Meaningful AI Mortality

Minimal public repository for the intrinsic mortality proof of concept and the associated conceptual framework.

Scope

This repository contains:

• PoC v1: intrinsic mortality via effective parameter corruption and progressive loss of hidden neurons.
• PoC v2: intrinsic mortality via cumulative neuronal senescence with persistent per-neuron vitality.
• Attack simulations showing that intrinsic mortality is bypassable at system level via restore, cloning, and code bypass.
• Core theory notes on model degradation, identity, and meaningful mortality.
• LaTeX source and PDF for the paper.

Repository structure

• simulations/code: executable PoC scripts and attack experiments.
• simulations/results: figures and summary outputs.
• theory/core: formal model definition and mortality framework.
• theory/identity: identity-focused note.
• theory/security: attack-surface note.
• paper/tex: LaTeX source and compiled PDF.

Technical setup

• Dataset: Iris (scikit-learn)
• Architecture: MLP 4 -> 16 -> 3
• Implementation: NumPy + matplotlib + scikit-learn

Install dependencies:

python -m pip install -r requirements.txt

Run the main scripts:

python simulations/code/intrinsic_mortality_poc.py
python simulations/code/intrinsic_mortality_poc_v2.py
python simulations/code/intrinsic_mortality_attack_experiments.py

Main empirical result

The project shows that intrinsic mortality can be made endogenous at the model-instance level, but remains local. Restore, cloning, and implementation-level bypass preserve continuity beyond the death of any single execution trajectory.

Citation

Paper archived on Zenodo:

Jusef Khamlichi (2026). Toward Meaningful Mortality in AI Systems: Intrinsic Degradation, Identity, and Succession. Zenodo. https://doi.org/10.5281/zenodo.19556905

Related paper files

• paper/tex/meaningful_ai_mortality_v4.tex
• paper/tex/meaningful_ai_mortality_v4.pdf