Swarm Intelligence: A Reading Note
Chapter 5: Swarm Intelligence
Previous sections explored connectionism and evolutionary mechanisms, each revealing how groups generate capacities no individual member possesses. In neural networks, each unit performs trivial computation; aggregated and connected, they produce powerful inference. In genetic algorithms, unremarkable individuals evolve collective adaptability through selection, crossover, and mutation.
Ethology reveals a third form. Flocking in birds and schooling in fish exhibit collective intelligence without central control. These groups may contain social roles such as leaders and followers, but no individual directs the whole. Each agent acts on local information, yet the collective moves as a coherent, orderly system, as if governed by a single mind. Humans display analogous behavior at busy intersections: when the light changes, pedestrians from four directions move simultaneously, following implicit rules, navigating without collision or coordination.
The social behavior of gregarious animals differs from ours in a fundamental respect: it operates without deliberation. Animals act for survival, communicating through instinct rather than inference about others' intentions. The book's example is termite nest construction. Humans build houses through a chain of cognition: architects design, engineers test structural integrity, authorities issue permits, managers coordinate labor. Termites, without any central command, construct arch-shaped nests millions of times their own body size. The mechanism is simple. Each termite carries soil to the area with the strongest pheromone concentration and deposits it there. This creates positive feedback: random initial searching leads to pheromone accumulation at certain points, which attracts more deposition, forming pillars. As multiple pillars emerge, the strongest pheromone gradients shift to the spaces between them, guiding termites to bridge the gaps and produce arches. Human arch construction, by contrast, requires shape recognition, material selection, method evaluation, and role assignment. Two kinds of intelligence, built on entirely different substrates.
Communism, in theory, aspires to a termite-like world where each person performs a role and collective strength emerges from the sum. But human cognition is not termite cognition. We noted earlier that both evolution and the mind are adaptive systems: evolution adapts populations to environments, the mind adapts individuals. When both operate simultaneously, individual minds seek to maximize personal benefit and externalize costs. Under a communist system, individuals discover that not working yields greater reward than working, and productivity collapses. Game theory formalizes this. It describes how individual choices at the micro level aggregate into group-level dynamics at the macro level. The Nash equilibrium (the strategy where no player benefits from unilateral deviation) becomes everyone's default. The irony is that Nash equilibrium presupposes theory of mind, the capacity to infer what others will choose. The very cognitive ability that distinguishes us from termites is what prevents us from achieving their seamless cooperation. Yet selfishness and altruism need not conflict. If incentive structures guide the Nash equilibrium toward cooperative outcomes, human groups can exhibit collective intelligence without requiring individuals to abandon self-interest.
Numerous species demonstrate robust adaptability with brains far simpler than ours, some containing only a few hundred neurons. Relying on instinctual behavior alone, they have survived millions of years of evolutionary pressure. The intelligence these species exhibit (bat echolocation, the hexagonal geometry of beehives, the hovering flight of hummingbirds) is no less remarkable for being unconscious. Bionics (a fusion of biology and electronics) allows us to implement in machines the solutions evolution discovered over millennia, capabilities we never evolved ourselves.