Life as Art: Cellular Automata in Digital Practice

The first people to watch Conway’s Life run on a teletype terminal in 1970 were doing something that did not yet have a name. They were not programmers executing a task — they were observers of a self-contained universe, watching patterns form and dissolve and reorganize, following no designer’s intent except the four rules. The paper curled out of the teletype with gliders and still lifes and explosions printed on it, and the people watching recognized something aesthetic in what was happening. The mathematics was visible. The structure was beautiful. And the beauty was not added from outside but generated from within — a product of the rules themselves.

This is what connects Life to the broader tradition of computer art: not just that artists later used Life as a medium, but that Life established the principle that made digital art interesting. Beauty could be generated, not merely represented. A program could be a brush, not just a tool.

Vera Molnár: The Machine Imaginaire

Vera Molnár came to algorithmic art before she had a computer. Born in Hungary in 1924, she moved to Paris in 1947, where she was drawn to geometric abstraction — the tradition of Mondrian, Malevich, the Constructivists. By 1959, she had begun making what she called “machine imaginaire” works: drawings produced according to explicit rules she had set for herself in advance, working through all the permutations of a system by hand. The constraint was self-imposed, but the spirit was already algorithmic: instructions first, execution second, artist as rule-maker rather than mark-maker.

In 1968, she knocked on the door of the Paris University computer center and asked to use their machine. She later described the look she received from the department head — a look that suggested he was considering whether to call for a nurse. She was granted access anyway.

She learned Fortran. She learned to generate plotter drawings — line art executed by a mechanized pen on paper, following instructions she had coded into the system. The results were striking: series like À la recherche de Paul Klee (1969–70) and Transformations (1976) explored the systematic variation of geometric forms across hundreds of states, producing the visual experience of a rule thinking through its own implications. Each drawing in a series was a record of the system’s state at a particular moment; the series as a whole was the system’s behavior made visible.

Molnár’s work is not Life-based — she used her own rule systems rather than Conway’s rules. But she is the intellectual ancestor of every artist who has used CA as a creative medium, because she was the first to articulate and practice the principle: the artist’s job is to write the rule. The art is what the rule produces. The artist is not the author of the image but the author of the system that generates the image.

Molnár died in 2023 at the age of 99, still working. A section of the 2022 Venice Biennale had been devoted to her plotter drawings from the 1970s. She had lived to see her “machine imaginaire” become the standard paradigm of digital art.

Harold Cohen and AARON (1973)

Harold Cohen was a British painter who had exhibited at the Venice Biennale and the Documenta before he became interested in artificial intelligence. The question that consumed him was: what does a painter know? Not what they feel, not what they intend, but what they actually know, in the sense of rules and heuristics and encoded responses to visual situations?

In 1971, he presented an initial prototypal painting system at the Fall Joint Computer Conference, and was invited to Stanford’s Artificial Intelligence Laboratory. In 1973, AARON was born.

AARON is a rule-based program — not a neural network, not a generative adversarial network, not anything that learns in the current sense. It is a set of rules Cohen wrote, refined, and extended over four decades, specifying how to construct images: how figures inhabit space, how marks of different kinds relate to each other, what constitutes a visually stable composition. Cohen also built the output devices: plotters and painting machines that executed AARON’s instructions with actual brushes and pens on actual paper and canvas.

AARON’s connection to cellular automata is philosophical rather than technical. AARON does not use Life’s rules. But Cohen was articulating the same question that Life raised: what is the minimum specification for generativity? If you encode enough knowledge about mark-making and composition into rules, can the rules themselves produce art? Cohen believed yes, and spent his career proving it. By the time of his death in 2016, AARON had been running continuously for over forty years, the longest-running AI system in art history, and it had produced works exhibited at the Tate Gallery, the Brooklyn Museum, and the Computer History Museum.

The question Cohen was asking — can rules generate meaning? — is the question that Life also raises, in a different register.

Some artists used Life itself, not just its principles. The aesthetic appeal was direct: Life patterns are beautiful. Gliders have a visual elegance that is not incidental but structural — the form follows from the function. The Gosper Glider Gun, with its double-lobed oscillating body and its regularly spaced glider emissions, looks like something a jeweler might design. The R-pentomino’s chaotic explosion and eventual resolution into still lifes has the character of a miniature drama.

In the early 1970s, Life patterns appeared as decorative elements in technical publications, on department bulletin boards, and as screen art on the first graphics terminals. The community that had gathered around Gardner’s column was producing thousands of configurations, and many of them were appreciated as much for their visual properties as for their mathematical interest.

The more formal use of Life as a visual medium came later, as the generative art movement developed a vocabulary for discussing what it meant to make art with algorithms. Jared Tarbell’s work in the early 2000s — web-based pieces that used cellular automaton rules alongside other algorithmic processes — exemplified the new practice: Life’s rules were components of a larger system, producing texture and structure within pieces that also incorporated particle systems, noise functions, and other generative elements.

