Limitations in Perceiving Intelligence

This section further explores the limitations in perceiving intelligence. As intelligent agents, human individuals inevitably interact with, perceive, and engage with themselves and other intelligent agents, consuming a significant portion of their time and energy. This process includes, but is not limited to, observing and analyzing one’s own thoughts and opinions, interacting with others to integrate into society, living in a city, understanding national development, joining or managing a business, etc. In the Intelligenism theoretical framework, entities such as the human body, brain, other individuals, society, nations, businesses, forest ecosystems, or marine systems are also defined as intelligent agents.

Interactions, perceptions, and engagements between human individuals or other intelligent agents with themselves or other intelligent agents can be categorized into three types:

1. Intelligent agents observing, interacting with, or engaging with their own sub-level intelligent agents.
2. Intelligent agents, as sub-level components of a larger intelligent agent, observe, interact with, or engage with the larger intelligent agent.
3. Intelligent agents observing, interacting with, or engaging with other intelligent agents without a hierarchical relationship.

1）Intelligent agents observing, interacting with, or engaging with their own sub-level intelligent agents：

Human individuals cannot perceive the inner workings of their brains or the complete functioning of their bodily organs. For example, humans cannot perceive their digestive system’s absorption of food. This can be attributed to the evolutionary process, where humans did not develop sufficient neurons to transmit detailed status information from sub-level modules (bodily tissues). When perceiving their own operations, humans often rely on vague sensations, such as pain or itching, to receive feedback about the conditions of their sub-level systems. In many seemingly normal scenarios (where issues may not trigger pain), individuals typically cannot access detailed operational information or signals from other sub-level intelligent agents. This may stem from an evolutionary adaptation in biological organisms (humans) that prioritizes overall survival, as excessive information transmission to the brain could overload it. If humans cannot allocate sufficient decision-making resources to other external environments, they may lose competitive advantages in nature. During the process of generating thoughts or feelings, we cannot sense the detailed operational states of neurons (sub-level intelligent agents). We can only grasp some thought patterns that are close to the final answer.

In the scenario of riding a bicycle, it isn’t easy to articulate how balance is maintained; yet, as intelligent agents, we achieve it. During cycling, the human intelligent agent typically focuses on the process itself. Concentrating on how hands and feet exert force or move often disrupts balance. In this process, the muscles of the hands and feet may form habitual responses, and the perception of cycling remains incomplete; yet, this state enables effective cycling.

This phenomenon is not limited to cycling. In swimming, running, or esports, the mind often focuses on the overall state, and concentrating on specific details while maintaining good overall performance is challenging. Michael Polanyi, in his book Personal Knowledge, refers to this state as the internalization of knowledge.

In these examples, human thought is still part of the intelligent agent’s construction. Perception itself can be an output of a specific task for an intelligent agent. When we use our perception to sense overall operations, we influence the system’s functioning, altering the intelligent agent’s output and potentially affecting task execution. Similar to Heisenberg’s uncertainty principle in quantum mechanics, observation affects the quantum state, making it unobservable in its pre-observation state. Thus, when an intelligent agent is part of a higher intelligence degree intelligent agent or uses its own intelligence to perceive itself or its sub-level intelligent agents, it influences the observed intelligent agent, resulting in a state different from its pre-observation state.

This may be one reason why an intelligent agent cannot modify itself without external influence, as it cannot observe itself statically while maintaining its intelligent structure. Similarly, a person cannot perform brain surgery on themselves, as their thoughts and behaviors change during the process, which can affect the surgery.

2）Intelligent agents, as sub-level components of a larger intelligent agent, observing, interacting with, or engaging with the larger intelligent agent：

The second type of observation and interaction can be exemplified by stock investors predicting and trading in capital markets. As sub-level intelligent agents in the stock market, investors both participate in the market’s operations and observe, analyze, and predict it. While some participants can gain probabilistic advantages through analysis, the market’s fluctuations cannot be fully predicted, demonstrating the limitations of intelligent agents observing and interacting with a larger intelligent agent.

3）Intelligent agents observing, interacting with, or engaging with other intelligent agents without a hierarchical relationship：

In another scenario, if an intelligent agent observes another as an independent third party, it can do so without influencing the observed intelligent agent, provided the observed intelligent agent is unaware of the observation. For example, a human observing a computer-generated artificial intelligence program’s operations can study, modify, upgrade, or evaluate it without the program being aware of the observer’s presence. However, even in this case, the observer cannot fully grasp the entire process of intelligence, as the program undergoes extensive parameter adjustments and interactions among neurons and networks before producing results. While we may partially understand the significance of these interactions, the parameter settings of intelligent program networks are more a result of coherence than a specific substantive meaning.

The Presentation of the World: Coherence and Balance

I believe the world’s presentation is based on a continuous process of coherence and balance among various entities. In this ongoing presentation, since atoms, stones, trees, or humans can be broken down into smaller components, this construction—from smaller components aggregating into larger entities—can be seen as a form of organization. The coherence of these organizations with their external environment is a necessary condition for their stable existence. When most observed organizations exist stably, the world’s presentation can be considered to have reached a local balance. When this balance is disrupted, environmental changes may lead to the dissolution, reconstruction, or internal reshaping of organizations, achieving a new balance. This cycle of coherence, shaping, balance, imbalance, reconstruction, and rebalancing, over sufficient time, manifests as the “eternal” world before humanity.

Within this framework, any human theory or formula is an interpretation or expression of the world, but cannot represent the world itself. This aligns with the views presented in the “Philosophical Foundations of Intelligenism” chapter (referred to as the “Philosophy Chapter”), which argues that no theory can be asserted as absolute truth, but rather that every theory has theoretical adaptability based on specific individuals or organizations within particular external environments. Some theories have broader adaptability, while others are limited to specific scenarios. Before the emergence of geometry and physics, humans had been building houses for many years, indicating that these fields were not necessary conditions for house construction. While these disciplines later provided significant guidance for building more robust structures efficiently, trial-and-error and inductive methods had already enabled humans to build houses without a thorough understanding of geometry or physics. Modern studies of ancient houses built without such knowledge reveal they still adhered to geometric and physical principles.

In the context of artificial intelligence programs, consider a set of 50 points on an XY coordinate system, all lying on the line Y = AX + B. Humans can derive the formula Y = AX + B using two or more points to determine the exact positions of all 50 points. An artificial intelligence program, constructed with neurons and network layers, can use a subset of these points as a training set to adjust the parameters of its neurons. When given an X value from a non-training-set point, the program can output the correct Y value without deriving the formula Y = AX + B, instead relying on a set of neuron parameters. Similarly, ancient humans, without knowledge of physics or geometry, used a trial-and-error process akin to neural feedback and parameter adjustment to acquire house-building skills.

Billions of years ago, the world’s matter likely underwent a similar process of “trial, feedback, and adjustment” to achieve coherence and form the broader balanced state we now recognize as the universe.

As mentioned earlier, intelligence degree is an indicator of an intelligent agent’s ability to align with its environment. Ancient humans mastering house-building through trial and error was a process of environmental coherence (exploring and utilizing external material and environmental laws). A scientist exploring a theory is also a process of environmental coherence, which can be generalized as a manifestation of an intelligent agent’s intelligence. Extending this further, intelligence may have existed before humans or most organisms, as it is the habitual way in which the world’s entities achieve coherence and balance. When humans discovered and defined this process through subjective initiative, it became known as intelligence.