Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Subjects: | |
| Online Access: | https://doi.org/10.5281/zenodo.18489477 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>This paper introduces a computationally grounded formulation of curiosity as a bounded, self-reinforcing operator acting on a fractal–holographic symbolic memory system. Unlike traditional curiosity models based on exploration incentives, novelty rewards, or policy optimization, the proposed mechanism operates entirely within memory dynamics and is triggered solely through engagement.</p> <p>We present a formal reinforcement operator that improves recall fidelity, compression efficiency, basin stability, and interference resistance through repeated use—without retraining, gradient descent, or loss minimization. Empirical evaluation demonstrates that curiosity-driven reinforcement produces monotonic improvements during early engagement, followed by natural saturation that prevents runaway consolidation or representational collapse.</p> <p>The system exhibits graceful degradation under interference and perturbation, preserves invariant symbolic structure, and avoids averaging or blending failure modes common in dense embedding-based memories. Crucially, reinforcement converges without external regularization, establishing curiosity as a bounded, use-dependent consolidation process rather than an optimization heuristic.</p> <p>These results provide a functional, testable definition of curiosity applicable to symbolic AI systems and contribute a new perspective on memory-centric intelligence architectures independent of biological or reward-based framing.</p> <p><strong>Notes on Availability:</strong><br>The implementation is proprietary to RAIT Enterprises and is currently deployed within a symbolic AI system under active evaluation through ARC benchmark testing. Code is therefore not publicly available at this time.</p> <h2><strong> Keywords</strong></h2> <ul> <li> <p>Curiosity</p> </li> <li> <p>Symbolic AI</p> </li> <li> <p>Fractal Memory</p> </li> <li> <p>Holographic Memory</p> </li> <li> <p>Memory Consolidation</p> </li> <li> <p>Interference Resistance</p> </li> <li> <p>Attractor Dynamics</p> </li> <li> <p>Basin Stability</p> </li> <li> <p>Computational Curiosity</p> </li> <li> <p>Use-Dependent Learning</p> </li> <li> <p>Non-Gradient Intelligence</p> </li> <li> <p>Artificial General Intelligence Foundations</p> </li> </ul>