to be those who excel in controlling the board. This is a colloquial statement, which I will later explain using the theoretical framework of Go technology. Of course, Shin's fighting ability, or computational power, is very strong. When he first entered the world stage, he was known for his combative style, and many of his losses during that period were due to over-aggression even when he had the advantage (I remember wanting to face him during the draw ceremony of the first Nongshim Cup, as he was the only non-Chinese player among the top eight, and also remarkably young. At that time, I had not been training for many years and was gradually withdrawing from frontline competition, while Shin was just starting out). After the release of Go AI (especially after KataGo was open-sourced in 2019), Shin began to learn from AI, gradually compensating for his weaknesses and shifting his style to a controlling one, achieving a remarkably high win rate. In fact, the various "styles" in Go are more for entertaining discussion than precise categorization. The so-called "controlling style" refers to consistently steering the game towards victory without engaging in fierce battles, leaving the opponent powerless, a strategy that avoids the inevitable fluctuations in complex calculations, smoothing out the game with a margin for error, thus maintaining a high win rate. But not all "controlling" players can reach such heights. The last peak of the controlling style was, of course, Lee Changho, whose peak period was characterized by seemingly ordinary, yet opponent-controlling moves. Now, with AI reanalysis, it's clear that Lee Changho's strength was not in endgame, but in situation judgment, with many games previously thought to be reversed by his endgame actually being cases where Lee more accurately assessed his advantageous position, while opponents and spectators misjudged the situation (details on how he achieved this can be discussed later). In the later stages of Lee Changho's career, the perception that he changed his style to become more "combative" was because, under his influence, younger players' judgment and sense of winning and losing also gradually strengthened. When there's no significant gap in judgment ability, controlling the game becomes difficult, often requiring combat (calculations with multiple uncertain outcomes) to resolve issues, leading to a decline in his win rate and dominance (this was also when I participated in national team studies on Lee Changho). That is to say, the key to controlling the game is "judgment". Here, I propose that the area where Shin Jinseo most leads today is in "judgment", clearly enhanced through training with AI. Professional players often say, "Judgment is the hardest aspect of Go." So what exactly is judgment, why is it the hardest, and how can it be improved? Due to space limitations, this discussion can only be brief here, with a more comprehensive discussion possible in the future. When compiling the "Chinese-English Dictionary of Go Terms", I found that many terms in Go lack precise definitions, and imprecise terms cannot lead to accurate knowledge. When talking about the technical domain of Go, terms like "mid-game ability" lack clarity, whereas the concept of "judgment" is relatively clear. Seven years ago, in my article "AlphaGo—The Future of Go," I first proposed the three elements of Go technology: intuition (sense), calculation, and judgment, which precisely correspond to AlphaGo's algorithm structure. The capabilities and techniques of human Go playing can all be encompassed within this framework of three concepts. Last year, I wrote a paper titled "An Analysis of the Knowledge Structure in Go," in which, beyond the main theme, I incidentally mentioned how learning from AI could bring about progress and potential for improvement in these three areas for human players. The conclusion was: in the aspect of intuition, continuous progress can be made through the accumulation of experience; there is very little room for improvement in calculation; however, there is significant potential for advancement in judgment, with the possibility of future theoretical developments. The memorization of all josekis can be categorized under intuition. The first phase of learning Go from AI involves learning the AI's sense of play, which means watching AI's moves to generate better impressions of the board in similar or near-similar situations. Since the opening phase of Go is where "similar or near-similar situations" occur most frequently, it is also the part of intuition that is easiest to learn, commonly referred to as memorizing josekis (though this is not entirely accurate). At this stage, human players relearn josekis, a necessary task for becoming a professional player, but this is not the entirety of learning. Beyond opening moves, there is room for updating the intuition related to mid-game transitions, local shape assessments, thickness and thinness relations, and knowing when to tenuki, among others. This part of training is primarily based on accumulating experience; the more one sees, the more one recognizes. Of course, moving from seeing a lot to truly understanding also requires careful analysis and understanding of unfamiliar or incomprehensible moves, to truly transform them into one's own intuition. Some players and fans mistakenly believe this is all there is to learning from AI, thus assuming that good Go play post-AI comes down to memorizing sequences, a clearly mistaken and easily spread misleading viewpoint. Discarding old josekis and learning new ones from AI, thereby forming a new sense of play in the openings, is just the first step or phase in