Picture a swarm of 200 particles that keep jostling, aligning their velocities, and pushing toward a common goal. This swarm isn't just darting around in space: it's learning to predict chaos. An external signal, something like the unpredictable behavior of weather, is fed in through a single driver particle, and the rest have to guess where it will move next. Such a swarm is a physical reservoir computer, and its forecasting ability depends on how it handles entropy.
It turned out that the absolute magnitude of entropy doesn't matter much. What matters far more is how strongly the external signal alters the system's thermodynamic response. When the driver is connected, the heat flow into the environment rises by thousands of times, and the system's entropy by tens of times. It is this relative increase, not entropy itself, that correlates with the swarm's performance.
This finding changes how physical computers are designed. Instead of the costly training of every candidate, suitable regimes can be selected quickly by measuring how the system responds to an external signal. Such screening applies to swarms of active matter, photonic systems, and spin systems.
A swarm of particles that once looked like nothing but a chaotic jumble turns out to be a delicate instrument. Its entropy works like a compass pointing to computational ability. And the best compass isn't the one that shows a lot, but the one that responds keenly to an external signal.
The reservoir's maximum performance is achieved in a regime where heat flow into the environment is minimal in the absence of a signal, but increases thousands of times when the signal is applied.
The idea of using a swarm of particles for computation echoes the concept of 'computing matter' from Stanisław Lem's novel The Invincible, where a collective of microscopic machines processes information.