What connects the brain, the immune system, a city, an ant colony, and a large language model? The answer, explored intensively at the recent AI-ABC Workshop on Complexity Science, Complex Adaptive Systems, Neural Networks and AI, is this: all of them are complex adaptive systems — networks of interacting components that give rise to organised behaviour no single part could produce alone.

The workshop was a rich, hands-on exploration of ideas that sit at the intersection of science, technology, and philosophy of mind. It was, by any measure, a very successful one — the kind of gathering that generates more questions than it answers, and does so productively.

What Is a Complex Adaptive System?

A Complex Adaptive System (CAS) is characterised by several defining properties. It is composed of many interacting agents or elements — neurons, cells, people, software modules — each following local rules. Yet the system as a whole displays emergent properties: patterns, behaviours, and capacities that cannot be predicted simply by examining the parts in isolation.

CAS are also adaptive: they learn, reorganise, and evolve in response to their environment. They are neither fully ordered nor fully random — they operate at what complexity theorists call the edge of chaos, a regime of maximal flexibility and computational power.

Examples span every scale of nature and society: ecosystems, economies, immune systems, cultures, brains, and — increasingly — artificial intelligence architectures.

Neural Networks Through a Complexity Lens

One of the workshop’s central threads was the relationship between biological neural networks and their artificial counterparts. Modern deep learning systems — transformers, recurrent networks, convolutional architectures — are, in a technical sense, artificial neural networks. But how closely do they mirror the complexity of the brain?

The brain is not merely a network of weighted connections. It is a living dynamical system operating across multiple timescales simultaneously, self-organised through development, experience, and ongoing metabolic processes. Its computational power arises not from any single module but from the coordinated dynamics of billions of neurons operating in concert — and in tension.

Artificial neural networks, by contrast, are trained rather than developed, optimised rather than adaptive in the biological sense. They can perform remarkable feats — image recognition, language generation, strategic game-playing — but they remain brittle in ways biological systems are not. They fail catastrophically on inputs slightly outside their training distribution. They do not generalise the way humans do.

The workshop invited us to ask: what would it mean to build AI systems that were genuinely complex and adaptive — not just large and powerful, but dynamically organised in ways that resemble living systems?

AI as a Complex Adaptive System

Current AI systems are often discussed as tools — instruments that humans use to accomplish tasks. But from a complexity science perspective, they are increasingly becoming participants in complex adaptive systems: embedded in social, economic, and informational ecosystems, shaping and being shaped by the environments in which they operate.

This framing has profound implications. In a CAS, there is no central controller. Behaviour emerges from the interactions of many agents following local rules. When AI systems enter social ecosystems — recommending content, mediating communication, influencing decisions — they become agents in a system whose emergent properties no one has fully designed or can fully predict.

Understanding AI through the lens of complexity science is therefore not merely a theoretical exercise. It is a prerequisite for responsible deployment. We need to ask not just “what does this model do?” but “what kind of system does it help bring into being?”

Coordination and Synchrony

A particularly generative theme was coordination: how do elements in a complex system synchronise, differentiate, and maintain coherent organisation over time? In biological systems, coordination dynamics — the science of how coupled oscillators and distributed networks achieve flexible, context-sensitive entrainment — offers a powerful framework for understanding everything from motor control to social interaction.

In artificial systems, coordination is typically engineered: attention mechanisms, synchronisation protocols, shared memory architectures. But the workshop raised the question of whether richer, more biologically-inspired coordination mechanisms might yield more robust and adaptable AI — systems capable of the fluid, context-sensitive integration that characterises human cognition.

A Productive Day

The AI-ABC workshop brought together researchers, clinicians, engineers, and thinkers from different disciplines, united by a shared conviction: that understanding complex systems — and building AI responsibly — requires frameworks that go beyond linear causality and simple input-output models.

The conversations were lively, the ideas were genuinely challenging, and the sense of intellectual community was palpable. (Photos from the day are coming — watch this space.)

If the goal of such a workshop is to expand the conceptual vocabulary of everyone in the room, this one succeeded. Complexity science is not a niche specialty. It is a way of seeing — and in a world increasingly shaped by adaptive systems both biological and artificial, it is a way of seeing we urgently need.