Peter Bauman (Monk Antony) reflects on Philip Galanter's classic 2003 definition of generative art before briefly speaking with the artist, Texas A&M professor and theorist about the definition twenty years later.
About the Author
Peter, an arts writer, is responsible for Le Random’s editorial branch.
Decoupling Generative Art with Philip Galanter
Peter Bauman (Monk Antony) reflects on Philip Galanter's classic 2003 definition of generative art before briefly speaking with the artist, Texas A&M professor and theorist about the definition twenty years later. Dialogue thrives on a foundation of shared language. For both participants and non-participants in the generative art space, grasping the nuance of the discourse can be a challenge. We only amplify this challenge with imprecise language, especially concerning the definition of generative art itself.
What do we mean when we say generative art? How do we determine whether a work is generative? Do we include non-digital practices? What is autonomy in art? What is an autonomous system? How we talk about and communicate the answers to these questions is vital to generative art's long-term success and acceptance into the art historical canon. We have to understand ourselves before we expect even interested parties to understand us. To our great luck, the foundation to our shared understanding has already been laid.
Generative art refers to any art practice where the artist uses a system, such as a set of natural language rules, a computer program, a machine, or other procedural invention, which is set into motion with some degree of autonomy contributing to or resulting in a completed work of art.
We can simplify this definition to four words: Art using autonomous systems. A word by word analysis of this simplified definition reveals several overlooked elements critical to the understanding of what generative is and is not. A) art
Uh oh, "What is art?" I tend to go with the if-an-artist-says-it's-art-then-it's-art definition but feel free to substitute your own.
What methods can generative artists use? This is critical. No method is specified, meaning that digital and analog practices are both equally valid as generative. Since 2003, Galanter has spoken of generative art "uncoupled from any particular technology," highlighting that generative art does not have to be code or computer based. Analog practices such as the quilters of Gee's Bend, Sol LeWitt and Carl Andre not only serve as canonical examples of the movement, they continue to inspire today's digital practices.
What is autonomy? According to Galanter, this means not requiring "moment‐to‐moment decision making or control by the artist." In other words, it is when an artist loses direct control over the process. Here is where we find much of the debate as to whether a work is generative. If artists themselves directly choose the constraint, then we typically do not consider this to be generative. In contrast, if the artist allows an external system to directly constrain specific actions, then we do consider that system to be autonomous and, therefore, generative. Next, we'll look at an example of both.
In fact, an identical work can be completed with both a non-generative and generative process. Consider a white, one-pixel work of art. An artist that chooses to constrain themselves by only working with white and using that to determine the pixel's color would not be incorporating an autonomous system into their work. This practice would not be generative because the "system" that chose the color was the brain of the artist and not external; the artist maintained control throughout. This identical white pixel, however, could also be created with a generative process. An artist could use an algorithm that determined the pixel's white color based on a timestamp of the work's creation. In this case, the individual pixel art was created using a generative process.
Autonomy is unfortunately not always black and white; it can be a matter of degree and there will be gray areas. I believe the key question to determine a creative process's generativity is:
Does the artist relinquish control to an external system at some point?
If this moment-to-moment control is relinquished, then the practice could be understood as generative. It does not mean that code, randomness or procedures necessarily had to be involved in the process as well. When in doubt, adhering to the artist's intentions is appropriate.
What are systems? Amy Goodchild considers there to be three types of autonomous system, "randomness, rules and natural systems."
In 2008, Galanter listed several examples of the three types: "Systems may include natural language instructions, biological or chemical processes, computer programs, machines, self‐organizing materials, mathematical operations, and other procedural inventions." An updated paper from 2019 includes "generative deep learning AI systems" as well.
The key point is the variety of different systems that can be considered generative. What Galanter's definition spotlights in our on-chain generative art community is that we are part of an artistic tradition that stretches back millennia. Our tools have changed from memorized basket weaving algorithms passed down through generations to computer code. Yet both can be seen as equally generative. Through this shared understanding we can better grasp generative art's place and importance in and throughout art history.
In 2003, Galanter wrote of decoupling generative art from "any particular technology." At that point in time, he was referring to technology like computers and code.
In 2023, by recalling this broader definition of generative art, we are again reminded to decouple generative art from the latest technology, NFTs. While NFTs have played a critical role in unlocking decades of pent-up demand for digital art and broadly enabling the resurgence of this movement, they are not the defining feature of generative art.
