Algorithmic Art as a subset of Generative Art

For many years, I’ve classified my work as generative art. However, in recent conversations, people often respond with: “Oh, you mean like AI art”. This is understandable, and upon explaining the differences between generative art and AI art, we often reach a common understanding. Yet, it’s evident that the term “generative art” is increasingly associated with AI-assisted art, and rightly so. AI tools can generate images from simple text prompts, embodying the essence of “generative.”

After some thought, and opting for a practical point of view now that AI techniques have become more widely known, perhaps a more precise term for my type of work is “algorithmic art”.

I see algorithmic art as a subset of generative art. Generative art is a broad term that encompasses all art that is created in an automated fashion. But unlike AI-assisted generative art, algorithmic art involves deep integration with and deliberate control over the algorithms, a distinction crucial to most classic generative artists.

In the 1960s, pioneers like Vera Molnár and Frieder Nake began using code to create art, leveraging computers, oscilloscopes, and plotter machines to produce images impossible to draw by hand. Their work was grounded in rules and instructions, with computer programs generating visuals based on these parameters. This marked the birth of generative art, where the “generative” aspect referred to the computer program, not the artist’s hand. The artist designed the rules, thus acting as the designer, with randomness adding slight variations to each visual output.

The 1970s saw continued experimentation with artists like Harold Cohen, who developed AARON, a program capable of generating complete abstract paintings. Although AARON wasn’t using modern AI techniques, it relied on a vast set of rules defined by Cohen, producing countless variations of paintings. AARON was a standalone program, not connected to the internet or learning from datasets, but purely rule-based, creating new outcomes with each run. Some of these outcomes, or “editions”, are now esteemed pieces in museums such as the Stedelijk Museum Amsterdam and the Tate Gallery.

With the advent of Artificial Intelligence, especially AI image generators, the definition of generative art has become blurred. Isn’t AI also an algorithm that generates art?

AI art utilizes machine learning algorithms trained on extensive datasets of images. The training process results in an abstract representation of the input, encapsulated as numbers (“weights”) in a data model. Artists can use these models to generate new images by querying them with text prompts, without needing any programming skills.

That’s not to say that AI art is devoid of creativity. Artists can guide the AI model by providing clever prompts, steering the generated outcomes in a particular direction. Some artists, like Ganbrood, have their own unique process to create artworks that are original, otherwordly, and unmistakenly “Ganbrood”. The artist’s creativity lies in the prompts they provide and in the curation of the generated outcomes. Often, some manual post-processing is involved.

AI art is often classified as generative art because it is autonomously created. However, traditional generative art differs significantly in its deep integration with, and control over, the algorithms. Therefore, distinguishing between AI-assisted generative art and code-based generative art is crucial, as their principles and methods vary greatly.

Currently, there is a lively scene of generative artists who create art with code. As opposed to AI art, these artists write clever algorithmic systems that generate art. They don’t use AI models and they don’t use prompts. They find beauty in the code itself and in the outcomes that the code generates.

Take this series of artworks by Jacek Markusiewicz, titled “Barbarians”. Each artwork is 100% generated with code. The artist has created a single algorithm that generates these unique images, with each edition being a different output.

Considering all of the above, a fitting definition of algorithmic art could be:

“Algorithmic art is created by an autonomous system executing an algorithm, where the artist carefully designs the boundaries of its computational space and optionally defines the influence of randomness.”

Let’s break it down:

An autonomous system operates independently, e.g. without human intervention. In algorithmic art, this system is typically a computer program executing code autonomously. Although an algorithm could be executed by humans (e.g., Sol LeWitt’s Wall Drawings), the key point is that the algorithmic system drives the artwork.

An algorithm is a sequence of operations defined by the artist. In algorithmic art, these instructions are often implemented in a computer program that generates the artwork based on code.

In how far these algorithms are all designed by the artist or not can be a point of discussion. Some artists combine existing ones, while others write their own.

The algorithm that the artist has implemented is the core driver of the artwork. Often, an artist introduces randomness in the algorithm to create variation in different outputs. Randomness can be applied in various ways, such as randomizing colors, shapes, or positions of elements in the artwork.

The decisions of where to apply randomness, or to what extent randomness should influence the artwork, are made by the artist. The theoretical range of outcomes by the algorithm and applied randomness is collectively called the computational space. The boundaries of this space are carefully determined by the artist.

It’s not strictly necesssary to include randomness in algorithmic art, but it’s often used to create variation for different editions of the artwork.

This way, an interesting dynamic presents itself. An algorithmic artwork is both the result of the artist’s intentions and of the autonomous system’s execution. The artist defines the parametric space in which the autonomous system can work and the autonomous system adds unpredictable elements to the artwork.

This interplay between the algorithm’s strict rules and randomness creates a dynamic tension, where the artist balances predictability and chaos to produce compelling art.

The rise of Web3 platforms like Art Blocks, Fxhash, and Verse has brought renewed attention to algorithmic art. These platforms allow artists to mint algorithmic art as NFTs, enabling unique ownership of algorithmic art stored on the blockchain, along with its provenance. The stored artwork is essentially a snapshot of the algorithm used to create it.

While collectors are typically more interested in the algorithm’s outcome than the algorithm itself, there’s an argument that the algorithm is the true artwork, as it embodies the artist’s primary effort. The generated outcomes are autonomously created, beyond the artist’s direct control.

Algorithmic art’s blend of autonomous systems and artist input creates intriguing dynamics in the art market. On Web3 platforms, artists upload their programs and define the number of editions. Once uploaded, the process is out of their hands, and collectors purchase editions, often unaware of the exact outcome due to the program’s randomness.

Both artist and collector must trust the algorithm to produce engaging art, with an element of surprise for the collector. Rare outcomes, often intentionally included by the artist, can enhance the artwork’s value.

Some platforms allow collectors to curate outcomes before purchase, democratizing the process and giving collectors influence over the final editions. Market trends indicate that the algorithm’s outcome, rather than the algorithm itself, is generally regarded as the artwork. Certain editions are valued more highly, with a secondary market where collectors trade these unique pieces.

The meaning of generative art has evolved significantly with the advent of AI. While AI-assisted generative art has gained prominence, traditional algorithmic art remains distinct in its deep integration with and control over algorithms. Understanding these differences is crucial for appreciating the rich diversity within generative art. Algorithmic art, with its unique blend of artist intention and computational creativity, continues to captivate and evolve, offering new possibilities in the ever-expanding digital art landscape.