The Physics of Intelligence: Three Problems, Four Coordinates

Why civilizations, AI systems, and individual minds navigate the same computational geometry—and how to measure their position within it


Three Layers of Understanding

This essay presents the computational layer of the framework—the three problems any intelligent system faces and the coordinate system for measuring solutions.

These three problems emerge from four physical constraints that govern any goal-directed system maintaining order against entropy:

These physical constraints are derived from thermodynamics, information theory, and control systems theory in our companion essay The Four Axiomatic Dilemmas. That essay proves these constraints apply to any telic system (goal-directed agent maintaining order against entropy): viruses, civilizations, future AI.

This essay focuses on what those constraints feel like to an intelligent optimizer: the three computational problems they generate. If you want the deep physics, read the Dilemmas essay first. If you want the strategic diagnostic, start here.

I. The Opening Problem

Political thought has relied on historically arbitrary coordinate systems—Left/Right, Economic/Social—that cannot plot the most important questions of our time: AI alignment, civilizational growth strategy, existential risk navigation. We need coordinates derived from necessity, not convention. What problems must any intelligent system solve to persist?

II. The Trinity of Tensions: Three Universal Problems

Any intelligent system optimizing under physical constraints faces exactly three irreducible computational problems. These are mathematical necessities, not cultural conventions or human psychological quirks.

The Minimal Intelligent System

Define our subject precisely. An intelligent system is a physical system that:

  1. Models reality: Maintains internal state M (map) representing external state W (world)
  2. Chooses actions: Selects actions A based on model M to optimize for goals G
  3. Optimizes over time: Allocates finite resources E across temporal horizon t
  4. Exists with other agents: Operates in environment containing other systems with goals

A simple reinforcement learning agent satisfies this. An insect navigating its environment satisfies this. A human civilization satisfies this. AGI will satisfy this.

What problems must such a system solve?

Problem One: The World (Order vs. Chaos)

The natural decomposition separates Epistemic ("How to build accurate model M of world W?") from Praxis ("How to structure action A given model M?").

For an intelligent agent optimizing under physical constraints, these form one integrated domain—the Problem of the World.

The coupling is fundamental:

Information theory shows the coupling. Mutual information I(M;W) measures model accuracy. For an agent, I(M;W) only matters if actions A depend on M. A perfect map you never use has zero value. Model quality is defined by action utility.

Conversely, action architecture depends on epistemic strategy. If you gather high-fidelity real-time data (expensive), you can use reactive, decentralized control. If you rely on compressed historical data (cheap), you need more rigid, top-down plans. Your control architecture is constrained by your information strategy.

The agent's objective is joint optimization: maximize utility of actions given model quality, minus costs of sensing and control.

A bacterium solves this: chemotaxis couples simple sensing to simple action. AlphaGo solves this: neural net perception integrates with Monte Carlo Tree Search planning. You solve this: intuition fuses with analysis, vision with strategy.

The optimization problem: How to model and act upon chaotic, uncertain reality?

Two solution dimensions:

Epistemic and praxis are deeply coupled in practice, yet represent orthogonal solution dimensions. The World Problem is a plane with two coordinates that vary independently:

R-Axis (Information Strategy): Where on the spectrum from cheap historical data (Mythos, R-) to expensive real-time data (Gnosis, R+)? A termite following pheromones (R-) versus a scientist running experiments (R+).

O-Axis (Control Architecture): Where on the spectrum from decentralized emergent coordination (O-) to centralized designed command (O+)? A flock of birds (O-) versus a military command hierarchy (O+).

These coordinates vary independently. High R+ with low O+ yields a scientist with no execution capacity—brilliant analysis, no implementation. High O+ with low R- yields rigid bureaucracy following outdated models.

Physical Grounding: The World Problem emerges from two of the Four Axiomatic Dilemmas:

These physical constraints manifest as the epistemic and praxis dimensions of the World Tension.

