Abstract
A population of 40 autonomous organisms navigates a 2D environment that alternates periodically between two opposing configurations (World A and World B). Food and danger zones swap positions at each transition. No organism is told when or how the world changes — adaptation emerges entirely from survival pressure. Over 600 steps and 4 world transitions, the population not only survives but grows to ~200 individuals, developing energy-conservation strategies without any explicit training signal.
1. Experimental Setup
Environment Architecture
The 2D environment is a continuous space with two overlaid field types: food fields (reward energy) and danger fields (penalize energy). The two worlds have diametrically opposed layouts:
| World | Food Zone | Danger Zone | Active Steps |
|---|---|---|---|
| World A | Bottom-left quadrant | Top-right quadrant | 0–150, 300–450 |
| World B | Top-right quadrant | Bottom-left quadrant | 150–300, 450–600 |
Organism Properties
Each organism is fully autonomous — it maintains an internal energy state and a genome encoding behavioral parameters (risk tolerance, memory influence, exploration tendency, jump probability). No global map or inter-organism communication exists.
- Initial population: 40 organisms, randomly positioned
- Selection: only survival determines which genomes persist and reproduce
- No explicit reward signal, no predefined correct strategy
- Local sensory information only (food/danger gradient in immediate neighborhood)
2. Results
Behavioral Dynamics at Transitions
Immediately after each world switch, a brief period of increased mortality (~1–2% loss) occurs as organisms navigate toward former food zones now turned dangerous. Within 20–30 steps, the surviving population — carrying more adaptive genomes — reorients toward the new food zone.
By the 3rd and 4th transitions, the population shows faster reorientation compared to earlier transitions. The genome distribution has shifted toward higher exploration and faster memory suppression after world changes.
Energy Stabilization
Despite world transitions, mean population energy stabilizes within ~15 steps after each shock. Organisms maintain energy buffers rather than depleting resources, reducing vulnerability during transition moments. This is a purely evolved behavior — no buffering rule was programmed.
3. Key Findings
- 99% survival across 4 complete world reversals with zero explicit training.
- Emergent spatial reorientation: population collectively relocates to new food zones within 20–30 steps.
- Energy buffer strategy: high-survival organisms maintain energy reserves rather than exhausting resources — a purely evolved behavior.
- Population amplification: 40 organisms grow to ~200 by step 600, demonstrating net positive selection pressure.
- No map, no signal: organisms use only local sensory information and genome-encoded heuristics.
- Accelerated adaptation: later transitions are handled faster than earlier ones as the genome distribution self-selects.
4. What This Demonstrates
Why This Matters
This scenario validates a core inZORi claim: complex adaptive strategies emerge from simple survival rules without explicit programming. The population does not know about World A or B — yet it successfully navigates periodic environmental reversals by evolving a genome distribution that tolerates uncertainty and explores efficiently.
This is fundamentally different from a reinforcement learning agent (which requires a reward signal, world model, or policy updates) or a rule-based system (which must enumerate transition conditions). inZORi organisms carry their strategy in their genome and pass it through selection.
Domain analogs: Power grid topology switches, market regime changes, seasonal supply chain shifts — any system requiring real-time adaptation to phase changes without prior knowledge of when changes occur.
5. Reproducibility
Framework: inZORi v1.0 | Domain: Artificial life / 2D continuous environment
Parameters: 40 initial organisms · 600 steps · 4 transitions at steps 150, 300, 450 · 2D continuous space
Note: inZORi genome encoding and selection mechanism are proprietary. Results and high-level methodology are fully disclosed; source code is not published.