Very nice commentary, REALLY appreciate the comparisons with modern AI play, adding another dimension of interest and usefulness. A couple of fine bits of yose too. The "almost" losing move in the centre was the one I wanted to play there too! So should have realised it was your mistake, LOL! Didn't even think of the attachment...
Thank you mister Redmond. I enjoy these reviews of your personal historical games. I could include basically all games before AI in that category but I think that would not be appropiate for the really old games.
Really enjoyed this, thank you Michael. As a fellow Redmond I want to maintain the high standard of the Redmond name in Go lol. I noticed your middle name is also Irish, so I assume Irish ancestry. I'm from Dublin.
In chess I saw when computers became "the thing", like a new Moses with 10 new commandments. But then, people started saying: "well, the computer says this variation is good, but I don't like to play that kind of resulting position. So, I'll do differently." That is because people realized that the evaluation was based in a certain calculation power impossible to any human (which includes the opponent!) and in order for it to work they had to memorize some strange moves that made no apparent sense. But in Go, it looks like top players just don't care. Maybe it is the "sensei culture" so present in the East, only now replaced by AI as a new sensei.
Maybe. Our understanding of the opening was always a bit fuzzy, and I think people realize that it is impossible to reproduce the calculation of a computer, but are enjoying the challenge of these new openings. We will surely see more developments in the AIs and the way people use them and I hesitate to predict if the trend will continue indefinitely or not.
@@MichaelRedmondsGoTV I am more familiar with shogi (Japanese chess) than chess, thus I compare go to shogi. I feel that in go it is often much easier to learn from AI plays than in shogi, because "many fight patterns are quite local and many variations are not too long to follow". Many variations of direct 33 josekis for instance can be evaluated without the global situation elsewhere. Traditional osae with only two stones and extension instead of double hane looks bad almost always regardless of the global board position. In this sense, AI josekis are simply new variations proposed by new stronger players. Nothing qualitatively different from the new josekis before AIs, like Dosaku presenting new josekis in this sense. In shogi (or chess), it looks much harder for a human to play like AlphaZero, probably because the patterns, if any, are essentially global and the variations are extremely high with extremely risky moves abundant in checkmate bombs. Let alone mimicking, it is difficult to abstract the patterns like joseki in the first place. It is true that some AI-originated human implementable openings have become so popular as one of them, Elmo enclosure, to receive the New Move Award from Shogi Association. However, the general trend is that the top pros do not simply mimic AI openings, and I elaborate the approach below. By the way, there is of course sensei culture also in shogi, but still the introduction of AI moves in human games seems very different from that in go. Also, the context looks very much different before and after the introduction of open-source AIs. Today, as you explain in other videos, the players can explore variations deeply with the assistance of AIs and it looks quite far from simply mimicking the AI moves. Even when a human-created variation loses 1 or 2 % of the winning percentage, it is usually much more important that the opponent is not familiar with the variation. Thus, particularly the world ranking No 1 Shin Jinseo seems to have created many original moves that do not appear in AI vs AI game records. The players simply playing as Marcelo suggested "mimicking AIs without deep understanding", I feel, generally do not get good results. The use of AIs today by top pros as I understand are more like conventional study groups (kenkyukai), only that other group members are the AIs. This study group approach seems to be adopted by the majority of top shogi players -- you test your hypotheses of human implementable strategies using AIs and get advice from AIs to review your games.