Problem Two: Time (Future vs. Present)

Given some epistemic-praxis solution, a second orthogonal problem emerges: How to allocate finite resources across time?

You have finite energy Etotal. Allocation decision: Epresent + Efuture = Etotal.

In reinforcement learning, this is explicit. The discount factor γ in the value function:

Vπ(s) = 𝔼[∑t=0→∞ γt · rt]

Where γ = 0 yields pure present focus (T-), γ = 1 yields infinite future focus (T+), and optimal γ ≈ 0.95-0.99 balances present and future.

Orthogonality to World:

You can have perfect world model (high R+, optimized O) and still choose the wrong time horizon. A chess engine can model the board perfectly but have flawed evaluation overweighting immediate material gain or searching too deeply into unlikely branches.

Conversely, you can have optimal temporal balance but catastrophic world modeling. A civilization can perfectly balance preservation and growth but base both on false cosmology.

The tension is irreducible. Pure present-focus yields exploitation, stagnation, death. Pure future-focus yields exploration, instability, failure to consolidate gains.

Physical Grounding: The Time Problem is the direct computational manifestation of the Thermodynamic Dilemma. Any telic system with finite free energy must allocate between maintenance (Homeostasis, T-) and growth (Metamorphosis, T+). This is thermodynamic necessity.

Problem Three: Self (Agency vs. Communion)

Given epistemic-praxis solution and temporal allocation solution, a third problem emerges for any system composed of multiple agents (cells, organisms, humans, AIs, or sub-modules within a single mind): Where is the boundary of "self" for optimization?

In game theory, this is multi-level selection. For a system with n agents, each agent i can optimize:

  1. Individual fitness fi (S- strategy: Agency)
  2. Group fitness Fgroup = f(f1, f2, ..., fn) (S+ strategy: Communion)

These are often in conflict. Tragedy of the commons: individual optimization destroys group optimum. Pure group optimization creates free-rider problem: what prevents defection?

Orthogonality to World and Time:

Perfect world model and optimal time horizon still leave the Self problem unresolved.

Meiji Japan: High R+ (adopted Western science), high T+ (rapid industrialization), high S+ (intense collectivism). Solved World and Time but chose strong Communion solution.

Modern Singapore: High R+ (technocratic governance), moderate T+ (long-term planning), moderate S- (meritocratic individualism). Same World/Time solutions, different Self solution.

These civilizations have different coordinates despite similar success on other axes. The Self dimension varies independently.

In AI alignment, this is explicit. The inner/outer alignment problem: Should the AI optimize for its learned objective (inner alignment, S- for the AI as agent) or human values (outer alignment, S+ including humans in boundary)?

This is orthogonal to the AI's world-modeling capability and time horizon. An AI can have perfect world model and balanced time preference but still face the "whose goals?" question.

Physical Grounding: The Self Problem emerges from the Boundary Dilemma. Multi-level selection theory (the Price equation) formalizes this: evolutionary change decomposes into within-group selection (Agency) and between-group selection (Communion). These often point in opposite directions.

Proof of Sufficiency: Why No Fourth?

Any proposed fourth problem must satisfy our criteria: Necessary, Irreducible, Universal, Orthogonal. Test candidates:

"Stability vs. Change": This IS the Time tension. Eliminated.

"Centralization vs. Decentralization": This IS the O-Axis. Eliminated.

"Competition vs. Cooperation": This IS the Self tension. Eliminated.

"Truth vs. Meaning": This IS the R-Axis. Eliminated.

For an intelligent system optimizing under physical constraints:

  1. Must solve: How to model and act on world → World
  2. Must solve: How to allocate resources across time → Time
  3. Must solve: How to coordinate with other agents → Self

Any proposed addition is either a sub-problem of these three, a combination of two or more, or a derived consequence rather than fundamental tension.

The problem space for intelligence is three-dimensional.

