Here's an idea that has been stewing in my head for months.
I implore you to ignore my shortcomings and just think the general idea over to see if there are some merits:
Let's start with a few questions:
Is it possible to build neuromorphic processors by taking advantage of the turing completeness of cellular automata by using a hardware implementation.
If a hardware cell could independently change its state according to a ruleset, wouldn't that mean we would have a clock-less massively parallel computer similar to the brain ?
If we could create a hardwired cellular automata that implements rule 110. Couldn't we perform arbitrary computation ?
Now here's the main idea:
The idea revolves around a 3D gel-based substrate containing conductive particles, which interact according to localized, decentralized rules inspired by cellular automata. In this concept, the nodes or neurons are not discrete components like those found in traditional hardware-based neural networks. Instead, they emerge from the conductive hydrogel substrate through local interactions between its constituents.
These conductive particles create pathways within the hydrogel matrix, altering the material's overall conductivity and enabling the formation of neuron-like structures. The particles self-organize according to rule sets without any external influence. As they establish new connections or sever existing ones, the information flow within the hydrogel matrix shifts, effectively implementing the cellular automata ruleset.
Each "cell" or region within the gel operates autonomously and communicates with neighboring cells. The cellular automata rules govern their behavior, collectively resulting in the processor's emergent computational capabilities.
Some literature:
All-Polymer Conducting Fibers and 3D Prints via Melt Processing and Templated Polymerization
https://pubs.acs.org/doi/10.1021/acsami.9b20615
Neuromorphic Computation Using Quantum-Dot Cellular Automata
https://ieeexplore.ieee.org/document/8123685
