Instructions Beyond Code: Prompts on Generative Art

Author／Christopher Adams

"Generative art is like a huge factory where the machines started talking back. As the computers got faster, the art got better. Now it's like the art is making itself, and we're just here to watch." — A Large Language Model

If you're a student of art or a practicing artist, you've learned all about art movements and styles that develop over generations. Now we see new technology and trends emerge every day. Using computers to make art is nothing new, but now "generative art" is the hot new thing. So what is generative art today?

A "lecture-performance" is the art of talking about art. Artists who want to teach and communicate make it part of their practice. It's one way to deconstruct knowledge and critique institutions. This article is a new category that I call a "prompt-performance." It’s a collaboration and dialogue with artificial intelligence on the topic of computers and art.

What is generative art?

Generative art is an art form where the development process is governed by an autonomous system, such as a set of instructions, an algorithm, a computer program, or a machine learning model. The artist defines the rules that generate the art, but the specific output is determined by the system's programming, often by introducing elements of randomness, feedback, or interactivity.

Generative art uses computers to produce the art, but also to conceive the art. The computer is a tool to make shapes, but also to shape ideas. This is because the computer is the only tool that can function like the mind

How does generative art work?

Generative art uses a combination of design strategies for the development of form using code:

1. Repetition builds complex visual and spatial structures by reproducing and combining basic elements and modular pieces into shapes, series, and patterns. Floor tiles, wallpaper, the printing press, the copy machine, the silk screen, and the calculator work by repetition. The computer processor excels at it. Repetition can also extend through time with music and animation. Repetition is implemented in programming languages with loops and recursion, which can be used to build tessellated patterns and tree-like structures.

2. Transformation manipulates form to create something new. Geometric transformation moves, rotates, scales, reflects, and distorts points, lines, shapes, surfaces, and volumes. Numerical transformation converts, transcodes, and operates on pixels and other image data. Transformations happen in 2-D, in 3-D, in color space, in “data space,” and in time.

3. Parameterization introduces variables, constraints, and randomness. The artist chooses the inputs and the range of possible values, and the manner of manipulating them, or lets the computer decide. Adding randomness yields unpredictable results and allows for an infinite number of outputs

4. Visualization represents complex information as images and graphics. Data can be queried, filtered, reduced, and mapped into new forms. Values can be differentiated by size, brightness, texture, color, orientation, and shape. Maps, diagrams, graphs, and illustrations are all effective visualization techniques.

5. Simulation models natural or artificial systems, behaviors, or environments. Physical processes, material properties, fluid dynamics, agent-based models, genetic algorithms, and artificial intelligence can all be computed.

As we ascend these design strategies from repetition all the way to simulation, digital computers demonstrate ever greater advantages in terms of processing speed, parallelization, memory bandwidth, storage capacity, and computational power.

What does generative art look like?

Generative art can be plotted, printed, screened, projected, and fabricated. Plotter printers, cathode ray tubes (CRTs), LCD screens, LEDs, digital projectors, e-ink, laser printers, inkjet printers, and 3-D printers are productive output media for generative art.

Generative art can look like anything or like nothing at all.

What were the influences of generative art?

We can compress the history of generative art into a series of interrelated and successive movements over the past 100 years:

1. Cubism and Futurism: Formal precursors; the fragmentation of perspective and the rise of technology.

2. Dada and Surrealism: Ideological precursors; randomness, automatism, ready-mades, the unconscious mind.

3. Suprematism, Op Art, Color field, Minimalism: Structural precursors; geometric abstraction, optical effects, form as subject, repetition and variation.

4. Abstract Expressionism: Spontaneity, emotion, gesture.

5. Kinetic Art: Movement, interactivity.

6. Conceptual Art: Process over quality, ideas over aesthetics.

7. Serial Art: Multiples, sets, series, editions, prints, objects.

8. Analog and Digital Computer-Based Art: Science, technology, cybernetics, software, robotics.

9. Net Art and Demoscene: Connectivity, interactivity, community.

10. Augmented Reality and Virtual Reality: Real-time, 3D, immersive.

11. Artificial Intelligence: Machine learning, image generators, large language models.

12. Blockchain-Based Art: Crypto art and NFTs.

The design space for generative art is so large that it can include almost all art movements, media, and styles.

What are the critical issues in generative art?

There are a number of important debates and artistic decisions that can be considered in determining the formal qualities of a generative artwork or series of works.

