Slime Mold
A brainless organism grows a trail network between Munich landmarks - and rediscovers the map.
A single-celled organism with no brain, no neurons, nothing that could be called a plan, is placed on a map of Munich with food at the landmarks. Left alone, it grows a network of tubes between them, then prunes it, and arrives at something close to the U-Bahn an engineering team spent decades designing.
The concept
Physarum polycephalum explores by spreading everywhere, then reinforces the tubes that carry the most flow and lets the rest wither. There is no global optimiser; each piece of the organism follows a local rule, lay down trail, follow trail, thicken what is used. Out of that comes a network that balances total length against robustness, the same trade-off a transit planner agonises over.
The math
The simulation uses an agent-based scheme: thousands of particles each sense the deposited trail field $T$ a little ahead and to each side, and steer toward the strongest:
then deposit a bit of trail where they step. The field diffuses and decays each tick ($T \leftarrow (1-\rho)\,(T * G)$, a blur $G$ times a decay $\rho$), so unused paths fade and used ones sharpen. Deposit, sense, diffuse, decay: four local rules, no map of the city anywhere in them.
Why it stays strange
The organism is solving, approximately, a hard network-design problem, the kind that is formally intractable at scale, using only chemistry and feedback. It rediscovers real rail maps not because it understands cities but because reinforcing flow and pruning waste is what good networks are. Intelligence, here, is a property of the process, not of any part.
Further reading
- Tero et al. (2010), "Rules for Biologically Inspired Adaptive Network Design" (the Tokyo rail study).
- Jones (2010), on multi-agent Physarum models.
@article{tero2010,
author = {Tero, Atsushi and others},
title = {Rules for Biologically Inspired Adaptive Network Design},
journal = {Science},
volume = {327},
number = {5964},
pages = {439--442},
year = {2010}
}