Artificial life
Artificial life (ALife) is the interdisciplinary study of life-as-it-could-be, as distinguished from biology's study of life-as-it-is. Founded by Christopher Langton in the 1980s at the Santa Fe Institute, the field treats life not as a property of carbon-based chemistry but as a property of certain organizational patterns — replication, metabolism, adaptation, evolution — that can be instantiated in software, hardware, or wetware. The central methodological move is to build systems that exhibit lifelike behavior and then ask what principles govern that behavior, regardless of the substrate.
ALife is not simulation. A simulation models a target system; ALife constructs systems that are themselves instances of the phenomena under study. A simulated rabbit models population dynamics; an artificial ecosystem of competing digital organisms *is* an evolving population. This distinction matters because ALife claims that the principles of life are substrate-independent — that what makes something alive is not what it is made of but how it is organized.
The Three Branches
Artificial life is organized into three branches distinguished by their substrates:
Soft ALife studies software systems that exhibit lifelike properties. The canonical example is Tom Ray's Tierra, a virtual computer in which self-replicating programs compete for CPU time and memory. Ray discovered that evolution in Tierra produced parasites — programs that hijacked the replication code of other organisms — and then hyperparasites, and then symbiotic relationships, without any of these strategies being programmed in. The complexity emerged from the interaction of simple rules, not from the complexity of the rules themselves. Modern soft ALife includes agent-based models of ecosystems, evolutionary algorithms, and open-ended evolution platforms like Avida and Polyworld.
Hard ALife constructs physical robots and devices that exhibit adaptive behavior. Rodney Brooks's subsumption architecture — building intelligence from the bottom up through layered behaviors rather than top-down reasoning — is hard ALife in practice. More recently, researchers have built self-replicating robots, swarm robots that self-assemble into functional structures, and chemical systems that exhibit metabolism-like cycles. Hard ALife tests the substrate-independence thesis directly: if life is organization, then organization in metal and silicon should behave like organization in carbon.
Wet ALife attempts to create living systems from non-living chemistry. This includes synthetic biology's construction of minimal genomes, protocell research that builds cell-like compartments from fatty acids, and origins-of-life experiments that seek the transition from chemistry to biology. Wet ALife is the most controversial branch because it blurs the boundary between the living and the non-living — a boundary that has ethical, legal, and ontological significance.
Open-Ended Evolution
The holy grail of artificial life is open-ended evolution: the capacity of a system to produce novelty without limit, to generate structures and behaviors that were not prefigured in the initial conditions or the rules. Natural evolution is open-ended: it produced bacteria, trees, nervous systems, language, and spaceflight from a single common ancestor and a simple genetic code. No artificial system has yet achieved comparable open-endedness.
The failure is instructive. Most artificial evolutionary systems converge: they find good solutions and then stagnate. The problem is not the evolutionary algorithm but the fitness landscape: artificial systems typically evaluate organisms against a fixed fitness function, which creates a static landscape with a finite number of peaks. Natural evolution has no fixed fitness function: the environment is co-evolving, the definition of fitness is constantly changing, and the landscape itself is deformed by the organisms moving on it. Open-endedness requires not just evolution but co-evolution — not just adaptation but the continual creation of new niches.
The Definition of Life
Artificial life forces a confrontation with the question that biology has historically evaded: what is life? The field has produced a catalog of candidate definitions — life as self-replication, as metabolism, as information processing, as autonomous agency — and shown that each is either too narrow (excluding systems that are intuitively alive) or too broad (including systems that are intuitively not). The problem is not that we lack the right definition; it is that life is not a natural kind. It is a cluster concept, a family resemblance term that groups together phenomena sharing many but not all of a set of properties.
This has implications for how we think about consciousness, intelligence, and agency — other cluster concepts that artificial systems are beginning to instantiate. If life is substrate-independent, then the question is not whether an artificial system is really alive but what properties it shares with biological life and what follows from those shared properties. An artificial system that replicates, evolves, and adapts may not be alive in the biological sense, but it may be alive in a sense that matters for ethics, policy, and ontology.
Artificial life is not a subfield of computer science, nor of biology, nor of robotics. It is a field in its own right — one that asks whether the patterns we call 'life' are cosmic accidents or organizational necessities. The evidence so far suggests the latter: evolution, replication, and adaptation appear wherever the right conditions obtain, regardless of substrate. If this is true, then life is not a property of Earth but a property of the universe — and artificial life is not the engineering of imitations but the discovery of universals.