Cybernetics and systems theory are fundamental disciplines for understanding the changes in things, and understanding these fundamental disciplines can help us better understand the world.
This article is from BEDROCK member Jimmy and his team's discussion records.
Control and Feedback
The study of control theory and systems theory began with the possibility space.
All possible changes in things constitute their possibility space, and things that are outside this space are impossible to achieve, such as turning stones into gold (not considering nuclear fusion). The goal of control theory is to control the conditions so that things change towards the desired goal, that is, control is the narrowing of the possibility space.
According to the method of control conditions, it can be divided into random control, memory-based control, conjugate control, and so on.
Random control is based on luck, such as finding the key that can open the door in a bunch of keys, which can only be tried one by one. It is worth noting that the correct conditions must be within the range of attempts. If the correct key is not in the keychain, no matter how many times you try, it will not work. The characteristic of random control is that the possibility space of the system only narrows when the target value is reached, and the possibility space does not shrink when the target value is not reached.
If you can remember which keys have been tried is incorrect, exclude them when trying again, and this is memory-based control. The key point of this method is to carefully distinguish, if you classify the correct key as an incorrect one, it will never be possible to find the correct key.
Conjugate control refers to expanding the scope of the control condition by means of transformation and substitution. For example, the weighing of elephants by Cao Chong is to convert the weight of elephants into the weight of stones, making it measurable.
Negative feedback control refers to continuously measuring the difference between the current state and the target, and taking measures to reduce this difference continuously, so that the accumulated controls can ultimately achieve the goal. People grabbing things with their hands, eagles grabbing rabbits, and missiles destroying attack targets all use negative feedback control.
The biggest advantage of negative feedback control is that it can accumulate the results of multiple controls, which is equivalent to amplifying the control ability. For individuals, don't make too precise plans in advance, because objective things are always changing, and you can only take one step at a time, constantly adjusting your actions based on the latest situation, and gradually approaching the goal.
In contrast, positive feedback is the process of gradually moving away from the goal (or the equilibrium state, the initial state), such as an arms race. Positive feedback is often referred to as a vicious cycle because it easily leads to the system getting out of control and collapsing.
BR Research：
How to identify positive and negative feedback and how to obtain core variable information? The difficulty of operation in reality is very high, such as how to define the R0 value of a liquidity crisis, and the identification of critical values is also a problem.
How to identify positive and negative feedback and how to obtain core variable information? The difficulty of operation in reality is very high, such as how to define the R0 value of a liquidity crisis, and the identification of critical values is also a problem.
Positive feedback may not ultimately lead to negative feedback either. For example, in a downward cycle, if a company goes bankrupt, there is no opportunity for negative feedback. If it falls into a liquidity trap, there may be no external force to generate new negative feedback.
In investment, cyclical product investment utilizes negative feedback, while studying competitive advantages utilizes positive feedback.
Information, Thinking, and Organization
"Knowing" is actually the process in our minds of increasing or decreasing the possibility space of things changing. In most cases, the possibility space decreases.
Information is the negative logarithm of the ratio of the change in possibility space. When the possibility space is reduced, the information content is positive; when the possibility space remains unchanged, the information content is zero; and when the possibility space increases, the information content is negative.
BR Research:
For example, in the original belief that the market operates under rules, but issues such as domestic education and platform governance (negative information) occur, the possibility space for the future market to operate according to market rules increases. The possibility space for individuals and companies to arrange consumption, production, and investment also increases.
The author later wrote a book about "China's super stable structure." Most of the feudal dynasty's ruling structure was stable, but occasionally it would collapse. This intermittent collapse mechanism is actually the basis of China's overall stable structure for thousands of years (unification, centralization, etc.).
The form of information in the transmission process is called a signal. Different forms of signal transmission have different capabilities, such as light, sound, electromagnetic waves, etc. The transmission paths are also vastly different. The transmission of information is the transfer of changes in possibility space.
Sufficient information is needed to implement control. A mute person is not necessarily deaf, but congenital deafness is highly likely to cause muteness because they have not collected enough information to control their vocal organs. The inability of people to control their own heartbeat and blood pressure is also due to similar reasons. If blood pressure is converted into visual signals and trained, one can learn to control their blood pressure.
The same words have different meanings for different people. This phenomenon is called the subjectivity of information. The human brain often subjectively processes received information, and human understanding of objective events is generally different.
The transmission of information involves channel and channel capacity issues. The maximum amount of information that can be transmitted by a channel in a unit of time is the channel capacity of that channel. Channel capacity depends on the speed of the channel and the control ability of people, as well as the size of the possibility space of the channel and people's understanding of it. When a certain amount of information needs to be transmitted in a unit of time, the selected channel capacity should ideally be equal to the amount of information to be transmitted. For example, if there are hundreds of colored lights at a crossroads, it may be more difficult to identify effective information.
