Some people like to play chess, where one must know the simple rules and the objective of the game. But what about the ancient Chinese board game, which has evolved over the course of many centuries, known as Wei-chi or Go? Can we identify simple rules and the objective of the game? Possibly, but it is exponentially more difficult. Interestingly though, we can view Wei-chi as a kind of artificial intelligence (AI), because we cannot easily explain the intelligence and thinking of it.
Wei-chi provides an operational framework for both offensive and defensive operations equally well. Due to its adaptability, Wei-chi has been applied in many different scenarios throughout history. In this way, we can see how Wei-chi has found usefulness even outside of its original intent as a board game. This is ideal when considering that armed forces must be prepared to confront adversaries from any domain, within the multi-domain operations (MDO) environment, who have increased access into other domains due to technological developments such as unmanned aerial vehicles (UAVs) or cyberwarfare capabilities.
“Wei-chi is employed operationally through its pattern matching process; we extract patterns from the known environment and compare them with patterns derived from our understanding of the adversary. This pattern matching process provides us the ability to extrapolate future actions based upon past behaviors.” — Dr. David Lai
Through its adaptation to both modern and ancient settings, Wei-chi provides insights into MDO that are important for military commanders. A Wei-chi board is separated into nine regions, with no pieces on the board at the beginning of the game. Players play on vacant intersections or points on the board. Each player has Wei-chi pieces: one black and one white. The Wei-chi pieces move around and are captured by surrounding an opponent’s Wei-chi piece between two Wei-chi pieces of the same color or bounded on three sides by Wei-chi pieces plus a corner point located between one Wei-chi piece and another. If a player has no legal moves anywhere on the board, we refer to this as “check” in Wei-chi terms. Similar to how the Chinese use the “Three Warfares” strategy of the People’s Liberation Army (PLA): Public Opinion Warfare, Psychological Warfare, and Legal Warfare.
Three examples of how the Chinese use their “Three Warfares” strategy are:
1. South China Sea. From 2013, China has been steadily reclaiming and fortifying islands in the South China Sea. The reclamation efforts and deployment of military assets on the islands illustrate China’s attempts to undermine the psychological capacity of the other claimants (Vietnam, Philippines, Brunei, and Malaysia) to resist its own claims. The navy has also used maritime militia in the area to sow confusion. It has implemented an aggressive messaging campaign with diplomatic pressure, news media, and other media to promote narratives emphasizing the validity of its claim and warning others against hostility. China has also tried to advance stories that portray China as a champion of international law in the South China Sea, as well as its own interpretations of international law, in order to challenge other countries’ positions and undermine the arbitration process.
2. COVID-19. In 2020, Chinese lawyers sued the United States (U.S.), claiming the U.S. was covering up COVID-19, an example of three warfares being conducted as if it were a military operation.
3. Confucian Institutes. Confucian Institutes are cultural and educational institutions dedicated to the promotion of Confucianism and Chinese culture. The institutes provide instruction on such topics as Chinese history, philosophy, calligraphy, and painting; traditional music; and Tai Chi Chuan. It’s no secret that China has military ambitions, but the Confucian Institutes are really about spying. Like leaving behind military personnel who can blend in with local populations, Confucian Institutes are a covert tool of military intelligence gathering. Espionage is all about getting information without being noticed and these institutes offer an excellent cover for Chinese espionage efforts.
Pattern-Matching in MDO
MDO is a military concept for multi-service, multi-force, and multinational operations against a highly integrated agile adversary within the land-air-sea-cyberspace-space domains, as well as the electromagnetic spectrum (EMS) and information environments. It requires the military to draw on favorable conditions in the beginning stages of a crisis. MDO seeks to deliver multiple dilemmas to an enemy by employing capabilities across these domains in unison. These dilemmas are created by applying all elements of national power (diplomatic, informational, military, economic) to achieve goals without having to rely on brute force or overmatch alone.
