It is challenging to navigate the ever-increasing complexity of our time. Dave Snowden has built a complete [framework](https://www.wikiwand.com/en/Cynefin_framework) in order to analyze and understand such systems around and inside us. It is especially useful in times of recurring challenges, like pandemics, environmental emergency situations, supply chain crisis and social collapse (war). Reading generative art resembles many aspects of the tactics of understanding dynamic complex systems. On the *perceptual level*, generative structures can appear as sophisticated, complex visual systems that can seem both familiar and alien-like the same time. On a *taxonomical level* (structure, underlying mechanisms), these systems can be understood at their core behavioural mechanisms, on the level of functions, instructions and algorithmic strategies. Then there is another layer for interpretation, which is a much wider *contextual level*, where history of the cultural scene, set and settings of the artwork and the artist, social and political integration plays role. This post covers some of these attributes with a focus on an epistemological approach.

**access.**

On the perceptual level, new forms of media, unconventional platforms, never before seen formats, different expressions and corresponding trends arise. As we see the transformation of the human attention span, combined with new ways of selection mechanisms and recommendation systems, one interesting aspect of perceptual observation would be time itself. While temporal aspects of an artwork can work on almost non-human extremities (think of some of the pieces of contemporary [music composers](https://universes.art/en/specials/john-cage-organ-project-halberstadt), emergent, long-form generative and conceptual art), we can see an emerging trend of the compression and decrease of focused attention recently, perhaps as a survival strategy against our oversaturated environment. Few decades ago, we spent many hours observing a single image, so the intimate relationship with those artefacts were developed on a long term, gradually unfolding, subjective process. As responsive media became widespread and more accessible through personalisation, [micro-interactions](https://www.nngroup.com/articles/microinteractions/) and similar agencies in the digital domain, we are in a transition towards consuming much more material with less intimate contact that lacks repetitiveness, gradually built confidence and subjective involvement. We founded our cultural backgrounds step by step to understand paintings, even photographs and drawings, through constant omnidirectional discourse along the cultural fields and developments. Now, while shifting from physical-only objects to hybrid, digitally signed networked media, we have to integrate systems thinking, dynamic properties, conditional structures, even [game theory](https://www.wikiwand.com/en/Game_theory) and evidence based speculation strategies into our cognitive resources. This leads to the second level of reading generative artworks.

::tezos-storage-pointer[]{contract="KT1KEa8z6vWXDJrVqtMrAeDVzsvxat3kHaCE" path="token_metadata::547993"}

On the *taxonomical level*, a generative piece is composed of different automated, functional and outsourced methodologies. Computer science, math, physics, gestalt, social engineering, all became part of the arsenal that takes part in the creation of a piece. Generative art usually incorporates algorithms from its simplest forms (adding two numbers) to more complex (using pre-trained machine learning models run by several machines independently) combinations of formalisable methods. An algorithm originally meant to describe a solution to a problem, so when we see colours, forms and shapes of a piece, we can also see the intention, variability, symbolic possibility field behind their actual configuration, depending on the level of our algorithmic literacy. The images are built up with textual code, so the original formulations of the exact steps can be traced back to discrete elements in the system. The accessibility and transparency of these elements define explicitly our experience of the observation and affordance of these projects. The nature of code (open / closed source) it has been built with, the quality and structure of the underlying datasets (public / privately controlled) are active components of the whole, extended body of a particular piece. Today, transparency, traceability becomes more and more problematic, as we see the wider acceptance and smart monetisation of closed, black-box like AI tools, where nor the artist, nor the audience have access to the data and algorithms that the system was trained on and the [intellectual properties are questionable](https://www.engadget.com/dall-e-generative-ai-tracking-data-privacy-160034656.html) to say the least. However, in general, an algorithmically literate person can read a generative artwork closer to its full completeness than one without such skills.

> *”What people call “AI” is actually a long historical process of crystallizing collective behavior, personal data, and individual labor into privatized algorithms that are used for the automation of complex tasks: from driving to translation, from object recognition to music composition. Just as much as the machines of the industrial age grew out of experimentation, know-how, and the labor of skilled workers, engineers, and craftsmen, the statistical models of AI grow out of the data produced by collective intelligence. Which is to say that AI emerges as an enormous imitation engine of collective intelligence.” -*[*source*](https://www.e-flux.com/journal/101/273221/three-thousand-years-of-algorithmic-rituals-the-emergence-of-ai-from-the-computation-of-space/)

**practice.**

The methods used for generative arts are a mixture of individual decisions, personal experience and community based solutions. These are ideas, thoughts, philosophies, belief systems, moral innovations, source code, video tutorials, microblog feedback mechanisms, reusable software libraries, math equation implementations, openly usable authoring tools with very different software architectures and corresponding community structures. The realtime feedback mechanism affects the trends and the behaviour of generative art ecosystems, they are constantly adapting and evolving, it is encoded in their dna.