BR Research：
It is effective to use in work to reduce the expression of invalid information in communication, and it is important to be aware that invalid information can interfere with the reception of valid information. Therefore, communication should be as concise as possible, and all information should be relevant.
Interference in information transmission can be classified into three types: control interference occurs during the controllable state process of the communication channel, natural interference or noise occurs during the natural transmission of the signal, or external factors affect the discernible state of the channel, and subjective interference occurs during the reception of the signal.
The process of eliminating interference is called filtering.
The most intuitive filtering method is to repeat the information along the same channel. This method is effective in excluding random and occasional interference, but it cannot exclude systematic and regular interference that occurs within the same channel.
Another better filtering method is to use completely different channels to transmit the same information, and then compare and analyze the results. During World War II, the Allies received the message that "the Germans cannot yet manufacture atomic weapons." At the same time, they received another message from an unrelated channel that "the Germans are using thorium (tu) in toothpaste." Thorium was a chemical element needed to manufacture atomic weapons at the time. From these two completely unrelated channels, the conclusion was confirmed that the Germans could not yet make atomic bombs. In addition, organisms that reproduce sexually ensure the effectiveness of DNA information transmission through dual channels (father and mother) compared to organisms that reproduce through mitosis.
Different channels refer to the various links of the channel composition being as different as possible. Otherwise, it is still possible to be affected by specific interference in those same links. In the story of Zeng Can's murder, the channels through which Zeng Mu received information were different people, but they had the same information source: a person with the same name as Zeng Can had committed murder.
Due to the existence of interference, the amount of information can only decrease during the transmission process and cannot increase. Generally speaking, the longer the transmission channel and the more links, the more likely it is to be affected by interference.
BR Research：
When conducting research, it is important to compare information from multiple sources, such as experts, companies, peers/competitors, upstream/downstream, top-down, etc.
If you have a positive outlook, it is also important to value information that is negative.
To eliminate interference, the "impedance filtering method" is often used. Impedance filtering is to find the essential difference between interference and the signal carrying information, and to use a device or method to block the interference signal while allowing the signal carrying information to pass smoothly. For example, a radio can receive signals within a wide frequency range, but through filtering, only the wireless frequency corresponding to a particular program is played.
BR Research：
Example: Browse-based IR vs. Search/Push IR
When the interference encountered during information transmission is mainly subjective interference, the filtering method commonly used is to transmit the information and its importance together. For example, when shouting for help, the volume represents the importance of the information.
Feedback filtering: Using the useful signal received and the channel to interact with each other to block useless signals. For example, the human brain only focuses on the information of the visual focus among the many pieces of information collected by the eyes.
Synchronous filtering: Using the synchronization between the signal and the channel switch to filter. That is, the channel is only opened when the signal needs to be transmitted, and closed at other times.
In addition to distinguishing truth from falsehood, more advanced thinking patterns are also needed. For example, in military operations, some important intelligence is often difficult to obtain, and they can only be obtained through a few channels with significant interference. Some famous military theorists have pointed out that major military actions cannot be entirely dependent on intelligence, and commanders must adhere to their own beliefs.
BR Research：
When collecting information for investment, it is necessary to use the above methods to distinguish truth from falsehood as much as possible, while also relying on one's own common sense and logic to judge the truthfulness of the information.
There are several forms of information processing:
Syllogism, which consists of three simple judgments, each containing a certain amount of information. The essence of syllogism is to process the information of the major premise and minor premise into the conclusion.
Probability theory, which obtains useful information through mathematical calculations.
Free association: first, move from information A to another information B that has something in common with A, and then move from B to another information C that has something in common with B (but not necessarily with A), and so on.
In the thought process, co-constraint control is very important. From the perspective of control theory, a person's thinking space can be divided into two parts: the image space and the concept space. When a person is engaged in image thinking, the movement of information is in the image space. The concept space represents the range of information movement during abstract thinking. In fact, even the simplest reasoning involves the coordination of these two spaces. In this coordination, co-constraint control is very important. For example, when we hear that someone has taken poison, we immediately think: this person is going to die.
BR Research：
Investment is also an attempt to make judgments about the future based on the study of the current situation. Since the future has not yet happened, no one knows exactly what will happen. In control theory, this actually involves important probability judgments and conjugate control, and the process of imagining the future based on the present. However, this mapping process is actually full of possibilities for error, especially when projecting very long-term futures based on current conditions.
The process of organization is actually a process of a system excluding many other possible ways of connecting and only selecting one or several ways of connection from an unconnected state. For example, organizing a chaotic crowd into a queue. A system must acquire a certain amount of information in order to organize itself.