Wei-chi pattern matching would be used as an operational framework to generate predictive analysis which can be used for early warning purposes or to identify opportunities within the MDO environment. It seeks to imprint enemy capabilities with specific patterns that Wei-chi pattern-matching can then associate with moves and pieces. Evaluation of Wei-chi patterns would take the form of moves, pieces and board positions in order to yield a higher degree of predictive analysis in MDO.
The Wei-chi board is used as an analogy for the operating environment; Wei-chi pieces are used as an analogy for friendly and enemy forces; Wei-chi moves are used as an analogy for threats. The Wei-chi board is divided into regions that represent multiple domains (e.g., maritime, air, land, cyberspace, and space). Each region has its own value assigned to it by identifying its importance to friendly forces relative to the Wei-chi piece’s position.
The Wei-chi board is a reflection of friendly and enemy capabilities, which we will represent with Wei-chi pieces. Each Wei-chi piece represents a specific type of force within a single domain, for example, an amphibious assault ship, tank, etc., again relative to its position on the Wei-chi board (e.g., where it is positioned). Wei-chi moves are representative of threats, such as submarines in choke points or landmines along bridge approaches.
The Wei-chi pattern matching model will use two concepts: Wei-chi pieces and Wei-chi moves; these components will be used to derive patterns and make associations between multiple domains and Wei-chi board positions.
Pattern-matching is like a lantern in that: it can dispel darkness and reveal hidden connections. It sheds light on the darkness so you can see what you’re doing. Pattern-matching is defined as the simultaneous identification of several known patterns present in data obtained from sensors, satellite imagery, human intelligence assets, etc.
Pattern-matching is fundamental within Wei-chi for anticipating the future actions of the enemy based upon his past pieces and moves. In Wei-chi, matches are played to completion so there is only one winner. The same principle should hold true for Wei-chi pattern matching: if we can anticipate what we expect from an adversary in a specific situation then we will be able to arrive at a winning state through a series of Wei-chi moves that seek to minimize exposure and/or vulnerabilities while maximizing relative strengths against identified threats.
A Wei-chi piece can be broken up into subunits which would allow for more distinct positioning opportunities. This would be what we call a smaller unit within a large unit or a higher level organization inside a larger organization. For example, many countries have several different land forces that they employ in battle such as infantry, armor, artillery, etc. These components of Wei-chi pieces would allow us to create Wei-chi moves which we can then measure with Wei-chi pattern matching. A Wei-chi pattern is an association between Wei-chi pieces and Wei-chi board positions, in our example, we would consider a Wei-chi pattern as a maritime threat and the specific location of the maritime threat (e.g., in choke points or in shallow water).
The Wei-chi board is part of an MDO environment consisting of various domains. The Wei-chi pieces represent threats that we would like to measure and Wei-chi patterns are associations between Wei-chi board positions and Wei-chi moves.
A Wei-chi pattern is simply an association between Wei-chi board positions and Wei-chi moves; it is based on land/maritime correlation. A Wei-chi pattern can be either a Wei-chi piece or positioning, positioning just means the general position of a piece (e.g, close range). If we consider Wei-chi as threat indicators then we would look at the land/maritime correlations out there that could impact our military operations.
Pattern-matching is also used to identify predictive indicators of the enemy’s capability to allow for early warning purposes. If Wei-chi patterns are associated with moves, we can predict Wei-chi patterns even before we make moves; this would be one of the primary goals of Wei-chi pattern matching. This would allow us to predict enemy behavior and provide opportunities for friendly forces to act first in order to take advantage of favorable conditions (e.g., crossing the border into Syria while avoiding detection).
Stages in this context are imaginary stages. Wei-chi could be used as a model to develop a pattern-matching algorithm and decision-making framework in MDO that allows the U.S. military to draw on favorable conditions in the beginning stages, and before, a conflict emerges. That is, we use Wei-chi as the pattern-matching algorithm for MDO by allowing Wei-chi pieces to represent one of many assets in each domain at each decision point, similar to how Wei-chi moves are played on intersections or points on the board. The Wei-chi board represents multiple domains at which we are allowed to operate. We could also use this technique when standing off against an adversary who is confined within several domains.
