A style that uses a self-governed autonomous system to compile compositions. Generative art is a broad category, created entirely by chance or through the mind’s eye. It’s an exploration of a transparent, participatory collective communication process. By leveraging the power of computing technology, artists can unlock the doors of creativity, lowering the time between intention and realization.
Table of Contents
What is generative art?
Generative systems help artists explore much broader ground faster than ever before, letting artists create thousands of ideas in milliseconds. The help of more powerful machinery and easier access to modern processing power is inventing new aesthetics regularly. This profound application of technology allows for the exploration of visual and auditory experiences, making art accessible even to the untrained.
Run within a set of artistic constraints; generative art fuels new on-chain platforms and pushes contemporary mediums’ boundaries. It is a form that has many techniques and approaches that can range from layering and compiling or being entirely created by code. Even artificial intelligence left to its own devices can make artworks. With generative art, language rules, machines, algorithms, or genetic sequences produce ideas, forms, shapes, or colors coupled with symmetry, pattern, and tiling systems. The systems can be digital, chemical, or manual and applied to various outputs.
Initially rejected by the cultural establishment marked as the domain of computer scientists and mathematicians. Generative art can be linked to increased access to technology, as computers and network connections have become common. Generative art is the exploration of systems rather than the production of content. Art that genuinely reflects our time.
Briefly looking at history
To understand generative art, we must take a trip back in time. You might find it shocking to know how long these ideas and technologies have been part of our existence. Early influences started to emerge at the end of the 19th century. Artists such as Paul Cézanne laid the foundation for Cubist principles. He introduced ways to play with geometry and conflicting vantage points. Fascinations with technology and machines arose with Futurism and Constructivism, which would become central components of generative art.
Early advancements in the 20th century ushered in Dadaism that used randomness to reject established policies and practices. The combined efforts of each movement’s initial ideas surfaced and challenged the very definition of art. In 1933, George David Birkhoff attempted to redefine aesthetics as a mathematical ratio between order and complexity. He stated that art was purely for communication purposes, and the result was trivial.
After the advent of the mainframe, Vera Molnar conducted experiments with systematically determined abstract geometrical paintings. Building her foundation in the arts, she set forth on the hunt for any computational power she could exploit. Once found, Molnar furthered her quest in generative art by establishing the research group art et Informatique. She was also learning Fortran and Basic, which Molnar applied to some of the first computer graphic drawings created by a plotter.
The earliest works of generative art were drawings with a strong emphasis on rhythm and repetition. Michael Doll was the first person to program a computer purely for art. One of the challenges faced by early artists was the limitation of output devices. The primary source at the time was the plotter. This limitation led to early art focusing on geometric forms and structures instead of more fluid content.
Practitioners prized the pure visual form above content as they considered the computer an autonomous machine. Less than two decades later, computers began to reach beyond military research and become a platform for artistic experimentation. During the late 1950s, Bell Labs and the RAND Corporation experimented with machines’ ability to create art and music. During this time, the US government also began research in artificial intelligence.
Since the government first developed computers to execute the complex calculations that were part of modern warfare, it was impossible to uncouple computer technologies from militarized control. The government and corporations researched ballistic trajectories, cryptography, and theoretical modeling. When programs failed to meet expectations, “AI winters” occurred and labeled research under “search algorithms” and “machine learning.”
Corporations, the military, and universities made computers possible. The immense machines required deep pockets. They were bolstering incentives to foster an extensive research program, recognizing the possibility that innovative results would ensure a competitive edge where computers were beginning to take hold.
The art, exhibitions, actions, and texts from the early days of generative art were meant to integrate computers into artmaking, reimagine artists’ artworks, and reimagine the political relationships of artists and their artworks. In other words, the governments tied it to art’s political role.
The military was so entangled in generative art that in 1963 during the trade journal Computers and Automation’s first “Computer Art Contest,” the jury awarded first prize to Splatter Diagram, a computer-generated fractal graphic created by the US Army Ballistic Research Laboratories.
Early pioneers of generative art
Several notable names come to mind when discussing generative art. Max Bense sought to quantify aesthetic experience and computerize creating and evaluating art. Gustav Metzger wanted to dismantle the myth that technology was rational or neutral. He tried to expose the destructiveness of consumerism at its very core. Even though they saw things differently, they both envisioned art as a way for computers to defy their warlike origins.