I have some additional comments regarding this within chess. In go, as I mentioned, it looks easier than in shogi to extract local play patterns per se -- copy-pasting the exact play variations pay off in many local fights. In chess, it looks as though more subtle fuzzy learning of playstyle looks much more fruitful. Many chess players were deeply fascinated by the new dynamic style of play by AlphaZero. Gary Kasparov may be one of the first to mention this regarding the game records published in Science Journal paper. But there are others like the grandmaster Karlsen as well ( ruclips.net/video/uPUEq8d73JI/видео.html ). AlphaZero seems to put much more importance on the piece "positions" than the piece values independent of the position. The conventional rule of thumb before AlphaZero was that piece values can be determined pretty much independent of the position, since in chess pieces, especially queen, can move very mobile all around the board. AlphaZero's playstyle revealed this is not at all the case. You want to maximize the mobility of your pieces and kill the mobility of your opponents' pieces, and for this purpose it makes full sense to sacrifice your pieces flexibly, just like sacrificing stones in go. I do not know whether AlphaZero leads to new opening books, but it certainly fundamentally changed the way the top players play chess in the world. In shogi, as far as my understanding goes, AlphaZero plays very much like the above, trying to eliminate the opponent's effective moves and maximize your effective moves, the strategy long known in reversi (Othello). However, I know almost no human pro that implements this style to a high level at this moment. There is a player Chida Shota who's main training method is to grasp the subtle patterns of AIs through sparring, and I do feel a lot of AI flavor in his moves. Another player Toyoshima Masayuki, the current Meijin and Ryuou (the two biggest titles), is also well known for his having changed the playstyle fundamentally through sparring with an AI program. However, since AlphaZero is not open source, no one has been able to spar with AlphaZero itself. Thus, like human cognition in general, it is quite often much more effective to grasp the fuzzy features underlying the playstyle of AIs than the concrete moves appearing in game records. In go, when people start thinking that the local exact patterns turn out not to be effective in learning, it will not only be the matter of mimicking AIs, but it will mean the fundamental change in learning go. The next generation genius kids may possibly come to learn value network globally, like AIs, rather than trying to learn precise local patterns like joseki. Shin Jinseo is apparently like that already to some extent -- he is able to play really surprising moves suggested by AIs but no human other than himself.
In fact, computers have completely changed chess. Even the rules have been changed. Adjourned games have been abolished, and today nobody is allowed to bring a mobile phone or the like into the tournament hall. Games not only for the World Chess Championship are played without any of the players calculating even one move themselves. In some cases, 100% of the prepared analyses are played. The former world champion Visnaswathan Anand was feared for this. Garry Kasparov also. I´m an Amateur with about 2300 Elo, and I also lost games, where my opponent simply didnt need to make a single move by his own. Just think of the chess databases with more than 8.000.000 games. The influence of Computers is still very little in Go compared with chess. I hope that this will remain so for a long time to come.
@@matthiasrichter3112 I would like that too, but I doubt that it will last more than a couple of years. Technology develops faster and faster every year, so I think it will not be long until Go reaches the situation chess faces today. The good thing is that fuseki is very different from openings in chess, so "preparation" in a chess sense is more difficult to achieve.
This game does not look at all like the O Meien I am familiar with. It is interesting to learn that he played much more conventionally than he has been since his top days to today.
It is sad to see computer's influence on fuseki pratically wiping out certain openings and making fuseki more limited to certain patterns related to hardcore territorial strategy. The sanrensei and chinese are now virtually nonexistent in top play. Takemiya's style, just to give one example, is a museum thing now.
Don't be too sad-there's plenty of richness and diversity in the AI style fuseki, too. Also, let's not forget that just 10 years ago, basically half of the all games consisted of either the Chinese fuseki or Orthodox fuseki.
