Why AI Companies Are Turning to Philosophy — And What This Means for the Future of AI
PSI Think Tank | Strategic Analysis
Artificial intelligence is entering a phase in which increasing computational power and larger models are no longer the only questions that matter. As AI systems become more capable of reasoning, interacting with the real world, and supporting increasingly consequential decisions, a different set of problems is becoming visible: What does a concept actually mean? How should context be represented? What constitutes a valid decision? How can a system distinguish between different layers of meaning? And how should reasoning be structured before an action is taken?
These questions help explain why philosophy is becoming increasingly relevant to advanced artificial intelligence. Philosophy is not entering AI because machines suddenly need philosophical opinions. Its relevance lies elsewhere. For thousands of years, philosophy has developed methods for examining concepts, categories, assumptions, knowledge, logic, meaning, causality, reality, and decision-making. Many of the problems now appearing in advanced AI systems are, at their core, problems of this kind.
The issue becomes particularly clear when examining ambiguous concepts. Consider the word “order.” Depending on context, it can refer to physical order, logical order, social order, political order, organizational order, legal order, or normative order. An AI system may possess information about all of these meanings. But knowing the meanings is not the same as determining which meaning is structurally relevant in a particular situation.
This distinction is fundamental.
The challenge is no longer simply to retrieve the correct definition. The deeper challenge is to determine which semantic dimension belongs to the decision space being analyzed.
This leads to a more general principle:
Structure → semantic space → definition.
Before a concept can always be defined adequately, the relevant structure in which that concept operates may need to be identified. This is one reason why ontology has become an increasingly important concept in AI architecture.
A particularly significant example is Palantir’s Ontology. Palantir describes its Ontology as a central system for human-AI decision-making and distinguishes it from a conventional data architecture. Rather than representing only data, the system is designed to represent decisions and their surrounding context, including data, logic, action, and security.
This development is important because it represents a movement away from a purely data-centric understanding of AI toward a decision-centric one. The question becomes not merely:
What data does the system have?
but:
What decision is being made, on the basis of what information, according to what logic, and with what consequences?
That represents a significant architectural development.
Another, philosophically more explicit example is Jachin. Jachin describes its research program as combining formal ontology, category theory, and symbolic reasoning, with the stated aim of using structures developed through centuries of philosophical inquiry as part of a reasoning infrastructure for AI. The importance of such approaches is not necessarily that they provide a finished solution to artificial general intelligence. Their significance lies in the direction they indicate.
The development of AI may increasingly move from:
Model → output
toward:
Model → structured context → reasoning → decision → action.
This changes the role of architecture.
A large language model can contain enormous amounts of information. But information alone does not automatically produce an adequate representation of a complex decision space. The model still needs to determine what is relevant, what belongs together, what distinctions matter, which assumptions are active, and how different elements influence one another.
This is where philosophy, ontology, semantics, logic, and systems architecture begin to converge.
The distinction between knowledge and structure becomes particularly important. A system may know thousands of facts about migration, economics, military strategy, climate, demographics, or technological development. Yet the existence of those facts does not by itself create a coherent strategic model.
A strategic decision requires a structured relationship between facts, actors, conditions, dependencies, objectives, constraints, risks, alternatives, and consequences.
Knowledge describes what is known. Structure determines how what is known becomes meaningful within a decision space.
This distinction may become increasingly important as AI systems move toward more autonomous forms of reasoning.
It also provides an interesting point of comparison with the 10DO Framework.
10DO does not attempt to define itself as a “philosophical AI.” Its purpose is different. The philosophical dimension forms part of its theoretical foundation, but the intended result is a structural architecture for the analysis and condensation of complex decision spaces.
The central question is therefore not:
What does philosophy say about AI?
but:
How can structural knowledge about meaning, order, context, and decision-making contribute to an architecture in which AI can reason more effectively within complex decision spaces?
This leads to a possible formulation of the 10DO AI approach:
10DO is a structural architecture for AI reasoning in complex decision spaces.
The distinction from a conventional ontology is important. An ontology primarily describes entities, relationships, categories, and structures within a particular domain. Palantir, for example, uses an ontology to create a semantic representation of an enterprise and connect data, logic, actions, and governance to operational decisions.
10DO approaches the problem from another direction. The starting point is not simply the question of what exists within a domain. The starting point is the structure of the decision space itself.
What structural fields are present? How do they interact? Which dimensions determine the strategic situation? Which meanings emerge within those relationships? Which distinctions are essential? And how can the resulting complexity be condensed without destroying the structure necessary for a meaningful decision?
This is where the concept of structural condensation becomes important.
Complexity cannot always be solved by adding more information. In many cases, additional information increases the complexity of the decision space. What is required is a method for preserving the relevant structural relationships while reducing unnecessary complexity.
The objective is therefore not simply to produce more answers. It is to create the conditions under which an answer can be structurally meaningful.
This may become one of the central challenges for advanced AI.
The future development of AI will probably not be determined exclusively by larger models, faster processors, or greater quantities of training data. These factors remain important, but another architectural layer is emerging around them: the organization of meaning, context, reasoning, decision-making, and action.
The transition can therefore be described in three broad stages:
The first stage was primarily concerned with computation.
The second stage increasingly focused on language, information, and knowledge.
The emerging third stage concerns structured reasoning and decision-making in complex environments.
At this level, philosophy becomes relevant again—not because AI needs to become philosophical, but because some of the deepest problems of intelligent systems are problems of structure, meaning, knowledge, logic, context, and decision.
The important question for the coming years may no longer be simply:
How intelligent is the model?
It may increasingly become:
Within what structure does the model reason?
And ultimately:
Who or what determines the structure of the decision space in which an AI system operates?
That question reaches beyond conventional AI engineering. It concerns the architecture of intelligence itself.
For this reason, the growing intersection between philosophy, ontology, semantics, reasoning, and artificial intelligence deserves close strategic attention.
The future of AI may depend not only on creating systems that can generate increasingly sophisticated answers, but on creating architectures that can determine which questions matter, which meanings are relevant, which structures govern a situation, and how complex information can be transformed into responsible decisions.
This is precisely the area in which structural approaches to AI reasoning may become increasingly important. And it is also the area in which the further development of 10DO AI should be examined.
PSI Think Tank — Strategic Perspective
The purpose of this analysis is not to present 10DO as the only possible solution, but to identify an emerging architectural problem in artificial intelligence and to examine where structural decision architectures may contribute to its development.