Both men imagined generative art as a means of catalyzing a more engaged community and a more transparent, participatory, and democratic world. Bense began writing about aesthetics in the 1950s, coming up with a way to calculate and quantify art. He envisioned it as a form of a communications network. Visuals are exchanged between audiences. He defined aesthetics as the unexpected element and that it unpredictably transmits information.
His theory looks at the relationship between possible variations and states that generative art’s aesthetic can be measured and predictably produced. The short essay “The Projects of Generative Aesthetics” covered schemas for evaluating these states and formulas for stimulating them. Coincidentally Bense also wrote the paper to accompany an exhibition of computer-generated graphics at the University of Stüttgart gallery.
In 1964 Bense began experimenting with “statistical graphics” using a programmable drawing machine known as the Zuse Graphomat Z64. With this machine and a simple algorithm, he told it to distribute eight dots inside a square and connect them with a closed line. The results were geometric line drawings highlighting the artificial production of probabilities, differing from normalcy using rules and programs.
Bense’s theory of generative aesthetics removed subjectivity from art. His ideas were that aesthetic judgment and inspiration were transparent and explicit forms of science that would transform the metaphysical discipline into a technological one.
Evolution of generative art
Generative art continued to evolve into the 1970s when the School of the Art Institute of Chicago created a department entitled Generative Systems, focusing on art practices using new technologies. The 1990s were a time when many things changed. Still a child, I was old enough to remember the onset of the internet and lived through the development of technologies. My exposure and experimentation with computers began in 1995. Back when Windows was good, and Roller Coaster Tycoon was cool. During this time, technologies such as Macromedia Flash started paving the way for generative art.
You probably didn’t know that you could use Flash as a generative art tool, but you could. Thanks to a little bit of code. ActionScript was a code that could apply directly onto graphics or frames. The language included basic actions like scaling, rotating, opacity changes and morphing one shape into another. These actions were programmatically applied when the play head hit a particular frame and triggered the code. An artist could loop these commands and repeatedly draw shapes onto the screen as you scaled and rotated them.
As a result, a new breed of artist-developer was born. Most notably, Joshua Davis, an art student at the Pratt Institute, inspired by creative coding. Technical limitations prevented Flash from editing bitmaps or pixels; artists had to work with the vector graphics and limited palettes of on-screen elements. Therefore a minimalist look was the style of generative art then.
Design by Numbers changed everything. While at the MIT Media Lab, John Maeda created and began developing an open-source platform to learn transferable programming skills to create art. Maeda recruited a couple of students named Ben Fry and Casey Reas to help work on it. They built a robust programming language that they would later call Processing.
Processing could take the properties of thousands of graphical objects. It was fast, easy to write, and easy to understand, packed with sample projects that outlined basic computer science principles such as loops, functions, and arrays. The idea was to help artists and designers learn to program. Early artworks saw artists exploring concepts based on patterns in nature.
Both Flash and Processing provided artists with new vehicles for creating art, and in the early 2000s, generative art took a twist as publicly available datasets became available.
A common thread across these tools is their element of randomization. While the artist manages and decides what goes into the artwork’s final form, the output of the automated system remains unpredictable. A simple analogy will be that of a musician: the musical scales are standardized, but the player is free to make nuanced variations when playing the piece. It is this dance between precision and uncertainty that makes generative art.
Ian Goodfellow wrote an essay on a new idea known as generative adversarial networks or GANs. It explained a system of two neural networks—a discriminator and a generator. The generator consumes a set of training data and trains itself to produce something that looks like what it taught itself. The idea is that the discriminator cannot tell another network created it.
Similarly, Procedural Modeling rapidly generates ideas and teaches a machine the parameters of what each object should look like—letting it do that hard work of creating lots of assets. While algorithmic art, code art, or procedural art is a step-by-step method of solving a problem, it allows artists to push the limits of expression.
Until recently, these techniques required deep technical knowledge and programming skills. Now it’s as simple as a Python notebook and Google cloud. Current applications span literature, music, products, architecture, and visual art.