I had some empathy for your other comments relating chess to go. Now, I think you are terribly wrong in really toooooo many ways Marcelo, blaming AIs for all the trends. As I already responded quite thoroughly to your last comment related to chess, not too many trends are qualitatively different from those before the AIs. As I elaborate below, the boring trend can be much more attributed to the fierce winner take it all environment in Chinese and Korean go than AIs. For the very example of sanrensei, it was already pretty much the museum thing long long before the AIs. Takemiya himself gave up playing sanrensei most of the time, though he continued to love nirensei. Nobody could ever show effective play with sanrensei against several modern responses even before the AIs. And, quite the opposite of what you claim, the Chinese style used to be much more boring than AI style fuseki of today. Really very high percentage of top pros adopted some of Chinese style fuseki for quite a while. Also, since the variety within Chinese style was much less than that within AI style fuseki, there were many games sharing something like 30 moves. It was like the player edging in some local sequence study within Chinese style had better results, similar may be said regarding the notorious avalanche joseki whose variation lengths are like over 30 moves. It is quite natural that the top players try to play according to the modern dominant strategy of the era, unless you can come up with a better strategy of your own. Half retired kind of Japanese players like old Takemiya that can earn a decent amount of money without very high winning percentage can do whatever they like, but especially the Korean and Chinese top pros, whose fame and income depended essentially on the results of international tournaments, have had to rely on the safe openings. You are also darn wrong thinking that AI strategies are more territory oriented than the traditional human's. If you look at the handicap games of Katago, a significant percentage of the game records show that Katago dominates the center despite giving three to four stones handicap to Fox 9ds. It is only that sanrensei for instance is not at all an effective strategy for building thickness. If AIs really love territory in the first place, why do they most often play star points in the first place????? They welcome the 33 invasions, which can be avoided if you started with komoku or 33. In fact, today, exactly due to the influence of AIs, komoku playing became much much more seldom, which seems to show that the top pros learned from AIs that too much territory orientation may be too thin in the long term.
The following may be a best example that the World Ranking No 1 (2020) can play with a neo cosmic style, depending upon the situation. ruclips.net/video/VXJ48f6nPnY/видео.html The point is that you do not force yourself into cosmic style. You always try to be as natural as you can and sometimes the game may result in cosmic style.
@@decidrophob When I wrote the comments you read, it was more to express my feeling on the subject of how AIs impact the way top players approach competitive games (go, chess etc.) and how I feel about the new paths resulting from that. I was not imagining that someone would write a ‘treatise’ in response to it. I can’t deny that players (of all levels) want to win and that they want to learn ways to maximize their chances of doing so. The will to win is an integral part of their activity. Trends, new joseki, new strategies always change over decades. New ideas take over older concepts as a new understanding of how to proceed. It was always like that. My sadness comes from the fact that (as it happened to chess 20 years ago), because of AIs (but not only because of it), go is going now through a phase when some of its ‘humanness’ is lost in a sense that all the comments/ideas/guidelines coming from players and analysts revolve around “AI X says it is good”, “AI Z thinks this is better” etc. etc. And that is my main point. Until now, the game was developed by people, who impart to it a little bit of pragmatism, but also a little bit of emotion, a little bit of human imperfection. Every new trend brings changes to what is currently done; some variations vanish, some new ones appear and so on. Maybe it will happen to sanrensei what happened to the King’s gambit and similar openings in chess. Nobody plays them at the highest level anymore (I personally think it sucks). But one way or another, the changes were proposed by people, not by people trying to play like a computer program just because it wins games easily. I don’t crusade against it, and, to some extent (mind that I’m only a 10k who learned the game two years ago) I also try to embrace it when I can, but only as a means to win, not because I think that it is emotionally pleasing. To me, the games of players like Dosaku, Go Seigen, Takemiya Masaki, Cho Chikun, Kobayashi Koichi, Lee Changho and Cho Hun-hyun are great because they carry within them the will of those players to bend the rules, each one in a different way according to their own personalities and creativity, and not by the dry use of some sort of ‘computer’s pragmatism’.