enhancing Go skills in the AI era. In terms of calculation, AI indeed cannot help us much, since human players' calculation abilities are limited by the capacity and efficiency of the human brain, or human limitations. Without using a brain-computer interface, there's not much room for improvement in human Go calculation abilities overall, though individual differences still exist and are unrelated to AI. The key issue is "judgment." In fact, during the matches between humans and machines in those years, we were able to fully understand that the greatest gap between human players and AI lay in the realm of judgment (not calculation).Why is it AI that teaches us to play at the 3-3 point? Is it because humans can't calculate the basic variations of 3-3? Of course not. It's rather that human players' judgments on territorial transitions were previously inaccurate. If we elaborate, it means that during transitions between territory and influence, the precision of judgments made using the two cognitive methods employed by humans ("logic" and "experience") differs significantly, leading to inaccuracies in judgment (whereas judgments in pure territorial battles don't deviate much because the same cognitive methods for judgment are used). The judgment capability in Go (corresponding to AlphaGo's Value Network), is an area where humans can improve through learning from AI. Humans lag far behind AI in this aspect and are far from reaching the limits of human abilities, therefore, there is significant room for improvement. AI's Value Network provides win rates, and naturally, humans can't directly convert board positions into win rate estimates like AI, so humans still need to use their own cognitive methods for improvement. Human cognition comprises logic and experience, where improvement in intuition primarily relies on the accumulation of experience; however, improvement in judgment has room for both to play a role. Human players can now undertake specialized training in judgment with AI as an excellent auxiliary tool, making such training one of the most important regular training methods. Of course, since some top players keep their AI training methods secret, it's not appropriate to discuss specifics here, but it's necessary to clarify the theoretical basis for why specialized judgment training is feasible and effective. The easiest way to use AI to train your own situation judgment ability is to use experience accumulation, and it is also the most straightforward training scheme currently available. Future logical methods may rely on new conceptual thinking, summarizing better, more effective Go strategies (discarding some past incorrect or imprecise concepts). Though this is likely not something that can be achieved in the short term. Theoretically, at this stage, devising specific training programs to improve judgment ability, with the main goal of accumulating correct experiential judgment, is very important and feasible for players still aspiring to compete at the top level. In this regard, Shin Jinseo has likely already taken the lead, with most players still in the initial phase of accumulating intuition (including some who are resistant to learning), while Shin has already advanced to a higher class. Shin Jinseo's ability to utilize the control-oriented style of play against many top players who also train diligently every day is mainly because his judgment ability is at least one notch higher. When entering complex battle scenarios, his win rate significantly drops (recent players who have defeated Shin told me their strategy was to engage him in combat). If we use AI's win rate fluctuations to explain, the so-called control-oriented style is about maintaining the board in a state where many move choices lead to slight differences in win rate. Mistakes here do not suddenly result in a loss. Whereas the so-called complex battle (or whatever other name it might have) is characterized by requiring almost every move to be precise in situations with many calculation branches and a single mistake could cause the win rate to plummet. Modern players, who have been rigorously practicing life-and-death problems from a young age, find it difficult to significantly surpass others in calculation, hence while fighters can achieve very high levels, their consistency is lower than that of control-oriented players (as determined by the principles of Go and human capability limits). Of course, from another perspective, if everyone's judgment abilities were similar, complex battles would become the main method for determining wins and losses (as seen in AI matches), at which point the so-called control-oriented style would no longer exist, nor would there be players with extremely strong dominance. In summary, Shin Jinseo's exceptionally high win rate relies on his control-oriented style of play. And the reason this approach has been hugely successful is due to his judgment ability is significantly ahead of the other top players (just like Lee Changho in his prime, who could predict certain changes in the game where one side was already leading while the opponent still felt the situation was evenly matched). The substantial improvement in his judgment ability has allowed him to avoid early collapses in prolonged battles, which are prone to reversals, and this enhancement in judgment is closely related to his use of AI in training. This means that currently, the most likely method to defeat Shin is to engage in complex battles; and to catch up with Shin in terms of skill, one needs to find new training methods to enhance judgment abilities with the aid of AI. Shin's ability to learn at a higher level through AI might not be