In an organization, such as human society, each individual is both an information source and receiver, and also a component of the social information channels. We must examine the entire control, feedback, and information transmission process of an object comprehensively. This is system theory.
System and its evolution
In systems and their evolution, new causal relationships arise in large systems:
Causal chains. By constantly exploring the reasons behind the reasons, one can get longer causal chains. On the one hand, in order to consider the overall situation, people have to include more and more factors into the object of study. On the other hand, the causal chains in nature have no end, and science must set appropriate limits for itself.
Probabilistic causality. Classic causality assumes that any cause necessarily leads to a certain result. But in fact, many connections between things in nature have randomness, and cause A may not necessarily lead to result B. For example, in quantum mechanics.
Mutual and self causality. For example, in an economic system, a lack of coal leads to the inability of a machinery factory to produce machinery, but a shortage of mining machines also leads to low coal production in the coal mine. In fact, when studying causal chains, we always find the closure of the causal chain. Causal chains that do not close almost do not exist in nature.
BR Research
Closure: has an endpoint. Can the final cause be found? The causal chain may not be a straight chain (with no end), but a network that can be closed.
It can also be understood as follows: in a closed system with limited variables, the causal chain is closed; in an open system, the causal chain is infinitely extended.
Causal network: an event may be the result of many causes working together, and this event may also cause many results. For example, ecology, economics, biological organizations, etc.
Establishing a relatively isolated system: First, when we trace along the causal long chain, we ignore those factors with small enough probability of influence and consider them as interference to the system and outside the system. Second, a relatively isolated system is as self-closed as possible, which is a mutual causality network. Third, based on our research purposes and the time scale of system changes, we capture the main variables of mutual causality and construct a system model. For example, if the time scale of the study is not very long, then we can study deer herds, forests, and predators as a system, excluding the effects of soil and climate.
By adopting a method of artificially defining a system, which is a scientific abstraction of the complex relationships between objective things, an infinite problem has been transformed into a finite problem for investigation.
A system will change from one stable state to another. If the new stable state has more than one possibility, it is called bifurcation. At this time, which stable structure the system evolves to will more or less have some randomness. When the system evolves and faces bifurcation, pure determinism is not applicable. Many times, small differences in initial conditions can lead to vastly different systems.
However, different stable states will eventually converge, which is called convergence. For example, most herbivores have evolved hooves.
Self-reproducing systems: Under certain conditions, the value of a variable increases faster as the value gets larger. Examples include nuclear explosions and avalanches.
Characteristics of self-reproducing systems:
There is usually a critical value for a self-reproducing process to occur, meaning that the system variable must be greater than this critical value.
There is an automatic growth chain with a causal relationship within the self-reproducing system.
Many self-reproducing systems are formed due to the disruption of negative feedback control mechanisms.
The phenomenon of variable growth in self-reproducing systems will not continue indefinitely and generally leads to system collapse, ultimately reaching a new stable state.
Self-organizing systems: The organizing process occurs automatically among a group of things or variables, without the need for external forces to intervene.
Characteristics:
There is initially an organizing core.
Self-organizing systems are unstable or semi-stable.
There is a causal selection chain within the self-organizing system.
The self-organizing process is irreversible.
Small differences in the self-organizing core can lead to significant differences in the final organization.
BR Research：
Self-organizing systems are a manifestation of positive feedback and have great power. Network effects are a form of self-organizing pattern, often becoming increasingly irreversible and difficult to replace by other networks once formed.
Self-organizing systems involve the automatic organization of a set of entities or variables without the need for external forces to intervene.
These systems have the following characteristics:
there is a core organization;
self-organizing systems are unstable or sub-stable;
there exists an automatic selection chain with causal relationships within the system;
the process of self-organization is irreversible;
small differences in the self-organizing core can lead to significant differences in the resulting large-scale organization.
BR Research：
Self-organizing systems are a manifestation of positive feedback and have tremendous power. Network effects are a form of self-organizing pattern that, once formed, becomes increasingly irreversible and difficult to replace with other networks.
Mathematical models of phase transitions can be categorized as leaps and gradual changes. For example, water freezing is a phase transition, while glass gradually hardens. The principle for distinguishing between leaps and gradual changes is whether the intermediate transition state is stable or not. If it is unstable, then it is a leap; if it is stable, then it is a gradual change. An avalanche is a leap, while a snowball rolling down a hill is a gradual change. An explosion is a leap, while combustion is a gradual change.
In the process of transitioning between two phases, two basic factors related to changes in conditions are important: factors that maintain the stability of the old phase and factors that establish the stability of the new phase. If the factors that establish the new phase are strengthened while the factors that maintain the old phase do not significantly weaken, the phase transition may occur through a leap; if the factors that establish the new phase are strengthened while the factors that maintain the old phase significantly weaken, the phase transition may occur gradually.