@@decidrophob Answering this comment, and adding a bit more on my newest comment above, this “cosmic style” will still appear from time to time, as much as sometimes someone uses the Bishop’s opening or the Two knight’s defense. But it will be more and more associated to a “weapon to throw the opponent off balance” or a “momentary thing” than to a style on which a top player would rely on to win titles, and that is because people will trust less and less it is good to win on a regular bases (inspiring less and less analysis and developments). Shin Jinseo is so strong right now that he can play 10-10 on the first move and probably still win, but that doesn’t mean it will become a ‘honte fuseki’. My addition to my last comment is that AIs play 4-4 points just as a means of accelerating development, much in a way Go Seigen and many other territorial players did since the 1930s. It is just a sort of springboard. There is a game I love by Kobayashi Koichi against Takemiya (Judan title, 3 april 1986, Go World n.45 p. 13) in which he plays the sanrensei as Black but aiming to develop fast and not to use it to develop a moyo as Takemiya himself would do. AIs are territorial, indeed. That is why, for example, they invade the corner so early. They like to grab some sure profit because their computational power allows them to calculate how to deal with any wall created by the opponent in the process.
Yeah. About openings. What the go players and computer scientists do not seem to understand is that the algorithms are following the road they are coded. The classic Chinese opening is still a winning opening, but not if your "brain" (heuristic code) is optimized to fight with a different opening. Why is that? It's because no "AI" is actually conscious, they have already forgotten how they opened after a few moves, and cannot coherently in language explain their "thought processes" (they can core dump their code on you, but that is no explanation) and they have no cognitive strategy, they only act _as if_ they have subjective strategy, but pseudo-random number generators in coded heuristics is not subjective cognitive, it is fake-cognition. Sure, it is powerful in a game. But so what? The beauty of go (or any competitive game) is only evident when you lose. Put another way, if I were a 9-dan Pro, I'd find a delicious thrill in beating AI with a Chinese opening, and if I lose by just the komi I'd still get a thrill. I'd be frickin ecstatic and treat myself to Mr Jiro's Sushi.
![ShinJinseo's modern cosmic style he forgot how to lose [game review] Goproyeonwoo](/img/1.gif)
Another supremely good game commentary Michael. The global perspective is so vital but many reviews tend to focus on fighting. Many thanks.
Very nice commentary, REALLY appreciate the comparisons with modern AI play, adding another dimension of interest and usefulness. A couple of fine bits of yose too. The "almost" losing move in the centre was the one I wanted to play there too! So should have realised it was your mistake, LOL! Didn't even think of the attachment...
I am in love with this type of video!!! Thanks Michael.
Great video! Full game commentaries is my favorite series :)
Thanks for sharing, this was a great game.
Thank you mister Redmond.
I enjoy these reviews of your personal historical games. I could include basically all games before AI in that category but I think that would not be appropiate for the really old games.
Great lecture.
Really enjoyed this, thank you Michael. As a fellow Redmond I want to maintain the high standard of the Redmond name in Go lol. I noticed your middle name is also Irish, so I assume Irish ancestry. I'm from Dublin.
In chess I saw when computers became "the thing", like a new Moses with 10 new commandments. But then, people started saying: "well, the computer says this variation is good, but I don't like to play that kind of resulting position. So, I'll do differently." That is because people realized that the evaluation was based in a certain calculation power impossible to any human (which includes the opponent!) and in order for it to work they had to memorize some strange moves that made no apparent sense. But in Go, it looks like top players just don't care. Maybe it is the "sensei culture" so present in the East, only now replaced by AI as a new sensei.
Maybe. Our understanding of the opening was always a bit fuzzy, and I think people realize that it is impossible to reproduce the calculation of a computer, but are enjoying the challenge of these new openings. We will surely see more developments in the AIs and the way people use them and I hesitate to predict if the trend will continue indefinitely or not.
@@MichaelRedmondsGoTV I am more familiar with shogi (Japanese chess) than chess, thus I compare go to shogi. I feel that in go it is often much easier to learn from AI plays than in shogi, because "many fight patterns are quite local and many variations are not too long to follow". Many variations of direct 33 josekis for instance can be evaluated without the global situation elsewhere. Traditional osae with only two stones and extension instead of double hane looks bad almost always regardless of the global board position. In this sense, AI josekis are simply new variations proposed by new stronger players. Nothing qualitatively different from the new josekis before AIs, like Dosaku presenting new josekis in this sense.