If we consider MDO operating environments with distinct geographical spaces, assets, and operators, then we must consider how our operators can apply capabilities effectively across these different environments without becoming constrained by organizational stovepipes. For example, we would use this in the early stages of MDO operations to gain an advantage by finding and exploiting connections between domains. We consider Wei-chi within an MDO context because we must consider how we can apply capabilities across disparate spaces before, during, or after a conflict emerges.
A Wei-chi algorithm consists of several stages. There are two sides to this game: Wei-chi players and their adversaries. One seeks to detect critical patterns that they can exploit throughout each stage while their adversary is simply trying to delay the final pattern-matching if at all possible. As with most games involving strategy, there are optimal strategies for both sides so long as they fully understand what options are available to their opponent.
Wei-chi is not meant to be used as a game in our algorithm because we are not presented with all of the same restrictions. Instead, we may use Wei-chi pieces to represent one of many assets in each domain at each decision point.
Shi or Energy
Let’s examine how to apply this algorithm against an adversary confined within several domains.
Domain 1: The Wei-chi board represents land warfare.
We allocate Wei-chi pieces accordingly within each domain representing multiple assets, e.g., an M1A2 tank.
Domain 2: The Wei-chi board represents air warfare.
We allocate Wei-chi pieces accordingly within each domain representing multiple assets, e.g., an F-22 Raptor.
Domain 3: The Wei-chi board represents space warfare.
We allocate Wei-chi pieces accordingly within each domain representing multiple assets, e.g., a Falcon 9 rocket.
We may also look at Wei-chi pieces on our board as multiple domains that we must touch in order to achieve victory within the game (or conflict we’re striving for) we are fighting.
The fundamental element within the algorithm is to detect patterns without revealing their own pieces while their adversary tries to delay them by attempting to move pieces around and potentially denying certain moves. If they make a mistake, then there is an opportunity for exploitation, but this also reveals information. This allows the algorithm to predict shi (pronounced shur or sure), which we could simply summarize as energy or energy potential. Where energy is defined as the capacity to do work. Yet, the key is knowing when or by how much, or even if you need to expend energy.
We can turn to physics for two simple equations that describe energy; thus, they describe shi:
- Kinetic Energy (K.E.). The K.E. of an object is the energy it possesses due to its motion.
2. W (Work)= F (Force) x D (Displacement). Work is a transfer of energy. When we use our energy on an object, we send some of it to it. This implies that the extra energy put into an object is known as “work.”
Elements of Wei-Chi Algorithm
Here we introduce elements from Wei-chi that will be used in our pattern-matching algorithm. Rather than use Wei-chi pieces as we would in the game to capture territory, we are using them as assets within each domain. Players are attempting to identify patterns without revealing their own assets or what they plan on doing, which could allow adversaries an opportunity for exploitation. This is similar to how we must think tactically in combat operations though we do not have perfect information either.
One of the most important elements of this pattern matching algorithm is that we have multi-domain operators who can apply capabilities across multiple operational spaces before, during, or after a conflict emerges. Those who are confined within specific domains cannot visualize operations between these spaces because they are overly constrained by organizational stovepipes.
The main dilemma for decision-makers is determining which patterns are likely to be most advantageous in a given situation. In Wei-chi, we consider all patterns even if we do not know what they mean or how we can use them. We train decision-makers to use the whole board instead of staying within organizational stovepipes. For this reason, Wei-chi is an excellent framework for developing this algorithm because it has many concepts that we would like to see in our pattern-matching strategy.
Wei-Chi Pattern Matching Algorithm
Adversaries will use Wei-chi-like tactics in MDO. This is why we must develop algorithms that consider all possible patterns regardless of whether we can visualize them or not. This challenge is even greater for Wei-chi players because we do not know what the adversaries plan to do as they might change their minds at any moment.