With generative art we can see a separation of planning and evaluation, instead of manipulating the substance directly, the artist forms some ideation, then build a framework around it, and let the evaluation unfold based on the parameters of the framework. In this sense, we can identify *choreography, conceptual art, instruction based art, programming, music composition, game design*, etc. as loosely coupled parts of this field.

![Generative Art Methods by Effective Complexity. Philip Galanter, 2019](ipfs://QmcKDCQrxogDqqbw7LkM1AG2MBpwaoEXisxtpLoWj7iaDS)

We can choose among many features that we find meaningful while thinking about the types, clusters and nature of algorithms that are being used for creating creative output. Identifying features, information extraction and similar tasks became really important while facing complex scenarios that are larger than what we can process from their completeness. Lets take an approach as a feature to identify the agencies of algorithms. We can point out two ways of constructing applications with them: the top down and bottom up approach. In this sense, top down algos are complex artificial intelligence, like neural networks and pre-trained systems, where the artists are using black boxes, tinkering with them, reverse engineering and trying to modify the outcome to fit their needs. Bottom up, or grass roots structures are the ones, where the artists start with some simple set of rules and let them grow, unfold and mature, waiting for their emergent properties to arise. Geometry, certain fields of mathematics, combinatorics, systems theory and conditional simulations are examples for these approaches. Or we can take another feature dimension: the entropy component, the transition from a chaotic to an ordered state, or vice versa. If we think of algorithms as formulated solutions that are reordering their environments over time, we can have *constructive* algorithms, such as world building simulations (cellular automata, l-systems etc) and we can have *deconstructive* algorithms where a system converges towards chaotic, erosive processes (such as subdivisions, noise systems, randomness).

These are just some examples of how we can access, recognize and integrate generative algorithms and computational thinking into our practice, but of course there might be many more interesting aspects that would help building a [vocabulary](https://github.com/jasonwebb/morphogenesis-resources) for a widely adopted, commonly used algorithmic literature.

**recognize.**

On the wider, *contextual level*, a generative artwork can mean many things. From proposing [alternative solutions](https://aljazaribook.com/en/) to existing social challenges, through [combining fine arts and the computers](https://piratefsh.github.io/2019/01/07/computer-art-history-part-2.html), the demo scene of the 80s, the glitch and abstract [constructivist environments](http://dextro.org/) of the 90s, [net art](https://jodi.org/) and computational pieces usually share a common aspect: most of them are influenced by counter culture and critical thinking — they are all questioning the status quo from time to time. In the early years these questions targeted the medium itself. Computational art was about to use computer for artistic practice instead of calculating only, making abstract visuals of debug views and wireframes, creating sound sculptures of glitch, errors and limits of the environment. The glitch scene and net art continued this tradition by exploring errors, limits and unintended interruptions of systems and new media. Today, many digital art projects question current distribution models, dead-end economical models, vendor lock-ins and [walled gardens](https://www.techtarget.com/searchsecurity/definition/walled-garden), with strange combinations of traditional information technology and more recently, decentralized ledger networks (blockchains).

> *“ The choices we are faced with today are especially important because digital technology so dramatically increases the ‘space of the possible’ that it includes the potential for machines that possess knowledge and will eventually want to make choices of their own.” -*[*source*](https://worldaftercapital.org/)

![Collective decision making structures](ipfs://QmcigKj4t9mXGmLRBthTnGoJSEdHPr8euWu4M2GvVyiRCV)

An interesting way of thinking about the position of a specific culture is to think of its wider context and history. In a recent [book](https://thenetworkstate.com/), Balaji Srinivasan shows that there are basically two ways of writing history. One of them is a top-down approach, while the other one is a bottom-up, self governing approach. While most traditional historical events are written “top-down” by the winners (who have the power to write and execute) throughout successful events in a particular field, the latter is building up context from an atomic level, by letting each and every change be logged in the system in an immutable way, so the participants don’t need social consensus to determine which event was true and which was false. Writing and exchanging generative art in networked communities has many aspects of this second approach. Since the creators are using self logging computational tools for their practice, they are constantly building up the micro-history, reputation, social relevance and accountability of their culture in an automated way, where each of their actions (thus, transactions) leave traces in a ledger structure. These logs act as references for community members, art historians, curators, entrepreneurs, researchers and for the artists themselves as tracing method for personal progress. In the meantime, the same logs also act as immutable building blocks that define consensus mechanisms within a social guild, they are transformative tools of cryptographically valid, self operating layers that are built on existing infrastructure, which can be integrated into our everyday life within reasonable energy and time.


This piece is an extended, updated version of an article, originally published on 'Becoming Human: Artificial Intelligence Magazine'. The text aims to understand generative intentions and methodologies within the visual arts from an epistemological perspective, by introducing early examples of artworks, algorithmic thinking and cognitive aspects of complex systems and decision making mechanisms. 

Generative Art Systems - an epistemological approach

Written by

Editions

Share

Details

stay ahead with our newsletter