In shogi (or chess), it looks much harder for a human to play like AlphaZero, probably because the patterns, if any, are essentially global and the variations are extremely high with extremely risky moves abundant in checkmate bombs. Let alone mimicking, it is difficult to abstract the patterns like joseki in the first place. It is true that some AI-originated human implementable openings have become so popular as one of them, Elmo enclosure, to receive the New Move Award from Shogi Association. However, the general trend is that the top pros do not simply mimic AI openings, and I elaborate the approach below. By the way, there is of course sensei culture also in shogi, but still the introduction of AI moves in human games seems very different from that in go.
Also, the context looks very much different before and after the introduction of open-source AIs. Today, as you explain in other videos, the players can explore variations deeply with the assistance of AIs and it looks quite far from simply mimicking the AI moves. Even when a human-created variation loses 1 or 2 % of the winning percentage, it is usually much more important that the opponent is not familiar with the variation. Thus, particularly the world ranking No 1 Shin Jinseo seems to have created many original moves that do not appear in AI vs AI game records. The players simply playing as Marcelo suggested "mimicking AIs without deep understanding", I feel, generally do not get good results. The use of AIs today by top pros as I understand are more like conventional study groups (kenkyukai), only that other group members are the AIs. This study group approach seems to be adopted by the majority of top shogi players -- you test your hypotheses of human implementable strategies using AIs and get advice from AIs to review your games.
I have some additional comments regarding this within chess. In go, as I mentioned, it looks easier than in shogi to extract local play patterns per se -- copy-pasting the exact play variations pay off in many local fights. In chess, it looks as though more subtle fuzzy learning of playstyle looks much more fruitful. Many chess players were deeply fascinated by the new dynamic style of play by AlphaZero. Gary Kasparov may be one of the first to mention this regarding the game records published in Science Journal paper. But there are others like the grandmaster Karlsen as well ( ruclips.net/video/uPUEq8d73JI/видео.html ). AlphaZero seems to put much more importance on the piece "positions" than the piece values independent of the position. The conventional rule of thumb before AlphaZero was that piece values can be determined pretty much independent of the position, since in chess pieces, especially queen, can move very mobile all around the board. AlphaZero's playstyle revealed this is not at all the case. You want to maximize the mobility of your pieces and kill the mobility of your opponents' pieces, and for this purpose it makes full sense to sacrifice your pieces flexibly, just like sacrificing stones in go. I do not know whether AlphaZero leads to new opening books, but it certainly fundamentally changed the way the top players play chess in the world.
In shogi, as far as my understanding goes, AlphaZero plays very much like the above, trying to eliminate the opponent's effective moves and maximize your effective moves, the strategy long known in reversi (Othello). However, I know almost no human pro that implements this style to a high level at this moment. There is a player Chida Shota who's main training method is to grasp the subtle patterns of AIs through sparring, and I do feel a lot of AI flavor in his moves. Another player Toyoshima Masayuki, the current Meijin and Ryuou (the two biggest titles), is also well known for his having changed the playstyle fundamentally through sparring with an AI program. However, since AlphaZero is not open source, no one has been able to spar with AlphaZero itself.
Thus, like human cognition in general, it is quite often much more effective to grasp the fuzzy features underlying the playstyle of AIs than the concrete moves appearing in game records. In go, when people start thinking that the local exact patterns turn out not to be effective in learning, it will not only be the matter of mimicking AIs, but it will mean the fundamental change in learning go. The next generation genius kids may possibly come to learn value network globally, like AIs, rather than trying to learn precise local patterns like joseki. Shin Jinseo is apparently like that already to some extent -- he is able to play really surprising moves suggested by AIs but no human other than himself.
In fact, computers have completely changed chess. Even the rules have been changed. Adjourned games have been abolished, and today nobody is allowed to bring a mobile phone or the like into the tournament hall.
Games not only for the World Chess Championship are played without any of the players calculating even one move themselves. In some cases, 100% of the prepared analyses are played. The former world champion Visnaswathan Anand was feared for this.
