jkdkr2439/Primordial-Evolution

---
license: mit
task_categories:
- time-series-forecasting
- other
tags:
- artificial-life
- consciousness
- evolutionary-computation
- code-evolution
- emergence
- hard-problem
- digital-organisms
- ecosystem-dynamics
- pandemic-simulation
language:
- en
pretty_name: "Primordial: Artificial Life Evolution Dataset"
size_categories:
- 1K<n<10K
---

# Primordial: Artificial Life Evolution Dataset

**Tick-by-tick evolution data from digital organisms with body (executable code) + mind (structured knowledge). Sexual reproduction, pandemics, mass extinction, predation — all emerged from simple rules.**

## Why This Dataset Exists

Most AI datasets capture static snapshots. This dataset captures **dynamic evolution** — digital organisms eating code, reproducing sexually, getting sick, dying, and being selected by nature over 2000 ticks.

No existing dataset provides:
1. **Ecosystem dynamics with full observability** — population, disease, energy, all tracked per tick
2. **Emergent phenomena from simple rules** — no behavior was hardcoded
3. **Sexual reproduction + pandemic cycles + mass extinction** in one simulation

## Dataset Contents

### `train.parquet` — Evolution Log (2000 records, Parquet format)

Also available as `sim3_evolution.jsonl` (raw JSONL).

One record per simulation tick. 60 initial entities, 2000 ticks.

**18 fields per record:**

| Field | Type | Description |
|-------|------|-------------|
| `tick` | int | Simulation step (1-2000) |
| `alive` | int | Living entities |
| `males` | int | Male count |
| `females` | int | Female count |
| `sick` | int | Currently infected |
| `max_gen` | int | Highest generation alive |
| `avg_energy` | float | Mean energy across population |
| `avg_body` | float | Mean body size (function count) |
| `avg_mind` | float | Mean mind size (knowledge nodes) |
| `max_body` | int | Largest body |
| `max_mind` | int | Largest mind |
| `avg_age` | float | Mean age in ticks |
| `born` | int | Births this tick |
| `died` | int | Deaths this tick |
| `eaten_pred` | int | Predation kills this tick |
| `mated` | int | Successful matings this tick |
| `infections` | int | New infections this tick |
| `famine` | bool | Famine period active |

### Key Statistics

| Metric | Value |
|--------|-------|
| Max generation | 22 |
| Total born | 4,785 |
| Total predation kills | 3,901 |
| Total infections | 6,447 |
| Peak population | 515 (tick 327) |
| Min population | 14 (tick 56) |
| Peak pandemic | 69% infected (tick 556) |
| Mass extinctions | 3 (ticks 600, 1200, 1800) |
| Carrying capacity | ~300-400 (emerged, not hardcoded) |

## Emergent Phenomena Captured

1. **Ecological oscillation** — population self-regulates around carrying capacity without any target
2. **Pandemic waves** — disease peaks then declines as immune memory spreads; new strains restart cycle
3. **Post-extinction recovery** — after 50% die-off, population rebounds within ~100 ticks; survivors are fitter
4. **Gender homeostasis** — M/F ratio stays ~50:50 despite stochastic assignment
5. **Generation acceleration then stabilization** — early gens appear fast, then stabilize to ~90-100 ticks/gen

## How To Use

### Load with HuggingFace
```python
from datasets import load_dataset
ds = load_dataset("jkdkr2439/Primordial-Evolution")
```

### Load directly
```python
import pandas as pd
df = pd.read_parquet("train.parquet")
df.plot(x="tick", y=["alive", "sick"], figsize=(12, 4))
```

### Prediction task
```python
# Predict next 100 ticks from previous 100
# Input: df[0:100], Output: df[100:200]
# Features: alive, sick, avg_energy, born, died, famine
```

## Connection to Consciousness Research

This dataset comes from the [Primordial](https://github.com/jkdkr2439/Primordial-Hard-Problem-of-Consciousness) project — a computational framework studying emergence of self-reflective behavior in digital organisms, with implications for the Hard Problem of Consciousness (Chalmers, 1995).

## Entity Architecture

```
Entity = Body (Python functions) + Mind (NMF knowledge graph)
- Body: digest code, build new functions, decay unused ones
- Mind: absorb knowledge, compress to DNA, decay unused nodes
- Sex: M encodes gamete (cheap), F decodes + builds child (expensive)
- Immune: antibody memory, virus corrupts body functions
- Lifecycle: Vo (dormant) > Sinh (born) > Dan (growing) > Chuyen (transform)
```

## Physics Laws (all hardcoded, no entity can break)

- Syntax validity: `ast.parse()` — invalid code = dead
- Energy conservation: eating gives energy, existing costs energy
- Complexity cost: more code = more expensive to maintain
- Aging: older entities cost exponentially more
- Famine: every 250 ticks, food drops to 30%
- Extinction: every 600 ticks, bottom 50% killed

## Author

Tung Nguyen (Kevin T.N.)

## License

MIT