On generative art and what humans and machines can share

The first paragraph of this text concluded with an intuition I marked as an inference, rather than a conclusion. When discussing the work Rain Blooms, by Kazuhiro Tanimoto, I suggested that between very different entities—a human and a computational system—the common ground may not be shared semantics, but a shareable process. It would not be necessary for both to feel the same for an aesthetic approximation to exist; it would suffice that they could trace the same dynamics of selection, transformation, and emergence.

This text tests that intuition against an entire artistic tradition. The question is whether the hypothesis holds beyond a single case: is it true that in a significant portion of generative art, what circulates, is reused, and accumulates is not a stable meaning but a process—rules, code, parameters, execution infrastructures? My thesis is that it is, but with a nuance that proves decisive: the shareable process does not eliminate semantics; it displaces them. What is shared ceases to be "what the work means" and becomes "how it produces meaning and variation." And that displacement is, precisely, what makes generative art a privileged territory for considering what a human and a machine can and cannot share.

1. What is meant by “meaningless process”

It is advisable to dispel a misconception from the outset. "Meaningless processes" does not imply that these works lack meaning. It signifies something more precise: that the system generating them operates without the need to comprehend what it produces. A cellular automaton executes local neighbourhood rules; it does not know it is producing a form that a human will interpret as an organism, an explosion, or a flower. Complexity emerges from the interaction of blind rules. Meaning is provided, subsequently, by the observer.

This clearly separates two planes that the conversation regarding art and artificial intelligence tends to conflate. The plane of the process: the rules, the code, the dynamics that produce the work. And the plane of meaning: what the work evokes, represents, or triggers in an observer. In mature generative art, these two planes do not coincide. The system inhabits the former; the human, the latter. And the thesis of this text is that what truly circulates between artists, platforms, and generations—what is inherited, branched, and reprogrammed—is the first plane, not the second.

2. A genealogy, not a catalogue

To support this, more than one example is required, yet not a flat inventory. A genealogy is needed: a line that demonstrates how the same idea is transmitted and transformed over six decades.

The conceptual prologue is Max Bense. At the Stuttgart School in the 1960s, influenced by information theory, Bense formulated the idea of generative aesthetics: art could be conceived as the regulated production of aesthetic states, stripped of the romantic concepts of genius or inspiration. His text Projekte generativer Ästhetik (1965) provides the foundational vocabulary: the machine calculates permutations of signs, and the human extracts aesthetic relevance from the combination. Bense functions as a manifesto, not a case study—it is unwise to burden him with the weight of the argument regarding emergence—yet he establishes the question.

The true matrix is Conway. The Game of Life (1970), popularised by Martin Gardner, crystallises the central argument of the entire subsequent tradition: a minimal set of local rules can generate behaviours of complexity impossible to anticipate through inspection. It is a “zero-player game”: once the initial conditions and rules are set, the system evolves independently. And what is decisive for this thesis is how it is shared: through rulestrings, patterns in compressed format, wikis, and simulators. What circulates in the Life community is not an image or a meaning, but an operative grammar that anyone can execute, vary, and extend. The portability of the process is maximised because a compact notation exists.

The next link is Karl Sims, and it is appropriate to treat him as a double block. In Evolving Virtual Creatures (1994), Sims does not design the forms: he programs a virtual genetic code and physical laws, allowing evolutionary selection to produce, over generations, morphologies and locomotion strategies that no designer had foreseen. In Galápagos (1997), that process becomes publicly shareable in a new way: the visitor to the installation does not share a closed meaning, but a selection function within a visible evolutionary system. They choose what survives. The collective “we” is produced through intervention in the process, not through consensus on meaning.

Explicit formalisation arrives with Casey Reas. In his Process series—documented in the Process Compendium—Reas transfers the logic of Sol LeWitt’s instructions to executable software. He commissions multiple distinct implementations of the same textual structure precisely to isolate three planes: interpretation, material, and process. The work is not a final image; it is the system and its possible realisations. And with Processing—the environment he created with Ben Fry in 2001—that process becomes not only shareable but teachable, versionable, and transmissible. Reas is the case where the thesis becomes almost literal: the work resides in the rules and their translations, not in a unique image.