Garry Kasparov also.
I´m an Amateur with about 2300 Elo, and I also lost games, where my opponent simply didnt need to make a single move by his own.
Just think of the chess databases with more than 8.000.000 games. The influence of Computers is still very little in Go compared with chess. I hope that this will remain so for a long time to come.
@@matthiasrichter3112 I would like that too, but I doubt that it will last more than a couple of years. Technology develops faster and faster every year, so I think it will not be long until Go reaches the situation chess faces today. The good thing is that fuseki is very different from openings in chess, so "preparation" in a chess sense is more difficult to achieve.
cool thank you
This game does not look at all like the O Meien I am familiar with. It is interesting to learn that he played much more conventionally than he has been since his top days to today.
With the use of AI, where stands the value of komi these days? Nice game revue -- made me nostalgic for the old days.
Most AIs are using 7.5 Komi and giving White an advantage but that evaluation is not always the case and depends on which program one uses.
@@MichaelRedmondsGoTV is it possible that Komi should be something like 7.0?
@@Herv3 While possibility of Jigo is kinnda cool in amateur games I think it would make pro tournaments and title matches difficult...
It is sad to see computer's influence on fuseki pratically wiping out certain openings and making fuseki more limited to certain patterns related to hardcore territorial strategy. The sanrensei and chinese are now virtually nonexistent in top play. Takemiya's style, just to give one example, is a museum thing now.
Don't be too sad-there's plenty of richness and diversity in the AI style fuseki, too. Also, let's not forget that just 10 years ago, basically half of the all games consisted of either the Chinese fuseki or Orthodox fuseki.
I had some empathy for your other comments relating chess to go. Now, I think you are terribly wrong in really toooooo many ways Marcelo, blaming AIs for all the trends. As I already responded quite thoroughly to your last comment related to chess, not too many trends are qualitatively different from those before the AIs. As I elaborate below, the boring trend can be much more attributed to the fierce winner take it all environment in Chinese and Korean go than AIs.
For the very example of sanrensei, it was already pretty much the museum thing long long before the AIs. Takemiya himself gave up playing sanrensei most of the time, though he continued to love nirensei. Nobody could ever show effective play with sanrensei against several modern responses even before the AIs.
And, quite the opposite of what you claim, the Chinese style used to be much more boring than AI style fuseki of today. Really very high percentage of top pros adopted some of Chinese style fuseki for quite a while. Also, since the variety within Chinese style was much less than that within AI style fuseki, there were many games sharing something like 30 moves. It was like the player edging in some local sequence study within Chinese style had better results, similar may be said regarding the notorious avalanche joseki whose variation lengths are like over 30 moves. It is quite natural that the top players try to play according to the modern dominant strategy of the era, unless you can come up with a better strategy of your own. Half retired kind of Japanese players like old Takemiya that can earn a decent amount of money without very high winning percentage can do whatever they like, but especially the Korean and Chinese top pros, whose fame and income depended essentially on the results of international tournaments, have had to rely on the safe openings.
You are also darn wrong thinking that AI strategies are more territory oriented than the traditional human's. If you look at the handicap games of Katago, a significant percentage of the game records show that Katago dominates the center despite giving three to four stones handicap to Fox 9ds. It is only that sanrensei for instance is not at all an effective strategy for building thickness. If AIs really love territory in the first place, why do they most often play star points in the first place????? They welcome the 33 invasions, which can be avoided if you started with komoku or 33. In fact, today, exactly due to the influence of AIs, komoku playing became much much more seldom, which seems to show that the top pros learned from AIs that too much territory orientation may be too thin in the long term.
The following may be a best example that the World Ranking No 1 (2020) can play with a neo cosmic style, depending upon the situation.
ruclips.net/video/VXJ48f6nPnY/видео.html
The point is that you do not force yourself into cosmic style. You always try to be as natural as you can and sometimes the game may result in cosmic style.