By thinking of generative art only in terms of NFTs, we make the movement smaller, we make it shorter and we ultimately make it insignificant.
The following is a brief conversation I had with Galanter about his background and definition today.
Peter Bauman: Can you talk about your background and what led you to generative art and academia?
Philip Galanter: I was one of those kids who took everything apart to see how it worked. Clocks. Radios. I had a chemistry set and a microscope. This led to an interest in what I would later learn to call emergence and complex systems. How do form and function arise from the interaction of simple parts, achieving the “more than the sum of the parts” effect?
So it was basically a playful creative drive that led me to systems and science. I had a latent interest in art and music too but it didn’t really flower until I went to college. There I would sign up every semester for one hour of composition private study. Under that cover I had access to the electronic music studio which included an Arp 2600 music synthesizer. As some may know, synthesizers have various modules with specific functions such as oscillation, filtering and amplification. Control voltages can also create all manner of timbral modulation. One can always play a synthesizer with a keyboard but what really captivated my attention was designing a configuration of patch cords and knobs such that the synthesizer would “play itself.” We didn’t call it generative music, or generative art, back then but that’s what it was. I also began my interest in generative art theory at that time developing a system called “meta composition” that was medium agnostic and could describe arbitrary aesthetic systems.
Peter Bauman: Regarding your generative art theory, I read your work as broadly campaigning for inclusivity in generative practices. Your definition is wider than most people's conception of what generative is. Why is inclusivity important to you?
Philip Galanter: I can see why you might say that, but it’s not a campaign to include various otherwise disparate examples in the canon of generative art. It’s merely a side effect of the defining feature of generative art: the artist giving up a degree of control to an external system. At the time I first published this theory of generative art in 2003, the typical positions were things like, “generative art uses random numbers,” or “generative art uses computers.”
What I wanted to point out was the unique issues generative art raises and how non-generative art didn’t share these issues.
A byproduct of that kind of thinking is the inclusion of random numbers, computers, metal corrosion, reactive dyes and simple tiles as all generative because the artist cedes control to the system. In other forms of art we expect the artist to exert expert, detailed control of the materials.
Peter Bauman: I'm in the process of writing a detailed timeline of generative art history. As a theoretician, you've thought deeply about the history of generative art and its connections to your definition. What are three events you think are the most critical to include in this timeline?
Philip Galanter: The first is both the first known example of art and the first example of generative art. I describe it in the 2003 paper as follows:
"Christopher Henshilwood of the South African Museum in Cape Town uncovered the oldest known art artifacts. Etched in hand sized pieces of red ochre more than 70,000 years old is an unmistakable grid design made of triangular tiles that would be clearly recognizable as such to Escher or generations of Islamic artists. While simple, tiling patterns are determined in detail by a system that takes away from the artist's influence and intuitive designs."
The second would be a burst of activity in the '50s and '60s that reflect the general art world's swing to the conceptual. It would include writers like William Burroughs, artists like Sol LeWitt, Ellsworth Kelly and Carl Andre, and musicians like John Cage and many others.
The third would be the current emergence of AI. There is always the danger of over-evaluating the present and thinking the current state of the art is a major inflection point. But assuming a big change is so unlikely as to be beyond a reasonable choice is also dangerous. Peter Bauman: Why do you see AI as such a turning point? Philip Galanter: The current state of AI art that most people know about are the so called prompt artists. I remain skeptical of that genre. (It’s generative, but superficial.) The main reason I think AI art really is a turning point in generative art is that AI will soon provide something entirely new: an entity with much greater intelligence that humans. Making art with AI provides a safe sandbox where we can interact with AI without dangerous side effects. Along the way it will prepare artists to comment on AI itself in an informed way. So if we can survive international war, climate change and the worldwide move towards fascist dictatorships, I’m convinced AI will present yet another opportunity for humankind to kill itself. And unlike most threats, this is something we have no experience with. Humans rarely get things right on the first try.
Philip Galanter is an artist, theorist and curator. He is an Associate Professor at Texas A&M University. Galanter's areas of interest include generative art, physical computing, sound art and music, complexity science and art theory. His work includes the artistic exploration of complex systems and the development of art theory, bridging the gap between the cultures of science and the humanities. Peter Bauman (Monk Antony) is Le Random's Editor-in-Chief.