A Promise Connecting Different Capabilities: Model Context Protocol
How AI Understands the World: Hierarchical Ontology
We once thought of AI simply as machines becoming smarter. But AI development has gone far beyond making fast accurate machines — it has begun to fundamentally change how we see, understand, and live in the world. From simple classification tasks, AI now writes, draws, converses, and performs complex tasks by connecting with various tools and programs.
AI is no longer a simple model doing one thing well but is becoming a complex ecosystem where components with various capabilities cooperate. This requires understanding AI as an interconnected system — a concept called Object-Oriented Thinking.
From One Model to Cooperating Multiple Models
Early AI had one large model processing everything. Recent AI has evolved to multiple small specialized models cooperating. Modern chatbots illustrate this: asking "what is today''s exchange rate?" triggers an internet search tool; asking "what flower is in this photo?" connects to an image analysis model; asking to "summarize this report" calls a summarization model. Each capability is specialized and connected — like software evolving from monolithic programs to modular microservices.
Object-Oriented Thinking: Reconstituting AI
Object-oriented thinking divides large programs into small parts each with clear roles and responsibilities — like building a house with specialized workers for bricks, electrical, windows, and roofing. In AI, small specialized models each handling one task and communicating with others is more efficient than one model doing everything. New functions can be added without retraining existing models — simply adding a new specialist module enables it to cooperate with the existing system immediately.
Model Context Protocol: Connecting Different Capabilities
When multiple models with different capabilities cooperate, clear rules are needed for how they communicate. Model Context Protocol (MCP) pre-defines what input different models receive and what output they produce — like meeting rules where everyone knows their role and speaks at appropriate times. For example, when asked "what is 25 × 134?", the language model sends a precisely formatted request to the calculation tool, which returns the result, which the language model then presents naturally to the user. MCP enables different models to have meaningful conversations without misunderstanding — making the whole system''s natural organic movement more important than any individual model''s speed.
Hierarchical Ontology: How AI Understands the World
Humans naturally understand objects in hierarchical systems — a cat is an animal, a mammal, of the cat family. Hierarchical Ontology provides AI with a similar "concept map": when encountering "dog," AI knows it belongs to "animal" → "mammal" → "pet." This enables: deeper and more accurate word understanding; handling unfamiliar concepts by inferring from known hierarchical context; AI defining its own role and selecting appropriate tools for each situation. Hierarchical ontology acts like a worldview guidebook for AI, enabling flexible and deep understanding across diverse situations.
Language Is a New Interface
Language models have evolved from sentence generators to central engines connecting and orchestrating diverse AI tools. Language is the most universal and intuitive interface between humans and machines. Rather than pressing complex buttons or coding, simply saying "tell me today''s weather" or "is there a cat in this photo?" triggers the appropriate tool chain. Language models grasp user intent, call external APIs or tools, and present results naturally. New tools can be added simply by teaching the language model their usage through examples — no complex interface redesign needed. This flexibility makes AI increasingly user-centered and accessible.
From Technology to Structure, From Structure to Ecosystem
AI development started from making better-performing models. Now the more important question is how well-made models connect and cooperate. Object-Oriented Thinking, Interaction Protocols, Hierarchical Ontology, and Large Language Models as central engines — together these give AI a flexible, human-centered cooperative structure. AI has become more than a tool solving difficult problems. It is making people think about what kind of society and world we want to create together. Designing structure is designing the possibilities we will unfold going forward. The next chapter examines the map of possibilities — where AI can go from here.