@@decidrophob When I wrote the comments you read, it was more to express my feeling on the subject of how AIs impact the way top players approach competitive games (go, chess etc.) and how I feel about the new paths resulting from that. I was not imagining that someone would write a ‘treatise’ in response to it.
I can’t deny that players (of all levels) want to win and that they want to learn ways to maximize their chances of doing so. The will to win is an integral part of their activity. Trends, new joseki, new strategies always change over decades. New ideas take over older concepts as a new understanding of how to proceed. It was always like that. My sadness comes from the fact that (as it happened to chess 20 years ago), because of AIs (but not only because of it), go is going now through a phase when some of its ‘humanness’ is lost in a sense that all the comments/ideas/guidelines coming from players and analysts revolve around “AI X says it is good”, “AI Z thinks this is better” etc. etc. And that is my main point. Until now, the game was developed by people, who impart to it a little bit of pragmatism, but also a little bit of emotion, a little bit of human imperfection. Every new trend brings changes to what is currently done; some variations vanish, some new ones appear and so on. Maybe it will happen to sanrensei what happened to the King’s gambit and similar openings in chess. Nobody plays them at the highest level anymore (I personally think it sucks). But one way or another, the changes were proposed by people, not by people trying to play like a computer program just because it wins games easily.
I don’t crusade against it, and, to some extent (mind that I’m only a 10k who learned the game two years ago) I also try to embrace it when I can, but only as a means to win, not because I think that it is emotionally pleasing.
To me, the games of players like Dosaku, Go Seigen, Takemiya Masaki, Cho Chikun, Kobayashi Koichi, Lee Changho and Cho Hun-hyun are great because they carry within them the will of those players to bend the rules, each one in a different way according to their own personalities and creativity, and not by the dry use of some sort of ‘computer’s pragmatism’.
@@decidrophob Answering this comment, and adding a bit more on my newest comment above, this “cosmic style” will still appear from time to time, as much as sometimes someone uses the Bishop’s opening or the Two knight’s defense. But it will be more and more associated to a “weapon to throw the opponent off balance” or a “momentary thing” than to a style on which a top player would rely on to win titles, and that is because people will trust less and less it is good to win on a regular bases (inspiring less and less analysis and developments). Shin Jinseo is so strong right now that he can play 10-10 on the first move and probably still win, but that doesn’t mean it will become a ‘honte fuseki’.
My addition to my last comment is that AIs play 4-4 points just as a means of accelerating development, much in a way Go Seigen and many other territorial players did since the 1930s. It is just a sort of springboard. There is a game I love by Kobayashi Koichi against Takemiya (Judan title, 3 april 1986, Go World n.45 p. 13) in which he plays the sanrensei as Black but aiming to develop fast and not to use it to develop a moyo as Takemiya himself would do. AIs are territorial, indeed. That is why, for example, they invade the corner so early. They like to grab some sure profit because their computational power allows them to calculate how to deal with any wall created by the opponent in the process.
By the way Master Redmond 9p Are you already retired at Go?
No
Yeah. About openings. What the go players and computer scientists do not seem to understand is that the algorithms are following the road they are coded. The classic Chinese opening is still a winning opening, but not if your "brain" (heuristic code) is optimized to fight with a different opening. Why is that? It's because no "AI" is actually conscious, they have already forgotten how they opened after a few moves, and cannot coherently in language explain their "thought processes" (they can core dump their code on you, but that is no explanation) and they have no cognitive strategy, they only act _as if_ they have subjective strategy, but pseudo-random number generators in coded heuristics is not subjective cognitive, it is fake-cognition. Sure, it is powerful in a game. But so what? The beauty of go (or any competitive game) is only evident when you lose.
Put another way, if I were a 9-dan Pro, I'd find a delicious thrill in beating AI with a Chinese opening, and if I lose by just the komi I'd still get a thrill. I'd be frickin ecstatic and treat myself to Mr Jiro's Sushi.