Scott Draves takes the process to a distributed scale. Electric Sheep (1999 onwards) is free software that functions as a form of collective artificial life: thousands of computers generate and render fractal animations, and users vote on which ones survive and reproduce in a global gene pool, with an archive of lineages. Here, the shareable process ceases to be a study rule and becomes social and technical infrastructure: the work exists as a gene pool, votes, and genealogies. The “we” does not share iconography; it shares an operative criterion of selection.

The penultimate step is Lenia. It is not merely a descendant of Conway: it is a continuous reformulation of the cellular automaton that produces plastic, resilient, and adaptive "life forms," with an open ecosystem of code, versions, and demos. Lenia demonstrates that the tradition did not end with Life, but continues to produce systems where the importance lies in the ecology of rules and parameters, not in stable iconographic semantics. It is proof that the lineage remains vibrant and continues to produce genuine emergence.

The conclusion is Tanimoto, not due to chronology but to argument. In Rain Blooms, Tanimoto insists that he works with an original cellular automaton, not a reproduction of classical models, and the presentation of the work underscores that what is designed is not a predetermined final form, but a system of local rules at the cell level. It is the contemporary formulation that best condenses the thesis: the artist designs the regime of emergence, not the image; the image and sound are products of the same dynamics; beauty appears as a consequence of the system, not as a prior objective.

3. The nuance that changes everything

Having reached this point, the thesis appears solid: throughout this genealogy, what is shared is the process, not the meaning. However, it is prudent to resist the absolute version, because the evidence qualifies it, and the nuance is more interesting than the raw thesis.

The thesis does not hold equally in all cases. In installations where access to the process is provided as an embodied experience—a pool with evolutionary creatures, an interactive environment, a strong ecological metaphor—curatorial semantics remain decisive for the public to understand what is at stake. The concept of "artificial life," of "ecology," of "non-human agency" does not emerge from the process on its own: it must be named, staged, and legitimised. In such cases, the process may be the transferable substrate, but the semantics are the layer of entry and mediation.

Therefore, the precise formulation is not 'process instead of meaning'. It is this: the shareable process shifts semantics from a fixed iconographic content to a procedural semantics. What is shared is no longer what the work represents, but how it produces meaning and variation. Conway's work does not 'mean' anything in the figurative sense; yet to share its rules is to share a way of generating behaviour, and that way possesses its own intelligibility. Meaning does not disappear: it changes location. It moves from the object to the process.

This has a consequence for portability that is worth noting. The process is shared most effectively when a compact notation exists: Conway's rulestrings, Processing sketches, Lenia's kernels and parameters. When the work can be reduced to a small set of instructions interpretable by different agents, sharing becomes cumulative, generational, communal. When it depends on unique hardware or unpublished software, the process becomes opaque and transmission breaks down. Shareability is not a binary property: it is a gradient, and it depends on whether the process has a transmissible form.

4. What, then, do human and machine share

I return to the underlying question of this series. If vital weight—what we observed in the previous text—requires a self that the machine lacks, what remains as a common ground between a human and a generative system? The answer this genealogy allows us to provide is: the process, not the meaning.

When a human contemplates a cellular automaton in evolution, something different occurs than when they contemplate a figurative painting. They are not reading a representation that another subject intended to convey to them. They are tracking a dynamic—how local rules generate order, how the system maintains itself on the edge between repetition and chaos—that the machine executes without understanding. Human and system are, in that moment, synchronised in the process: both attend to the same emergent dynamic. But they remain separated in meaning: the human injects sense, memory, vital weight; the system injects nothing, because it has no source from which to do so.

This is what makes generative art a privileged territory for the question of this series. Not because it demonstrates that the machine understands—it does not—but because it exhibits the exact boundary with unusual clarity: there is a shareable zone, the process, and a non-shareable zone, the lived meaning. Tanimoto's work is valuable precisely because it makes that boundary perceptible. It reveals a regime of emergence that human and machine can both track, while making it evident that only one of the two can experience what that regime produces.

This is the affirmative version of an argument that was critical in other texts of this series. There, I showed what the machine does not share with us. Here, I show what it does: a formal dynamic, a process of selection and transformation, a traceable emergence. It is not insignificant. It is, perhaps, the only honest form of common ground between such distinct entities: not to pretend that they share meaning, but to recognise that they can share process.