1. Pixels vs Paint: This debate is about the authenticity of materials. Should generative art look digital, machine-rendered, and original? Or can it imitate physical mediums, tangible materials, and traditional art styles?

2. Screen vs Print: This concerns how to display generative art. Showing the art on screens highlights its dynamic attributes and software-native qualities. Conversely, printing and framing generative art can create a compelling aesthetic experience, valuing the generative process more than running live software.

3. Geometry vs Gesture: This is a matter of the style of generative art. Should it stick to precise, mathematical shapes and patterns, or include more human touches, such as brushstrokes or freehand lines? This determines whether the work feels more analytic or more expressive.

4. Long Form vs Short Form: This is a question about the volume of outputs of generative art. Should artists let the generative algorithm randomly produce hundreds or even thousands of variations? Or should they curate the outputs into a smaller, more selective series? This can influence the perceived value of each piece, and the quality of the entire collection.

5. Static vs Kinetic: Should generative art be still, like a painting or a photograph, having one form? Or should it be kinetic, rendered frame by frame, animated, evolving, or even reactive? This affects how we engage with and experience generative art.

6. Fixed vs Responsive: This is a choice about how flexible generative art should be. Should it have a set resolution and aspect ratio? Or can it scale and resize itself to adapt to different display sizes and orientations?

7. Software vs Hardware: This is a decision about how to embody generative art. Should it be made only with software, which is easier to share and access? Or should it include hardware and robotics, which can make the art feel more vital?

8. On-chain vs Off-chain: This is a consideration of how to store generative art NFTs. Should all the art's data and software be on the blockchain (on-chain), which is more permanent but costly? Or should some parts be stored off the blockchain (off-chain), which is more convenient but might raise questions about the art's integrity and longevity?

which is more convenient but might raise questions about the art's integrity and longevity?

Is there a difference between generative art and art made with AI?

While both generative art and AI art rely on computational processes, there are significant differences. Generative (algorithmic) art is the direct use of code and graphics with a high level of artistic control, and it usually operates on geometric primitives like lines, points, and polygons. In contrast, AI (synthetic media) art trains machine learning models on large sets of data using neural networks, and it usually operates on pixels or other binary data.

AI art challenges our values of originality and artistry. Artists who practice "appropriation" art consciously borrow and adapt existing work. AI practices a "misappropriation" art that some call learning and others call stealing. Since AI can mimic all artistic styles or mediums, it robs any aesthetic of its original sense. Any art that an AI can copy becomes meaningless. For a true artist to use AI, the idea is what matters most.

If I can't code, can I still make generative art?

Art has no boundaries. You don't even need a computer to make generative art. All you need is an idea. This is the key to conceptual art. Artists define a set of rules for the work, and execute their plan without regard for subjectivity or creativity. Just like a machine. Again, the idea is what matters.

What is your impression of the generative art movement in Taiwan?

These last few years have seen a burgeoning generative art scene that is filled with genuine excitement and camaraderie. It's a heady mix of artists, hackers, designers, collectors, and blockchain users. There are more and more museum exhibitions, gallery shows, workshops, art salons, and artist talks devoted to generative art.

Does that constitute a movement? One key point of reference is the independent art-space movement that rose to prominence in Taiwan in the past two decades. The earliest artist-run spaces were founded in the 1980s, but by the 2000s, a critical mass of artists, curators, academics, and critics brought these issues into focus: independent art production and curation, government control of cultural resources, disused urban spaces, and funding for the arts. They had vigorous debates, but they also put their ideas into practice with institutional experiments, by founding independent art centers and project spaces.

Today, Distributed Autonomous Organizations (DAOs) are emerging as institutional alternatives for art production. A DAO is a decentralized organization whose collective decision-making is mediated through blockchain technology. What are the possibilities of a DAO for the Arts? On the one hand, it could pool financial resources, support diverse artists, and democratize curation. On the other hand, it could prioritize art as an investment, promote popular artists, and minimize unique curatorial perspectives. A DAO for the Arts is not an answer but an opening.

What are the critical issues for generative art today?

In contemporary art, particularly in terms of generative art, AI, blockchain-based art, and DAOs, several critical issues arise:

1. Art Production and Curation: Software and the Internet are changing how art is made and distributed. Generative art, AI, and blockchain technology demand new knowledge and skills from curators and exhibition spaces. DAOs could challenge centralized power structures and gate-keeping in the art world.