entirely conscious on a theoretical level, but it is deeply connected to his understanding, cognition, and feelings towards Go AI. What left a deep impression on me was an interview in 2021 where Shin Jinseo discussed his views on Go AI, which I still consider to be the most meaningful response from a human player in the era of AI. Shin Jinseo said: "Blue spot is the move recommended by AI, the one with the highest probability of winning. But beyond that, there is clearly something more." "I have finally realized that winning or losing is not everything in Go. Now, I focus more on the process of a game. Before a match, am I fully concentrating on the game? During the game, have I made any careless moves? When in a favorable position, have I become complacent? When in a disadvantageous position, have I given up too soon? When mistakes occur, have I tried my best to turn the situation around? In short, I have developed a deep respect for Go." (See the interview—Shin Jinseo: AI Lets Humanity Leap Over Fences, Pursuing a More Profound Art of Go). The "reality of the game" presented by AI calls on players to continually seek new understandings and languages. This search requires humble rationality as a virtue's foundation, while the arrogant will fall into contradictions and confusion. Here, the relationship between Go and humans begins to reveal itself. The reason Shin Jinseo embodies the gentleman's path of "Yàn Rán Ér Rì Zhāng" is partly because he didn't prematurely reveal his brilliance and stand at the pinnacle to be worshiped. Instead, he went through a long period of lows, even facing mockery from opponents during his difficult climb and pursuit, was toyed with by the whims of fate to the point of crying against the wall, was kept awake by baseless slanders, yet always maintained the dignity and grace of a Go player, keeping a balanced tone even when forced to retaliate. From a purely competitive utilitarian perspective, perhaps Shin's greatest weakness is his strong sense of responsibility. He clearly bears not just the responsibility of winning titles and breaking records, but also of maintaining the image of Go players and setting an example for the industry. These responsibilities can sometimes become burdens in matches, making it hard for him to enter a selfless state of clarity in finals, unable to completely set aside all responsibilities to focus solely on winning. Of course, his legendary performance in the Nongshim Cup proves that Shin can fully shoulder these responsibilities, demonstrating that one does not need to abandon them to win unburdened. When Shin was unjustly slandered that year, I was too angry and blocked some Go players who were smearing him, and through a friend in Korea, I informed Shin that there were many Chinese Go players who opposed the baseless smears and supported the honor of Go players. This was done to prevent misunderstandings from collective public silence. Similar to the Diana incident in the Korean Go community, I expressed support from Chinese Go players to Ms. Diana through Mr. Kim Seungjun, motivated by the same idea. For various reasons and rationales, we have done too little for justice, which is terrible for those smeared; at the very least, we should follow our hearts to express support, even if it is negligible compared to the harm they have suffered. When the COVID-19 pandemic broke out in 2020, my family worked on the front lines in hospitals, and I stayed in Wuhan during the lockdown, feeling helpless and useless. Consequently, I proposed initiating donations from the Go community to the disaster area. Surprisingly, we quickly received donation requests from Korea, with Shin Jinseo, then not yet twenty, being the first Korean Go player to propose donations to the Wuhan area, followed by Lee Changho and Choi Jung, each donating 10 million Korean won (~$7,500) to support Wuhan's fight against the pandemic (meanwhile, the donations also received significant support within the domestic Go community, but also sparked discussions of "refusing moral hijacking"). This reminded me of the 2008 Wenchuan earthquake when Lee Sedol and Cho Hanseung decided to donate all their prize money before the Asian Cup finals; I lost to Cho Hanseung by half a point in that tournament, how I wished to join them in their good deeds. In such moments and events, I feel we are part of the same industry community, and good examples are incredibly important for younger practitioners. The profession of Go players does not produce any supplies required for social and economic life, but it has received a lot of material rewards and some social attention. If Go players neither explore the boundaries of human spirit and thought, nor know how to give back to society, but only know to take excessively without concern for the vulnerable and focus solely on victory and defeat, then the foundation of legitimacy for this profession will soon weaken. Fortunately, we have players like Shin Jinseo. Whether it is his Go skills, his understanding of Go and AI, his attitude towards Go, his character and integrity, or his awareness and responsibility towards the industry and the inheritance of the way of Go, Shin Jinseo can indeed be regarded as a model in the Go community. Not yet twenty-four, he has gradually shown what is meant by the "gentleman's way, Yàn Rán Ér Rì Zhāng (secretly by brightly)." NOTE: This article was first written on WeChat in Chinese. Later Wenzhe Song posted an English translation on Facebook. Devin Fraze then hunted down the original article, checked the translation (likely not ai; checked against google translate and ChatGPT 4) and then edited the article and added citations.
