About the Model

Origins & Development

From Simple Questions to Profound Insights

The WHAT vs HOW model emerged from a simple question about data compression that led to a profound realization: our understanding of artificial intelligence was trapped in inadequate metaphors. We called it a brain, a library, a stochastic parrot—each falling short.

Through intense inquiry and analysis, we discovered that the key to understanding AI lies in a fundamental distinction: the difference between what it knows and how it thinks. This isn't just semantics—it's the foundation of a new framework that actually helps us comprehend and direct AI systems effectively.

Began with data compression research questions
Recognized limitations of existing AI metaphors
Developed the fundamental WHAT vs HOW distinction
Created the naval command structure analogy
Validated through practical application and testing

The scaffolding concept: building intelligence through structured protocols

The question is not whether machines think, but whether we can think clearly enough about how they think to guide them toward wisdom rather than mere knowledge.

— Core Philosophy of the WHAT vs HOW Model

Technical Foundation

The collaboration between innate policy and designed protocols

Grounded in Real AI Architecture

The WHAT vs HOW model isn't just theoretical—it's grounded in the actual architecture of modern AI systems. Large Language Models really do contain vast compressed world models (the WHAT) and sophisticated policy networks (the Captain) shaped by training data frequency and fitness functions.

The breakthrough insight is recognizing how these components interact and how we can leverage massive context windows to provide detailed operational protocols (the Admiral's orders) that guide the system's inherent capabilities toward specific, reliable outcomes.

Based on actual LLM architecture analysis
Incorporates training dynamics (frequency + fitness)
Leverages massive context window capabilities
Provides actionable guidance for system design
Enables predictable, reliable AI behavior

🧠

World Model Compression

The WHAT represents knowledge as distributed, holographic potential rather than discrete facts, enabling rich semantic associations and contextual understanding.

⚓

Policy Network Dynamics

The Captain's decision-making is shaped by dual forces: frequency of patterns in training data and fitness optimization for correct, coherent responses.

🎯

Protocol Architecture

The Admiral's protocols can scale to gigabytes of navigational data, providing unprecedented precision in guiding AI behavior toward specific goals.

🔄

Learning Mechanism

The self-correcting loop enables true learning through protocol rewriting, transforming knowledge into wisdom via reflection and adaptation.

Development Journey

Initial Inquiry

Research into data compression and knowledge representation revealed fundamental gaps in how we conceptualize AI intelligence. The search for better metaphors began.

Phase 1

Metaphor Analysis

Systematic analysis of existing AI metaphors (brain, library, parrot) revealed their limitations and the need for a more precise framework.

Phase 2

Core Insight

The breakthrough distinction between WHAT (knowledge) and HOW (process) emerged, along with the naval command structure analogy.

Phase 3

Framework Development

Detailed modeling of each component: the compressed world model, policy networks, operational protocols, and the self-correcting learning loop.

Phase 4

Validation & Testing

Practical application and testing confirmed the model's predictive power and its utility in guiding AI system design and interaction.

Phase 5

Future Implications

AI System Design

The WHAT vs HOW model provides a blueprint for building more reliable and predictable AI systems by focusing on protocol design rather than just scaling knowledge bases.

Human-AI Collaboration

Understanding the distinction between innate capability and designed guidance enables more effective human-AI partnerships, where humans focus on the "how" while AI provides the "what."

AI Safety & Alignment

The model reveals how to align AI systems by carefully designing operational protocols that encode human values and constraints, making AI behavior more predictable and controllable.

Educational Applications

The framework suggests new approaches to AI-assisted learning, where educational protocols can guide AI tutors to adapt their vast knowledge to individual student needs and learning styles.

Research Directions

The model opens new research avenues in understanding how AI systems learn, generalize, and can be guided toward increasingly sophisticated forms of intelligence and wisdom.

Ethical AI Development

By making the distinction between knowledge and process explicit, the model helps us think more clearly about responsibility, agency, and the ethical implications of AI systems.