Tag
This paper introduces Self-Organising Digital Circuits, using a topology-masked Transformer to configure lookup tables in Boolean gates, enabling circuits to self-assemble and self-repair around hardware faults. It demonstrates near-perfect recovery from soft errors and generalization to larger circuit scales, bridging biological self-organization with digital hardware resilience.
This paper proposes 'learnable novelty' as a unified principle underlying intelligence across statistics, complex systems, and adaptive behavior, and provides a differentiable estimator using reservoir computing that demonstrates complexity generation, abstraction, and exploration without supervision.
Sakana AI and collaborators introduce Smart Cellular Bricks, physical modular units running identical Neural Cellular Automata that collectively infer their global shape through local communication, with no central controller. The system demonstrates robustness to noise and failures, enabling shape classification and damage recovery.
This paper introduces Meta Neural Cellular Automata (MetaNCA), a framework that learns local update rules to self-organize the weights of neural networks without backpropagation, scaling to networks of 2 million parameters on MNIST and CIFAR-100 and generalizing to unseen architectures.
Introduces Neural Particle Automata, a method for learning self-organizing particle dynamics using smooth particle hydrodynamics perception, enabling particles to have local perception vectors for an update rule, analogous to Neural Cellular Automata but on continuous particle positions.
Introduces High-Res Neural Cellular Automata that operates on a coarse lattice and uses a Local Pattern Producing Network to generate high-resolution outputs, enabling efficient procedural generation.
This article explores neural cellular automata as a computational model inspired by biological morphogenesis and regeneration, demonstrating how simple local rules can lead to complex global behaviors.
Mariano Salcedo, a student at MIT's new Music Technology and Computation Graduate Program, is using neural cellular automata to generate self-regenerating visuals driven by music audio streams.