So now let’s introduce the Wei-chi algorithm that we are developing for MDO. There are five basic patterns to Wei-chi which are categorized into three broad categories: offensive, defense, and miscellaneous. The offensive tactics are referred to as “harms” while defensive techniques are called “guards.” The miscellaneous category contains moves that do not clearly fall into either category.
Offensive tactics refer to direct attacks to gain territory or indirectly attack reducing an opponent’s territory. These can be classified into encirclements (moves to make a territory smaller by completely surrounding an area), strikes (moves aimed at cutting up the opponent’s pieces or groups), and sacrifices (involves putting one’s own pieces at risk, usually for tactical reasons).
The defensive techniques are referred to as “guards.” These maneuvers block off attacker’s access to certain parts, weaken their structures (groups), and produce situations providing opportunities for counter-attacks. Guards include extensions (placing a piece in a certain spot to strengthen it, typically because it makes them harder to attack), weaves (moves that make it difficult for the opponent to capture a piece by placing a string of pieces between an enemy and a target), and multi-josekis (when many guards are used at once).
Miscellaneous refers to moves that do not clearly fall into either category which include “pulling back” (moving a piece away from the center), attacking weaknesses in the enemy’s structure, and capturing prisoners. Pulling back is done for development purposes while attacking weaknesses in the enemy’s structure are usually done just before starting an attack on their formation. Capturing prisoners is typically part of encirclement strategy as it gives you more room to work with.
To do this, the algorithm must be able to:
1) Pattern match— where we develop our own patterns without revealing information about how we plan to use them. Develop patterns through training and pre-positioning our assets to act like adversaries (i.e. red teaming). We want to observe how they move around without revealing any information about where our own pieces are or what we plan on doing.
2) Deception — where we consider what others are thinking when they are trying to match our patterns.
3) Exploitation — when we focus on the adversary’s mistakes, forcing them to reveal their hand or be in unfavorable situations where we can take advantage by taking away their ability to move around. For example, if an adversary does not know that we are intercepting messages then they might leave a message we can intercept. If we make this behavior pattern obvious, then we can exploit it by tracking down the adversary’s other messages leaving us with the advantage because we know what they do not.
Wei-Chi Intelligence is the Chinese Room
“From one thing, know ten thousand things. When you attain the Way of strategy there will not be one thing you cannot see.” — Miyamoto Musashi
The Chinese Room Argument, which is often referred to as the Chinese Room, was originally presented in a 1980 paper by American philosopher John Searle. It attempts to illustrate that a program cannot simulate a real understanding of a language, regardless of how intelligently or human-like the program may make its responses appear. The argument has elicited a wide variety of reactions from both philosophers and computer scientists; it has become one of the central issues in the philosophy of artificial intelligence (AI).
The Wei-chi algorithm we are developing for MDO will allow us to simulate Wei-chi and we can therefore apply Wei-chi algorithms across all domains without requiring operators to understand everything within each domain. This is similar to how the Chinese Room’s “intelligent room” does not need to understand Chinese in order to be able to carry out conversations with you in Chinese.
Wei-chi is a good example of pattern-matching across multiple domains in MDO. We can imagine Wei-chi strategies to be applied when we are attacking an adversary, defending against an adversary, or when we are avoiding combat with adversaries entirely.
Yet, the development of a complex algorithm, such as the one discussed here, illustrates the importance of integrating AI with human thinking, as a program cannot simulate real understanding, regardless of the intelligence. However, the combination of the two: human and artificial intelligence, provides us a greater chance of knowing ten thousand things by simply knowing one. Where we do not need to understand the meaning of the ten thousand things, just the identification of the one thing that can influence them all. This is fundamental to, not only this algorithm but to the concept shi.
For more reading on Wei-chi, read Uncovering Hidden Patterns of Thought in War: Wei-Chi versus Chess.