Casey Reas and Processing (2001)

Casey Reas is, more than any other single figure, responsible for the mainstreaming of generative art as an artistic practice. In 2001, working with Ben Fry at MIT, he created Processing — a programming language and environment designed specifically for visual artists, which made algorithmic image-making accessible to people without computer science backgrounds. Processing became the standard tool of generative art education and is now used by hundreds of thousands of artists and designers worldwide.

Reas’s own practice is deeply engaged with cellular automata and related systems. Early works like Cells (2000) and Edge (2000) employed simple rules to simulate organic growth patterns, allowing forms to evolve through iterative rule application. His Process series, begun in 2004, translated concepts from artificial life — simulated neurons, emergent behavior, local interaction — into visual systems that produced complex images from minimal instructions.

What Reas brought to the conversation was a precise aesthetic theory of why algorithmic generation was interesting, as opposed to merely novel. In his framing, the interest lies in the gap between the simplicity of the rule and the complexity of the output — the same gap that makes Life compelling. A piece that generates ten thousand different images from a three-line rule is interesting because of the relationship between rule and result, not because of any individual image. The art is the function, not the output.

Reas has explicitly cited Life and Conway’s rules as foundational to his thinking. The Processing community has produced hundreds of Life implementations, from simple educational tools to complex visual explorations. Processing made Life accessible as a generative medium in the same way that Gardner’s column made Life accessible as a mathematical object.

William Latham and Evolutionary Form (1990s)

William Latham came to computation from a direction opposite to Molnár’s and Cohen’s. He was a sculptor interested in organic form — in the logic of growth and variation that produced shells, bones, and the branching structures of plants. His early work at the Royal College of Art in the 1980s consisted of enormous hand-drawn evolutionary trees: diagrams showing how a simple form could be varied systematically through thousands of mutations, filling pages with organic-looking structures that had never existed in nature.

In the late 1980s, as a Research Fellow at IBM, Latham collaborated with mathematician Stephen Todd to build MUTATOR — software that formalized his evolutionary trees as a computational process. A starting form was represented as a grammar, the grammar was mutated randomly, and the results were displayed for Latham to select from. Selected forms were mutated again. Selected again. The process was exactly natural selection, with Latham serving as the selective pressure: he chose the forms he found aesthetically interesting, and those were the ones that reproduced.

The resulting sculptures — rendered as 3D computer graphics and later as physical objects — looked unlike anything produced by conventional means. They were organic in the way that sea creatures are organic, but they were alien, following the logic of an evolved system in a different environment. Latham’s work was exhibited at the Institute of Contemporary Arts in London and collected widely.

The connection to Life is the evolutionary principle itself. Life’s patterns do not evolve — the rules are fixed. But Karl Sims’s virtual creatures and William Latham’s MUTATOR sculptures are both downstream of the ALife tradition that Life inspired, in which evolution is understood as a computational process that can run on any substrate. Latham made that process visible as sculpture.

The Aesthetic Logic of CA

What makes cellular automata aesthetically interesting is not merely that they produce complex images. It is the character of the complexity they produce.

CA images have a specific texture: they are locally irregular but globally structured. Zoom in to any region of a Life simulation in mid-run and you see chaos. Zoom out and you see persistent structures — gliders, oscillators, stable regions — embedded in that chaos. This multi-scale structure, with order at one scale and disorder at another, is the same structure that makes natural textures (wood grain, rock formations, biological tissues) visually interesting. CA images have the quality of nature without being nature.

They are also temporal. A Life pattern is not an image but a process, and the image you print at any given generation is a cross-section of that process. Artists have explored this temporality in different ways: as animation (the obvious choice), as superposition (printing multiple generations in different colors on the same surface), and as sonification (mapping cell states to musical parameters to produce generative music).

The sonification tradition deserves its own mention. Composers including Brian Eno (not Life specifically, but cellular-automaton-adjacent generative systems), and numerous experimental musicians have used Life and Life-like rules to generate musical structures. The mapping choices — which cell parameters map to pitch, which to duration, which to timbre — are themselves aesthetic decisions, and different mappings produce strikingly different music from identical underlying patterns. The temporal structure of Life — patterns that stabilize, oscillate, or drift — translates naturally into musical form.

Contemporary Practice

Today, cellular automata appear in digital art across every medium and at every level of technical sophistication. Instagram and X host communities of generative artists, many of whom use Life-like rules as components of their practice. NFT platforms have hosted Life-based generative art collections. Gallery installations use large-format displays running Life in real time, sometimes at scales where individual cells are meters across.

The tools have changed. The principle has not. Life established that beauty could be generated — that a system of simple rules could produce complex visual structures that were worth looking at for their own sake. Every digital artist working with algorithmic generation is working in the tradition that Life helped define.

Molnár said once that she was interested in “controlled disorder” — art that occupied the region between pure order (boring) and pure randomness (meaningless). Life occupies exactly that region, and it occupies it in the minimal possible way: the simplest system known that produces controlled disorder at every scale.

That is why artists keep returning to it. It is not the most powerful generative system available. It is the clearest demonstration that generativity is real.

Further Reading