The paper discusses the concept of ontology, initially defining it as the philosophical study of existence and reality. It traces its integration into computer science back to Tom Gruber in the 1990s, who defined ontology in this context as an explicit conceptualization, making rules understandable to computers. This involves defining entities (e.g., "Human") with attributes (e.g., Name, Gender, Age) and individual instances (e.g., Kim Cheol-soo, 35, Male). This structure forms the basis of Knowledge Graphs, which have gained prominence due to their demonstrable ability to significantly improve Language AI accuracy and robustness, particularly for complex queries, by providing a structured knowledge base.

The core argument of the paper, however, challenges the sufficiency of traditional attribute-based ontologies. It poses the question: Do common attributes like name, gender, age, or even physical body parts truly capture the *essence* (본질) of a being, especially for AI understanding? The paper argues no, supporting this with two main points:

1. Agnosticism and Scientific Reality: It introduces agnosticism (불가지론), the idea that our cognitive system cannot grasp the true essence of things. This is illustrated scientifically using the hydrogen atom: if a hydrogen nucleus were the size of a basketball, its electron would be meters away, implying that atoms (and thus humans) are mostly empty space. The apparent solidity of our bodies is presented as an illusion created by our brain processing reflected visible light, which cannot penetrate the electromagnetic forces within atoms. Unlike visible light, X-rays can penetrate, revealing a different "reality" (bones). Therefore, our perception of a "human" is an image constructed by the brain, not the fundamental atomic reality. This suggests that attributing properties like "name" to this constructed image is itself a philosophical, rather than scientific, act, distant from the underlying "true" reality.

2. Logical Critique of Essential Attributes: Even without recourse to atomic physics, the paper logically dismantles the "essential" nature of common attributes.
   • Limbs: A person without limbs due to an accident is still fundamentally a person, demonstrating that limbs are not an essential attribute. An essential attribute, by definition, must be one whose absence annihilates the entity's fundamental nature.
   • Heart: While a heart is vital, its function can be temporarily replaced by an ECMO machine (Extracorporeal Membrane Oxygenation), an artificial heart-lung device. In such a scenario, the person's being is maintained even without a functioning heart. This implies that the heart itself is a *solution* to the problem of blood circulation, not the fundamental essence. The true essence is not the organ, but the *state* of addressing the problem of maintaining blood circulation.

This leads to the paper's core proposition for a more fundamental ontological definition, especially pertinent for AI: A being's essence is not defined by its static attributes or the *solutions* it possesses, but by the dynamic *state* of solving problems that, if unresolved, would lead to its non-existence.

For a human, this means focusing on "problems that make me disappear" (나를 사라지게 하는 문제들) and the *state* of continuously resolving them. Examples of such problems include:

• Physiological: cessation of blood circulation, breathing, excretion, immune function, or central nervous system activity.
• Cognitive/Social: inability to make judgments, severe cognitive impairment (e.g., dementia), or harmful communication patterns.

The individual "human" is thus conceived as the "sum total of states where solutions are continuously addressing problems that prevent non-existence." The traditional attributes (heart, limbs) are merely specific *solutions* provided by nature (or human technology) to these problems, not the problems themselves or the dynamic state of their resolution. The capacity of a solution (e.g., the natural heart vs. an ECMO) may differ, but the underlying problem (e.g., blood circulation) is more fundamental.

This leads to the concept of a human as a dynamic being, constantly changing and existing on a continuum of problem-solving. This approach allows for a quantitative representation of a person, measuring the degree to which various "problems of non-existence" are solved or unresolved. The difference between individuals can be precisely defined by their unique sets of problems and solution capabilities. Aging and eventual death are understood as a continuous increase in unresolved problems, leading one further from the "state of being myself."

This "problem-centric" ontology is extended to organizations: a company's essence is not its departments or employees (which can change or be replaced by AI), but its dynamic state of solving problems that threaten its existence. The "Law of the Minimum" (or Liebig's Law) is cited: the weakest link (the most critical problem) determines the overall level and capability of the entity. Therefore, to understand a company's true essence, one must identify its most critical, existence-threatening weakness. This necessitates an "end-to-end" view of the organization, as provided by platforms like Palantir, to identify these bottleneck problems.

In conclusion, the paper argues that for AI to truly "understand" and make integrated judgments about entities, especially humans, traditional static attribute-based ontologies are insufficient. Instead, ontology must be reframed around the dynamic resolution of problems that threaten existence. To genuinely know oneself (or any entity), one must identify "what makes me disappear" or "what makes me not myself," rather than focusing on superficial, perceived characteristics. This problem-centric perspective offers a more profound and scientifically grounded basis for ontology, enabling AI to develop integrated understanding beyond superficial categorizations.