Multi-Scale Evolutionary Development and Cognition

Overview

A python implementation for cognitive evolution related to the Technological Approach to Mind Everywhere (TAME) ¹ and Darwin's Agential Material frameworks ², covering the following references:

Project Structure

• Experiments and implementations to publications are located in the ./examples directory
• Code for the Hybrid NCA-Framework is located in the ./musepy directory
• The evolutionary training stack depends on the mindcraft framework
• See also the LICENSE file for licensing information

System information

• Python 3.8
• Ubuntu 22.04 LTS

Notes

The mindcraft framework is under active development and the API might change in the future. The framework is not fully documented, but we are working on it.

WARNING: The mindcraft.io.Repr class is used to represent native classes as yml files, which might also use eval to evaluate the class.

Experiments

Evolutionary Implications of Multi-Scale Intelligence and NCAs ³

Implementations and examples to ³ are located in the examples/evolve directory.

Aging as a loss of Goal-Directedness ⁴

Implementations to ⁴ are located in the examples/aging-as-loss-of-goal-directedness directory.

Hybrid NCA-Framework Overview

A Hybrid Neural Cellular Automaton (NCA) introduced in³ and used to evolve collective, multi-cellular agents with multi-scale developmental competency from initial (direct) encodings is implemented in the musepy package (see ./musepy). The evolutionary training stack of musepy depends on the mindcraft framework (https://github.com/bhartl/NeurEvo). See also the ./examples/evolve/README.md for further details on how to use the evolutionary training stack.

Install

Virtual Environment

The musepy framework should be installed in dedicated virtual environments (venv) to not compromise the system's Python installation. With Anaconda, a corresponding muse venv can be generated as follows:

conda create -n muse python=3.8

which can be activated via

conda activate muse

Install as Editable Package

To install the musepy package, go to the root directory of the multiscale-evolution framework and run

pip install -e .

or for the cpu-only version run

pip install -e . --extra-index-url https://download.pytorch.org/whl/cpu

(recommended - the code is written for mpi4py parallelization -- see install detailed instruction for mpi4py or try running conda install mpi4py)

Dependencies

The repository depends on the mindcraft framework, see https://github.com/bhartl/NeurEvo for installation instructions (especially for mpi4py parallelization and acceleration, and licensing).

Examples and Jupyter Notebooks

To run the examples, analysis tools and Jupyter notebooks, the examples packages listed in the project.optional-dependencies of the pyproject.toml file are required. These can be installed via

pip install -e .[examples]

To use the Jupyter notebooks, the muse kernel needs to be installed via

ipython kernel install --name "muse" --user

Citation

If you use this code in your research, please cite the following publication:

@article{hartl2024evolutionary,
    author = {Hartl, Benedikt and Risi, Sebastian and Levin, Michael},
    title = {Evolutionary Implications of Self-Assembling Cybernetic Materials with Collective Problem-Solving Intelligence at Multiple Scales},
    journal = {Entropy},
    volume = {26},
    year = {2024},
    number = {7},
    article-number = {532},
    doi = {10.3390/e26070532}
}

References

Footnotes

1. "Technological Approach to Mind Everywhere: An Experimentally-Grounded Framework for Understanding Diverse Bodies and Minds",
   by M. Levin, Frontiers in Systems Neuroscience 2022, 16; https://doi.org/10.3389/fnsys.2022.768201 ↩

2. "Darwin’s agential materials: evolutionary implications of multiscale competency in developmental biology",
   by M. Levin, Cellular and Molecular Life Sciences 2023, 80, 142; https://doi.org/10.1007/s00018-023-04790-z ↩

3. "Evolutionary Implications of Self-Assembling Cybernetic Materials with Collective Problem-Solving Intelligence at Multiple Scales",
   by B. Hartl, S. Risi, M. Levin, Entropy 2024, 26(7), 532; https://doi.org/10.3390/e26070532 ↩ ↩² ↩³

4. "Aging as a Loss of Goal-Directedness: An Evolutionary Simulation and Analysis Unifying Regeneration with Anatomical Rejuvenation",
   by L. Pio-Lopez*, B. Hartl* M. Levin, Advanced Science 2025, https://doi.org/10.1002/advs.202509872 (* equal contribution) ↩ ↩²