Literature is produced by an author creating databases of words and randomly generating complete works. Music artists can set constraints on systems and feed them style examples to create unique compositions. Products and Architecture use computer-aided modulations to develop varying solutions quickly and efficiently. Visual artists create works from organic to artificial to chaotic or controlled via autonomous systems.
The world of generative art has been influenced by brilliant and talented individuals willing to stick their necks out and explore. There are by far too many artworks to list; however, here is a list of some of the more notable projects that have helped shape the artform’s history and shine a light on what is possible today and moving forward.
Unique generative art projects:
- Subdivided Columns Michael Hansmeyer explores how subdivisions can define and embellish this column order with an elaborate system of ornamentation.
- 8-corner, 1964, Max Bense illustrated the creativity of algorithms and the rational basis for aesthetically balanced forms.
- Computergrafik Computerplastik, 1970, Georg Ness was variations of a record for a purpose by the computer, selecting ideal variants.
- Structure de Quadrilatères, 1988, Vera Molnar
- Intersection Aggregate, 2004, Jared Tarbell is a fun visualization defining the relationships between objects.
- All Streets, 2007, Ben Fry plots all of the streets in the lower 48 United States: an image of 26 million individual road segments.
- Memories of Passersby, 2018, Mario Klingemann uses a complex system of neural networks to generate a never-ending stream of portraits created by a machine.
- RGB, 2020, Casey Reas created three drawings rendered with a plotter using Koh-I-Noor Rapidoplot.
- Stitched & Smeared Even Against . . . XO vol. 2, 2021, Andrew Benson exhibits looped brushstrokes, pixel smears, improvisational gestures, distortion map, feedback.
Mentionable generative artists
Much like the works mentioned before, there is an extensive list of highly talented individuals out there experimenting with generative art and learning as we all go. While not even a fractional representation of that list, here are some notable artists to look up to and enjoy.
- Max Bense
- German philosopher, writer, and publicist, known for his philosophy of science, logic, aesthetics, and semiotics.
- Georg Ness
- German academic who was a pioneer of computer art and generative graphics. He studied mathematics, physics, and philosophy.
- Vera Molnar
- French media artists are widely considered pioneers of computer art and generative art. The first woman to use computers in her art practice.
- Frieder Nake
- Mathematician, computer scientist, and pioneer of computer art. Best known for his contributions to the earliest manifestations of computer art.
- Michael Noll
- An American engineer, professor emeritus, and a very early pioneer in digital computer art and 3D animation, and tactile communication.
- Lillian Schwartz
- American artist considered a pioneer of computer-mediated art and one of the first artists notable for basing almost her entire oeuvre on computational media.
- Grace Hertlein
- Early computer artists focussed on applied maths and stochastic functions, creating works primarily involving formal structures and geometric abstraction.
- Muriel Cooper
- A pioneering book designer, digital designer, researcher, and educator. She was the first design director of the MIT Press.
- Sol Lewitt
- An American artist linked to various movements, including conceptual art and minimalism.
- Jared Tarbell
- Most interested in visualizing numerical processes and the life-like properties of complex, computational systems.
- Ben Fry
- Focuses on combining fields such as computer science, statistics, graphic design, and data visualization to understand information.
- Casey Reas
- American artist whose conceptual, procedural, and minimal artworks explore ideas through the contemporary lens of software.
- Michael Hansmeyer
- A post-modern architect utilizes algorithmic architecture techniques, generative art mentalities, and CAD software to generate complex structures.
- John Maeda
- His work explores the area where business, design, and technology merge to make space for the “humanist technologist.”
- Mario Klingemann
- German artist and Google Arts and Culture resident are known for involving neural networks, code, and algorithms considered a pioneer in using computer learning in the arts.
- Andrew Benson
- His experimental works are unmistakably digital, working primarily in digital video and animation, incorporating software processes and electronic devices.
- Anders Hoff
- A generative artist fascinated by patterns using the highly organized structure and then finds ways to disrupt it.
- Mark J. Stock
- Generative artist, scientist, and programmer combining elements of nature and computation, exploring the tension between the natural world and its simulation.
- Scott Draves
- Video artist and accomplished VJ, inventor of Fractal Flames, and the distributed computing project Electric Sheep. They invented patch-based texture synthesis and published the first implementation of this class of algorithms.
