Swarm Intelligence: A Reading Note
Chapter 5: Swarm Intelligence
In previous discussions, we explored the applications of connectionism and evolutionary mechanisms, marveling at the power exhibited by groups. In neural networks, each unit performs simple computations, but when aggregated and connected, they demonstrate formidable strength. Similarly, individuals in genetic algorithms may not possess remarkable capabilities, but the mechanism of evolution enables the group to develop adaptability.
In ethology, observations of flocking behaviors in birds and fish reveal another form of collective intelligence. These biological groups may have social roles like leaders and followers, but there isn't a central control system. All individuals act according to their will, yet they form a massive, harmonious, and orderly collective, as if controlled by a single mind. We observe similar human group movement at large intersections; when the light turns green, people from all four corners move simultaneously, following certain rules, moving in an orderly fashion without collisions.
The social behaviors of gregarious animals are not like our social behaviors, which involve considering various interests and others' thoughts. They act purely for survival, communicating with others instinctively. An example mentioned in the book is termite nest building. Humans need architects to design and test buildings for safety, obtain permits, and manage construction to build a sturdy house. Termites, without a central command, can build arch-shaped nests millions of times larger than themselves. Closer inspection of individual behavior reveals simple actions: carrying soil to the area with the most pheromones and depositing it where pheromones are strongest. This mechanism creates a positive feedback system, starting with random searching and leading to the accumulation of soil at pheromone-dense areas, forming pillars. As several pillars emerge, the strongest pheromone points are influenced by other pillars, guiding termites to build arch structures. In contrast, human construction of arches involves recognizing shapes, deciding where materials fit best, evaluating methods, and understanding who is building the door or roof, demonstrating a different kind of intelligence.
Communism, in theory, aspires to a termite-like world where everyone performs their role, contributing to powerful collective strength. Unfortunately, human cognition is far more complex than other species. We know that both the mind and evolution are systems—evolution allows populations to adapt to the environment, while the mind enables individual adaptation. Individuals, seeking to maximize their benefits, often externalize costs to others. Under a communist system, people realize that not working is more beneficial than working, leading to a lack of productivity and a dysfunctional society. Game theory describes this phenomenon: how individuals influence each other, and how, when the mind and evolution operate simultaneously, choices are made on a micro level, and how groups evolve on a macro level. The most self-beneficial choice, known as Nash equilibrium, eventually becomes everyone's choice. Interestingly, the premise of Nash equilibrium is the ability to infer others' thoughts—mind theory. Despite our capacity to understand others, this ability hinders our unity. However, selfishness and altruism aren't necessarily conflicting. If game theory is applied appropriately, guiding Nash equilibrium towards altruism, it can unite humans to complete tasks and exhibit collective intelligence.
Numerous species in nature demonstrate robust vitality, with brains far less complex than ours, some with only a few hundred brain cells. Yet, relying solely on instinctual behaviors, they escape evolutionary elimination. The intelligence displayed by these species, selected by natural selection, is no less remarkable than that of humans. Bats use echolocation, the structure of beehives, the flight pattern of hummingbirds—humans, in awe, have learned to utilize these intelligences, developing various technologies. "Bionics," a portmanteau of biology and electronics, allows us to implement biological evolutionary intelligence in machines, despite not evolving these abilities ourselves.