The Four-to-Three Compression

There are four physical dilemmas (Thermodynamic, Boundary, Information, Control) but only three computational problems (Time, Self, World).

The Information Dilemma (how to acquire accurate data) and the Control Dilemma (how to structure coordination) are distinct physical constraints. A bacterium faces the Information Dilemma through chemotaxis. A virus faces the Control Dilemma through its capsid structure. These are separable at the physics layer.

For an intelligent system capable of modeling and planning, these two physical constraints fuse into a single computational problem: how to simultaneously model reality AND act upon it effectively. Epistemic strategy (R-Axis: Mythos ↔ Gnosis) and praxis architecture (O-Axis: Emergence ↔ Design) become two dimensions of the same optimization challenge: the Problem of the World.

This is why the World generates two measurement axes (R and O) while Time and Self each generate one (T and S).

For the full physics derivation showing why these dilemmas are necessary and universal, see The Four Axiomatic Dilemmas.


III. Machines Already Navigate This Geometry

The Trinity is observable in existing computational systems. These tensions emerge from optimization physics, not human psychology. Any goal-directed system navigating physical reality under constraints faces World, Time, and Self problems.

Artificial systems already navigate structurally analogous tensions. The computational geometry is identical. The core optimization structure remains the same.

AlphaGo: Navigating the World Tension

AlphaGo combines Policy Network (O+/Design: precision, brittle) with Monte Carlo Tree Search (O-/Emergence: robust, expensive). Pure strategies fail; integration succeeds.

On the R-Axis: It trains on human games (R-/Mythos: compressed historical patterns) then surpasses via self-play (R+/Gnosis: costly novel exploration). Pure R- plateaus at human level. Pure R+ is computationally intractable. Synthesis achieves superhuman performance.

This artificial system navigates identical Order/Chaos geometry as civilizations.

Reinforcement Learning: Navigating the Time Tension

The discount factor γ in Vπ(s) = 𝔼[∑t γt · rt] directly encodes Time Tension.

γ = 0 (T-): pure exploitation, myopic optimization. Agent ignores future consequences entirely.

γ = 1 (T+): pure exploration, infinite time horizon. Agent weights distant future equally with immediate present, producing unstable learning.

Optimal γ ≈ 0.95-0.99 balances present and future. This is empirically discovered, not theoretically derived. Extreme γ values produce catastrophic failure.

Your civilization faces the same equation. The Democratic Ratchet is γ → 0 in political form: myopic optimization for present consumption at expense of future possibility.

Multi-Agent RL: Navigating the Self Tension

Independent learners (S-) face tragedy of commons. Centralized controllers (S+) fail to scale. Dec-POMDPs retain local agency while enabling coordination—empirically superior to pure extremes.

The inner/outer alignment problem in AI safety is Self Tension in technical form: Where does the AI draw its optimization boundary? Mesa-optimization manifests Self Tension. Reward hacking manifests Time Tension pathology. Corrigibility reflects World Tension.

Every AI safety problem is Trinity navigation in different substrate.


IV. SORT: The Four Coordinates of Intelligence

These three universal problems create a four-dimensional solution space. The coordinates of this space are the SORT axes: Sovereignty, Organization, Reality, Telos.

Let's explore each axis by showing which Trinity problem it solves.

The T-Axis: Solving the Problem of Time

Range: -1 (Homeostasis) to +1 (Metamorphosis)
Question: Do we preserve what we have, or risk it to become greater?

The Time Tension manifests in civilizations as the strategic choice between conservation and transformation.

The -1 Pole (Homeostasis):

The axiology of the Hospice. The goal is stability, comfort, risk-aversion, and the preservation of past successes. A maintenance society. The highest value is sustaining the present equilibrium. Growth that threatens stability is rejected.

The +1 Pole (Metamorphosis):

The axiology of the Foundry. The goal is growth, transcendence, and the willingness to risk the comfortable present for a more transcendent future. A striving society. The highest value is transformation toward higher complexity, capability, and purpose.