2. Authorship and Originality: Generative and AI art raise questions about who creates the art: the human or the machine? This challenges our ideas about creativity and self-expression. There is also a controversy about whether AI-generated art can be copyrighted.

3. Data Ethics: AI art involves ethical issues about the sources of its training data, which artists cannot always control. There are concerns about copyright, moral rights, plagiarism, and infringement.

4. Economic Models and Art Markets: AI art is disrupting the commercial art industries. Blockchain and NFTs are changing how art is published and sold. DAOs are also introducing new ways to fund and invest in art.

5. Access and Inclusivity: Digital art can be more accessible, but it might leave out people who don't have access to technology, who face disadvantages, or who have disabilities. This digital divide can create a barrier in appreciating and participating in these new art forms.

6. Art Conservation and Permanence: Preserving digital art over the long term can be difficult, especially with so many dependencies on specific technologies, formats, and platforms. This poses a significant challenge for museums, conservators, and art collectors.

What are your sources? Where can I read more?

This article was written and edited with the assistance of a large language model (LLM). The following references provided numerous inputs.

Figure 1
After the Cave（2023）

After the Cave (2023) by Aluan Wang (王新仁). This "short-form" generative art collection includes 128 works that extend the parallel lines of Indigenous and Modern Art. In 1937, the Museum of Modern Art (MoMA) in New York held three simultaneous exhibitions: one for works by Cézanne; Twelve Modern Paintings by Paul Klee, Joan Miro, Jean Arp, and others; and Prehistoric Rock Pictures in Europe and Africa, featuring life-size facsimiles of cave art. This juxtaposition inspired a cheeky newspaper headline: "First Surrealists Were Cavemen". We can now say: the first generative artists, like the first philosophers, also emerged from a cave.

Figure 2
Walk in Progress #0 (2023)

Walk in Progress #0 (2023) by Liu Nai-Ting (劉乃廷). The artist's previous collections depict life-forms like octopuses and coral emerging organically frame by frame with natural, random variations. In contrast, this work is a singular study of a city landscape rendered in exacting detail, line by line and layer by layer. These four images show the stages of the picture's progression from line drawing, to color draft, to finished scene, to its final "deconstruction." The artwork can be displayed as live software, as a rendered animation, and as physical prints. This work is an homage to Festival on South Street (南街殷賑) (1930) by Taiwanese gouache painter Kuo Hsueh-Hu (郭雪湖).

Figure 3
INACTION （2023）

INACTION (2023) by Natalie.J. This "long-form" generative art collection is limited to 128 outputs. Color fields are assembled from vector fields that resemble both hard landscapes eroded over millennia and soft folds of fabric. How can we read "the art of inaction"? Like stone waiting to be carved.The expressiveness of the entire work is embedded in the code. The final set of images is produced autonomously beyond the artist's control.

Figure 4
The Ancestors （2021）

The Ancestors (2021) by Lai Tsung-Yun (賴宗昀). The artist trained a machine learning model on photographs of Taiwanese indigenous peoples. Their face tattoos were banned and even forcibly removed during the island's colonial past. We can read these works as addressing a historic injustice, but also as perpetrating a new disfigurement using artificial intelligence.

Figure 5
In Love With The World（2023）

In Love With The World (2023) by Hemilylan. Related to artist's previous series PFP of U (2022), which was inspired by web3's profile picture culture, these abstract portraits were executed with acrylic paint before being digitized. The works are categorized using metadata attributes such as accessories, background, and body type, similar to popular NFT collections. They also highlight the prevalence of avatars and the value of anonymity in blockchain communities.

Figure 6
Tree Ring #24 Semi-Finished（2022）

Tree Ring #24 Semi-Finished (2022) by Huang Mu-Yun (黃沐芸). This series of digitally altered photographs examines how artificial intelligence is used to write prompts about women and to generate images of the female body. It dissects the biases that go into and the distorted imagery that comes out of machine learning models. It is a study of self-identity and depictions of the self, as well as body image issues and social media manipulation.

Figure 7
a contemporary art center, taipei (a proposal)（2008）

a contemporary art center, taipei (a proposal) (2008) by Jun Yang. The artist occupied a temporary pavilion at the Taipei Biennial to host an experiment for an independent art space. This project was the genesis of Taipei Contemporary Art Center, founded in 2010. Other art spaces established around that time include Open Contemporary Art Center (OCAC) and TheCube Project Space. Today, we can ask what a Decentralized Autonomous Organization (DAO) for the Arts is, should, could, and would be.