Phase transitions can be represented by a mutation function. When the number of controlled variables is no more than 4, and the number of state variables is no more than 2, there are 7 basic models. These are the most fundamental patterns. When there are more than 5 controlled variables that affect the mutation, there are infinite types of mutation models.
The phenomenon of over-correction has strict conditions and can only occur when the phase transition occurs through a leap. For example, water must exceed 100°C to boil, and water vapor must be below 100°C to condense. However, not all situations that involve leaps require over-correction.
Extreme coexistence: under certain conditions, the subsystems of a large system may simultaneously exist in completely different phases. For example, water has a coexistence region of two phases, liquid and solid can coexist, but there is no intermediate phase. In a small island with many storms, insects have only two evolutionary outcomes: either they have no wings and hide in the grass, or they have strong wings and can resist the strong winds.
BR Research：
Human society itself, market fluctuations, technological progress, and even psychological states, among other factors, may present gradual and qualitative changes. Even for paradigm shifts at relatively large levels, different patterns of change may exist.
Many growth investments are essentially investing in new quality factors that are currently underway, but it is important to beware that these new quality factors may be replaced by updated factors someday.
Many value investments essentially invest in relatively old quality factors that tend to be relatively stable (as they often occupy the dominant position and are in the mature stage of cash feedback). Even slow gradual declines can often be adequately anticipated. However, once they are qualitatively replaced by new quality factors in some imperceptible way (black swan events often exist in the form of sudden, serious deviations from the trend line), they may cause significant losses because they imply severe underpricing.
The Black Box Epistemology
Control theory regards the object of human knowledge and transformation as a black box. For a black box (object), we use a system composed of a set of observable and controllable variables to describe it.
Control theory believes that there are two different methods of understanding the black box object. One is the method of not opening the black box, and the other is the method of opening the black box. For example, a clock can be observed for its pointer movement pattern or can be dismantled to study the gear meshing relationship inside.
The hypothesis about the internal situation of the black box based on the external input-output is called a model in control, which is just an assumption and does not necessarily represent the actual structure inside the black box. When there is the ability to open the black box, the original model often needs to be improved, which is the process of scientific progress. However, the new system is still a black box. Therefore, people always have to use some methods that do not open the black box to study and solve problems.
Mao Zedong's "practice-theory-practice" model is a negative feedback control process that continuously modifies subjective knowledge based on the difference between knowledge and practice. However, it is subject to five limitations:
The limitations of observable and controllable variables. If these are not improved, the theory cannot progress.
Clarity of the theory. That is, the theory needs to provide a certain amount of information so that it can be compared with practical testing. This is equivalent to the falsifiability of science.
The speed of the model's approximation to objective truth. In order for a feedback adjustment to be effective, the speed of the feedback adjustment must be greater than the rate of change of the object. Otherwise, oscillations will occur, moving from one extreme to another and not achieving the purpose of effective control. For example, if the detection speed is slower than the rate of material change, the correct model cannot be obtained. In physics, there is a very important assumption that the law does not change with time and space.
Over-feedback. Any negative feedback control system will transform from a stable process that approaches the target step by step into an oscillation process if over-feedback occurs. For example, optics oscillated between the particle theory and the wave theory until quantum field theory provided a correct explanation of the nature of light.
Determinable conditions. The error between practical results and theoretical results must reflect the degree of approximation between the theory and objective truth. For example, Newton's law of universal gravitation was almost abandoned because the gravitational force calculated according to the measured meridian did not match his theory. While practice is the ultimate standard for testing truth, when determinable conditions are not satisfied, people also use some intermediate standards. For example, Einstein once proposed the theory of "guide field" to explain the duality of light, but this theory does not meet the law of conservation of energy, so Einstein never published it.
The fewer the conditions required for a law to hold, the more universal the law is. With the increasing ability of humans to control nature, the universality of their understanding of laws increases.
When humans seek greater control over nature, science arises, and those things that humans cannot control but attempt to control become the first objects of scientific research. Therefore, science has a center, a starting point, which is the human itself, and the batch of variables that he can control initially.
Inspiration for Investment
Control theory provides a new way of understanding the world from a different perspective, many of which concepts can be directly applied to investment. For example, the system often has a stable state but also oscillates around this stable state, just like the financial crisis in economic development that swings
BR Research：
Soros' reflexivity is to find non-steady-state positive feedback and make money from it. However, he did not enter early, but waited until the critical point appeared before entering the market.
Judging the critical point is not easy to identify during the process, and it is also a process of trial and error. Different processes have different methods for judging the critical point, and it can only be analyzed specifically for each problem.
Comments