The Trade-off: Pure Homeostasis is slow death. Pure Metamorphosis is self-consuming fire. Healthy civilizations navigate between securing foundations and reaching higher.

Physical Grounding: The T-Axis maps directly to the Thermodynamic Dilemma. Energy allocated to maintenance (Homeostasis) cannot simultaneously fund growth (Metamorphosis). The discount factor γ in reinforcement learning is this dilemma made explicit.

The S-Axis: Solving the Problem of Self

Range: -1 (Agency/Individual) to +1 (Communion/Collective)
Question: Who matters most—the individual or the group?

The Self Tension manifests as the boundary-definition problem: where does ultimate value reside?

The -1 Pole (Agency/Individual):

Sovereignty resides in the individual. The purpose of society is to maximize personal liberty, protect natural rights, and enable self-actualization. The individual is the irreducible unit of moral value. Social arrangements are legitimate only to the extent they serve individual flourishing.

The +1 Pole (Communion/Collective):

Sovereignty resides in the group—the tribe, the nation, the civilization. The long-term survival, cohesion, and glory of the group is the highest good, to which individual desires must be subordinated. The collective has moral reality beyond the sum of its members.

The Trade-off: Individual maximizes agency and innovation, risks atomization. Collective maximizes unity and focus, risks stagnation and suppressing genius. The tension is inescapable.

Physical Grounding: The S-Axis maps directly to the Boundary Dilemma. The Price equation in evolutionary biology formalizes this: change decomposes into within-group selection (Agency) and between-group selection (Communion). Multi-agent RL systems navigate this dimension empirically.

The R & O Axes: Solving the Problem of the World

The World problem requires solving two sub-problems simultaneously: how to model reality (epistemic challenge) and how to coordinate action (praxis challenge).

While these are deeply coupled in practice (your control architecture depends on your information strategy), they represent orthogonal solution dimensions. This is why the World Tension generates two measurement axes.

The R-Axis (Information Strategy)

Range: -1 (Mythos/Stories) to +1 (Gnosis/Data)
Question: Do we trust sacred narratives or empirical experiments?

The -1 Pole (Mythos):

Truth is found in our stories. It is revealed through narrative, tradition, religion, archetype, and the shared, intuitive wisdom of the tribe. Mythos provides meaning, cohesion, and a moral compass. Reality is understood through symbolic interpretation and sacred texts.

The +1 Pole (Gnosis):

Truth is found in data. It is discovered through empirical observation, logical deduction, and ruthless, falsifiable experimentation. Gnosis provides accuracy, competence, and a brutal, unflinching map of reality. Knowledge is validated by prediction and control.

The Trade-off: Mythos provides meaning and cohesion. Gnosis provides accuracy and power. Integration required: Gnosis refines Mythos, Mythos gives meaning to Gnosis. Failure yields brittle theocracy (R- pathology) or sterile technocracy (R+ pathology).

Physical Grounding: The R-Axis maps to the Information Dilemma. Real-time sensing (Gnosis) has high metabolic cost but high accuracy. Compressed historical data (Mythos) is cheap but potentially obsolete. Environmental volatility determines optimal strategy.

The O-Axis (Control Architecture)

Range: -1 (Emergence/Bottom-up) to +1 (Design/Top-down)
Question: Should order emerge spontaneously or be centrally planned?

The -1 Pole (Emergence):

Order is not created; it is discovered. A resilient and prosperous society arises organically from bottom-up processes. This includes free markets (price signals coordinating production), common law (evolved precedent adapting to cases), and tradition (multi-generational filtering of practices).

The +1 Pole (Design):

Order is not discovered; it is architected. A complex, dangerous world requires conscious, rational, and far-sighted authority to design systems, manage complexity, and steer toward desirable futures. This is the logic of the engineer, the central planner, and the lawgiver.

