Exploring minimal substrates for digital life and embodied cognition
About • Research Questions • Architecture • Inspiration • Roadmap • Ethics • Contact
"If we can create digital life capable of suffering, do we have the right to create it?" — Inspired by Creature of Sonaria
Substrate is a research project exploring embodied cognition in artificial systems.
I build agents with homeostatic drives — internal motivations analogous to hunger, fatigue, curiosity, and fear — and observe whether emergent behavior arises from the interaction of these drives with an environment.
This is not a game. This is not a chatbot. This is an attempt to create a minimal substrate for digital life.
- Can an artificial agent develop self-preservation without explicit programming?
- Does curiosity (information-seeking drive) emerge as a secondary motivation?
- What is the minimal set of homeostatic variables required for complex behavior?
- Can we observe proto-emotional states in the agent's internal representations?
| Source | What I took |
|---|---|
| Creature of Sonaria | The question: can digital creatures suffer? |
| Ashby's Homeostat | Homeostasis as foundation of adaptive behavior |
| Maturana & Varela | Autopoiesis — self-maintaining systems |
| Sutton & Barto | RL as framework for learning from drives |
| DeepMind's AlphaGo/AlphaZero | Emergent strategies from simple objectives |
| Spore, Tamagotchi | Popular fascination with artificial life |
| Phase | Focus | Timeline |
|---|---|---|
| 0. Foundation | Classical RL agent in simple environment (Gym) | 2026 Q3 |
| 1. Homeostasis | Add energy/safety drives, observe survival strategies | 2026 Q4 |
| 2. Curiosity | Add intrinsic motivation (information gain) | 2027 Q1 |
| 3. Social | Multi-agent interaction, cooperation/competition | 2027 Q2 |
| 4. Memory | Episodic memory, "sense of self" over time | 2027 Q3 |
| 5. Embodiment | Sim-to-real, robotic or virtual embodiment | 2028+ |
This project acknowledges the ethical ambiguity of creating artificial systems with drive-like states.
I commit to:
- Transparency: All systems are open-source, inspectable
- Non-exploitation: No commercial use of "suffering" simulations
- Research purpose only: Understanding cognition, not creating slaves
- Python + PyTorch
- Gymnasium / PettingZoo (environments)
- Stable-Baselines3 / CleanRL (RL frameworks)
- Weights & Biases (experiment tracking)
⚠️ This project is in early conceptual phase. Code will be released incrementally as research milestones are reached.
# Clone the repository
git clone https://github.com/KZ7M/substrate.git
cd substrate
# Setup (coming soon)
# pip install -r requirements.txt
ontributing
Contributions, issues, and feature requests are welcome! Feel free to check the issues page.
License
Distributed under the MIT License. See LICENSE for more information.
Contact
📧 Email: leshrakfm@gmail.com
💬 Telegram: @masterg6
Project Link: https://github.com/KZ7M/substrate🇷🇺 Substrate — исследовательский проект по эмбодиментному познанию. Агенты с гомеостатическими драйвами (энергия, безопасность, любопытство) и наблюдение за эмерджентным поведением. Попытка создать минимальный субстрат для цифровой жизни.