5. Conclusion

Three conclusions.

The first, descriptive. In much of generative art—from Conway to Tanimoto, via Sims, Reas, Draves, and Lenia—what circulates, is inherited, and is transformed is not a stable iconographic meaning, but a shareable process: rules, code, parameters, execution infrastructures. The identity of many of these works resides in their architecture of rules, not in a final image.

The second, nuanced. The shareable process does not eliminate semantics: it shifts it. From a fixed content to a procedural semantics—how the system produces sense and variation. And the portability of the process depends on the existence of a compact notation that different agents can execute; where there is none, transmission breaks down.

The third, regarding the machine boundary. Generative art exhibits with unusual clarity what a human and a machine can share and what they cannot. They share the process: the dynamic of emergence that both can track. They do not share the meaning: the sense and the vital weight that only the human can inject, because only they have a source from which to do so. Recognising that boundary—without feigning understanding where there is only computation, without denying the common ground where it exists—is the precise way to consider the aesthetic relationship between entities.

Processes without meaning names, in the end, an honesty. The machine generates order without understanding it. The human understands without having generated it. And on the edge where both meet—the visible process, the traceable emergence—there is a form of art that does not pretend that the machine feels, nor that the human calculates, but that reveals, with a clarity few other practices achieve, where what can be shared ends and where what each contributes alone begins.

On open conversation

This text is the fourth in the series on regimes of relevance that I have been developing at NeuroArt: Cognitive Surplus. It expands upon the intuition regarding the shareable process that the first text outlined concerning Tanimoto, extending it to the tradition of generative art and artificial life. The following and final planned step addresses the limits of interpretability: why a system's regime of relevance remains opaque even when one has access to its internal mechanisms.

Should anyone wish to contribute from the fields of generative art, the history of computational art, artificial life, or complex systems theory, this notebook remains open.

Sources

Bense, M. (1965). Projekte generativer Ästhetik (Rot, Nr. 19). Stuttgart: Elisabeth Walther. (Formulation of generative aesthetics as the regulated production of aesthetic states.)

Gardner, M. (1970). Mathematical Games: The fantastic combinations of John Conway’s new solitaire game “life”. Scientific American, 223(4), 120–123.

Galanter, P. (2003). What is Generative Art? Complexity Theory as a Context for Art Theory. GA2003 – 6th International Conference on Generative Art. Milan: Politecnico di Milano. (Definition of generative art as the initiation of an autonomous system.)

Sims, K. (1991). Artificial Evolution for Computer Graphics. SIGGRAPH ’91. https://doi.org/10.1145/122718.122752

Sims, K. (1994). Evolving Virtual Creatures. SIGGRAPH ’94. (Morphologies and emergent behaviours through evolutionary selection.)

Sims, K. (1997). Galápagos. (Interactive installation of evolution guided by visitor selection.)

Reas, C. (2010). Process Compendium 2004–2010. Los Angeles: REAS Studio. (Documentation of software structures as processes: separation of concept, implementation, and process.)

Reas, C., & Fry, B. (2006). Processing: programming for the media arts. AI & Society, 20(4). https://doi.org/10.1007/s00146-006-0050-9

Draves, S. (1999– ). Electric Sheep. (Distributed artificial life with collective selection and lineage archiving.)

Chan, B. W.-C. (2019). Lenia: Biology of Artificial Life. Complex Systems, 28(3). https://doi.org/10.25088/ComplexSystems.28.3.251

Wu, S., Qu, H., & Zhang, Y. A Survey of Recent Practice of Artificial Life in Visual Art. Artificial Life. https://doi.org/10.1162/artl_a_00433

Kawahara Lino, Joana. “The Rule of the Bloom | Kazuhiro Tanimoto.” Right Click Save, 2026. https://www.rightclicksave.com/article/the-rule-of-the-bloom-kazuhiro-tanimoto-cellular-automata-neort-art-blocks

Esteban Ruiz, J. A. What matters and what is relevant. Public notebook at juanesteban.art, 2026.

Esteban Ruiz, J. A. The effect of itself. Public notebook at juanesteban.art, 2026.

Esteban Ruiz, J. A. Art as Structural Surplus: Toward a Relational Ontology Beyond Human Authorship (V2.3). PhilArchive and Zenodo, 2026.