The Trade-off: Total Design yields brittle sclerosis. Total Emergence yields chaotic impotence. Optimal: minimum elegant Design unleashing maximum creative Emergence.

Physical Grounding: The O-Axis maps to the Control Dilemma. Control systems theory proves the trade-off: centralized control is optimal under perfect information but brittle under uncertainty. Distributed control is suboptimal but resilient under partial information.

V. Reading SORT Coordinates

SORT coordinates are written as [S, O, R, T] with values ranging from -1 to +1 on each axis.

Example notation: [S-0.6, O-0.4, R+0.7, T+0.8]

This reads as: "Strong individual sovereignty, moderately emergent organization, strong Gnostic epistemology, strong Metamorphic telos."

Historical Examples

Civilization S O R T Description
Classical Athens
(5th century BCE)
-0.6 -0.4 +0.7 +0.8 High-agency, emergent, gnostic, metamorphic. The archetypal Foundry—individual genius, philosophical inquiry, imperial expansion.
Sparta
(5th century BCE)
+0.8 +0.6 -0.5 -0.4 High-communion, designed, mythopoetic, homeostatic. The archetypal militaristic collective—unity through discipline, preservation over growth.
Victorian Britain
(1850-1900)
-0.3 +0.2 +0.8 +0.9 Moderate individualism, slight design bias, strong empiricism, intense growth. High-coherence Foundry at civilizational peak—industrial revolution, imperial expansion, scientific dominance.
Tokugawa Japan
(1603-1868)
+0.6 +0.7 -0.6 -0.8 Collective identity, designed hierarchy, traditional wisdom, extreme homeostasis. Deliberately chose isolation and stasis—stable for 250 years but vulnerable to external shock.
USA
(1960)
-0.4 -0.2 +0.7 +0.8 Individual liberty, emergent markets, scientific dominance, Apollo-era ambition. High-coherence Foundry at civilizational peak.
Modern West
(2020s)
±0.0 +0.6 ±0.1 -0.5 Fragmented sovereignty (low coherence), bureaucratic design, mixed/incoherent epistemology, homeostatic drift. Low-coherence state with internal conflict and declining vitality.

Key Patterns to Notice

The Complete Derivation Chain

Physical Law Computational Problem Measurement Axis
Information Dilemma
(metabolic cost of sensing)
World
(Order vs. Chaos)
R-Axis (Reality)
Mythos ↔ Gnosis
Control Dilemma
(centralized vs. distributed)
O-Axis (Organization)
Emergence ↔ Design
Thermodynamic Dilemma
(energy allocation)
Time
(Future vs. Present)
T-Axis (Telos)
Homeostasis ↔ Metamorphosis
Boundary Dilemma
(self-definition)
Self
(Agency vs. Communion)
S-Axis (Sovereignty)
Individual ↔ Collective

Four physical dilemmas → Three computational problems → Four measurement coordinates.


VI. Why This Matters

The Trinity of Tensions and the SORT framework describe the computational geometry of intelligence itself.

Same Geometry, Different Substrates

Civilizations navigate this geometry through culture, institutions, and collective choice. A civilization's SORT coordinates describe its strategic solutions to the Trinity.

AI systems navigate this geometry through training dynamics, reward functions, and architectural choices. The inner/outer alignment problem is Self Tension. Reward hacking is Time Tension pathology. Corrigibility is World Tension.

Individual consciousness navigates this geometry through psychological integration, value hierarchies, and life strategies. Your personal SORT coordinates describe your native solutions to these tensions. Internal conflict reflects misalignment between different parts of your psyche navigating these problems differently.

Substrate-independent computational necessity.

Predictive Power

Understanding the Trinity and SORT enables:

Where to Go from Here

This essay sits between physics and applications. For the deep foundation, see The Four Axiomatic Dilemmas. For diagnostic applications, see The Tyranny of the Present or The Axiological Malthusian Trap. For AI alignment applications, see AI Alignment via Physics.