- Daniel Shiffman
- Is a computer programmer, a member of the Board of Directors of the Processing Foundation, and an Associate Arts Professor.
- Lauren McCarthy
- Artist and computer programmer creating artworks that use various media and techniques, including performance, artificial intelligence, and programmed computer-based interaction.
- Joshua Davis
- American designer, technologist, author, and artist in new media. He is best known as the creator of praystation.com, winner of the Prix Ars Electronica 2001 Golden Nica for “Net Vision / Net Excellence”.
- Robbie Barrat
- An artist and graphic designer working with artificial intelligence as a tool and a medium.
Generative art tools and resources
Tools like Artbreeder and Runway ML make GANs and other machine learning models more accessible. Artbreeder has already produced more than 54 million images and radically simplified making Art with GANs. Users of the tool click on images to “breed” them, dragging sliders back and forth to increase or decrease the amount of influence.
Runway ML is more advanced and designed to accelerate the movement of new algorithms and models from research to software. It’s easy to get started quickly and allows for the fine-tuning of multiple machine learning models to colorize black-and-white photos, transfer style from one image to another, recognize faces, turn doodles into photorealistic photos, amongst others.
“Dreamcatcher” allows designers to input design objectives, including functional requirements, material, manufacturing method, performance criteria, and cost restrictions. Its system then evaluates and presents many designs satisfying the requirements.
People will gain new skills and capabilities as assistive creation systems become more accessible. They are empowering novice artists and lowering entry barriers. Tools like HumOn and Humtap let non-musicians hum into their smartphones and quickly build a song around it.
Open-source software allows anyone to expand their creative toolkit for free. Open source datasets provide computational artists with the raw material to create new works or tools. Generative creation systems shift artistic focus to creating a process rather than a result. By leveraging computers to do the legwork, they can create near infinite variations in a given solution space.
Here’s a list of tools you can use to create generative art.
- Open Images Dataset
- Lucid Sonic Dreams
Generative art books
Surprisingly not a lot of books exist on the market for generative art. There are, however, varying amounts of books and online resources about the different languages, approaches, use cases, and styles of generative art. Here are some of the popular ones that I found.
Generative art books:
- Generative Art: A Practical Guide Using Processing
- The Nature of Code: Simulating Natural Systems with Processing
- Computers and Creativity 2012th Edition
- Fractals: Form, Chance, and Dimension
- Processing: A Programming Handbook for Visual Designers and Artists
- Art Forms in Nature: The Prints of Ernst Haeckel
Pushback and criticism
Much like anything else new and disruptive, it’s never short of pushbacks and criticism as attractive, groundbreaking, and innovative as generative art was; others were not thinking that same thing. Many said it was algorithmic overkill. Over the years, its perception that artificial intelligence takes over as the artist lets go of physical control by allowing a machine to complete it. The finished product is far from art.
Today artists find new ways to differentiate themselves in this seemingly infinite and repetitive imagery produced. The best way to do so is by reintroducing elements of analog art production. Sougwen Chung, Anna Ridler, and Helena Sarin are three great examples of artists who have done so. Sarin trains GANs on her drawings and paintings and uses those visuals generated by AI as the basis for works made with various analog processes.
Chung trained a neural network on her drawings and employed analog art processes to make her final work. She also programmed robots to draw alongside her on large canvases on the floor in a series of live performances. It’s straightforward to find repositories and use existing code. That’s how we learn. People want to see the creative exploration of the technology and its application. Therefore a generative art today must transcend realms.
Procedural modeling reduces the cost of generating images, video, audio, and other assets required for multimedia leaving more time for defining a range of possible outcomes, deepening our understanding of what makes something beautiful. Artificial intelligence will learn from us and become collaborators, suggesting ideas in real-time, helping to maintain a flow state.
General adversarial networks allow machines to learn about a dataset and generate incredibly realistic variations in a similar style. Creative Adversarial Networks allow devices to inject novelty into their outputs and even remember what novelty is helpful over time. Evolutionary algorithms make it possible to encode the fundamental processes of nature into the creation of art.
Educational platforms make it easier to learn new skills. The future is exciting, combining analog and digital, human and machine, instead of elevating one over the other. It’s a world that consists of generative art, blockchain technology, cryptocurrencies, and NTFs.