Emergent Models
Emergent Models (EMs) are machine learning systems based on dynamical systems (typically cellular automata) that compute by evolving an internal state over time. Unlike neural networks that rely on static parametric functions, EMs operate through local update rules, repeated over time and space.
Initial State Programming
The initial state encodes the program. With Turing Complete rules (e.g. Rule 110, Game of Life), arbitrary algorithms can be computed by modifying only the initial configuration
Adaptive Computation
Internal halting mechanisms enable variable-length computation, adapting processing time to task complexity, making recursion and dynamic behaviors feasible
Beyond Neural Networks
EMs generalize algorithms deeply, understand low-level patterns, and enable meta-learning through self-modification of internal state
Research
We thank the support of the Wolfram Institute and ResearchHub for this project.