In artistic practices working with algorithmic processes, levels of automation and procedurality impact the process both methodologically and conceptually. Giving over degrees of agency to machinic processes in artistic production raises questions as to what impact this may have on the relation between producer and the mechanisms and processes used to articulate the work. In this sense, automation can be seen as not only a means to an end (Parisi) but impactful upon the process from the stage of ideation. The procedural logic of algorithmic media is influential to the modalities of generation and decision-making during the process of artistic production. The following paper explores how various degrees of automation and procedurality are expressed through a selection of pertinent artistic methodologies.

The mechanisation of image production in the last century, the subsequent integration of movement, electronic signal, and the current turn toward algorithmic media reveal a tendency turning away from images as primarily optical media, in favour of automation and procedural logic. Human biological perception, thus, is no longer the primary audience for images, as we recognise that an ever-increasing amount of visual information is created and processed without passing through a human viewer. As a result, artistic practices have gravitated toward the production of images which prioritise the execution of sets of instructions. The term automation, as it is used in this research, is understood to be bounded, acknowledging that systems may function autonomously to a degree, but the aspect of autonomy is limited by the internal parameters of a system by virtue of its design. That is, automated forms of imaging are subject to the boundaries embedded into a given system by its designer. More highly automated forms of media, such as those employing machine learning, may produce unexpected outcomes, but it is the subject of debate as to whether or not an automated system may be considered to create autonomously. In this paper, these qualities are examined through paradigmatic examples in order to outline several distinct tendencies of how notions and methods of automation and procedurality have been integrated into artistic production and digital aesthetics.

Automation

“Machine-supported art also radically implicates the independence of machines from man.” (Weibel 17)

Automation has been one of the driving principles of the past century, influencing all aspects of life, including artistic modes of production. Peter Weibel outlines five stages of automation, summarised as follows:

First stage: machine-supported generation of images (i.e. photography)

Second stage: machine-supported transmission of images (i.e. printing)

Third stage: machine-moved images (i.e. cinema)

Fourth stage: image as electronic signal (i.e. television)

Fifth stage: machine-generated, calculable image of the computer (i.e. computer graphics) (17-18)

He conceives of the last of these stages as satisfying each of the previous four categories, with the addition of being autonomously produced by machines, separately from the intentionality of the artist. Combining this model with Manovich’s low-level and high-level automated media (Manovich 32-33), Weibel’s five stages progress upward from low-level to high-level automation. Partially-automated methods tend to be low-level, involving a pre-defined system or limited aspects of a work which are orchestrated outside the control of the artist. Placed here would be works which entail a restricted degree of autonomy, and in which the outcome is largely based on an initial data-set or system. Lower-level automated systems tend to be more predictable, limiting the agency of the machine or process being used. These are differentiated from higher-level automated media, which involve dynamic behaviour, and would potentially have the ability to affect outcomes based on the semantic content of the initial inputs. Such works, are generated on the fly or during run-time (Manovich 32) and may be responsive to user behaviour. At the hypothetical pinnacle of automated media would be works which satisfy one of the aims of a-art (artificial art) (Massey 11), which is to achieve an outcome which could not have been foreseen by the creator of the work. Practices involving higher levels of automation and procedurally would satisfy this to a degree, as in spite of being highly deterministic in some aspects, algorithmic systems may also be capable of producing surprising, unpredictable results, as  has been documented by many computer scientists and researchers working with machine learning and artificial intelligence (Lehman et al.).

Strategies of automation have been used in various avant-garde movements since the beginning of the twentieth century, from Dada, Surrealism, Conceptualism, early computer art, to contemporary practices. The Surrealist notion of automatism was particularly influential, advocating that artists relinquish conscious control over the artistic process, so as to arrive at art produced by the subconscious mind. Automatic writing, drawing, and painting thus involved artists developing methods to elude their own consciousness, often by employing highly systematised, rule-based techniques to surrender control by engaging with serendipity and randomness. In many instances, the artist expressly sought to hand over agency, intentionality, or control to a process, machine or system. One of the most famous and influential methods to materialise from this kind of aleatory (Carvalhais 146) approaches is the “cut-up technique”, a process in which a linear text would be dismembered at random and rearranged by the artist, influencing the creation of a new work through its starting conditions. This would be considered a low-level form of automation, yet due to the use of randomness, these processes are highly unpredictable. Unpredictability is another measure which will be discussed further in the following sections.

Practices employing chance, randomness and a lack of intentionality are nonetheless highly structured, employing sets of rules and parameters which frame the work. Conceiving of the artistic practice in such terms has parallels in cybernetic theory, viewing machinic and biological processes on a flat hierarchy, in terms of their systematic behaviour. The next section covers procedurality and the use of structured sets of rules in the production of art.

Procedure

With the turn toward algorithmic media, image production increasingly entails the execution of sets of procedures, integrating a procedural logic into the process. Procedurality is not medium-specific, nor tied particularly to the implements nor methodologies which guide image production, but rather, this term is descriptive of particular processes executed as a sequence of formal procedures.  As such, the machinic aspect of analogue artistic processes of automation may take form concretely through apparatus or through technique. Considering the relationship between an image and the code used to produce it in terms of instructions, the following examples give depth to our notion of automated artistic practices through their procedural aspects.

Julia Hoelzl and Rémi Marie point out that the algorithmic nature of digital images shares a lineage with the history of cartography (Hoelzl/Marie 99), which shifted from thinking of maps as representing the world in pictures to instead such a representation taking the form of a data set. After the loss of Ptolemy’s atlas, geographia, the maps were able to be redrawn by cartographers centuries later, based on the detailed data and instructions which Ptolemy had transcribed. In this sense, the systematised coordinates and measurements of features on the maps enabled them to be saved, transmitted and reiterated from numerical data into images. In another less-than contemporary example related to the history of procedural images, the canon of proportions outlined by Leonardo DaVinci in his Vitruvian Man describes the human body geometrically, as if to function as instructions for its construction. In each of these instances, mathematical formulae and the systematic cataloguing of the internal relationships enabled images to be transcribed, stored, transmitted and reiterated. Database-type images such as these or most digital images may be the result of complex processes, but although they involve some level of flexibility in their execution, they are for the most part highly predictable.

In early computer art practices, Vera Molnár is known for her “machine imaginaire”, which implemented operations for the production of visual outcomes, the artist herself taking on the conceptual role of a computer performing tasks based on a set of pre-defined rules (Broeckmann 141–142). Conceptual artists have also used rule-based systems in their work. Sol Lewitt’s practice is especially relevant, with his perspective that “The idea becomes a machine that makes the art.” (in Kosuth) In his instruction-based drawings, the artist gives directions for the construction of the work, which may be executed with some degree of variation. The openness of the artwork to being carried out in more than one way relates closely to the same property in computational images, that there may be variability in the results of a single algorithm. Considering concepts as mechanisms for driving artistic production, we see the close parallels between procedural logic and conceptualism, implementing cognitive processes in order to arrive at new conclusions in the process. Other artists such as Yoko Ono, with her Instruction Paintings and Lawrence Weiner with his declaration that “the piece need not be built.” touch upon the latency of instructional works, which, in the case of conceptual works, may be enacted by the viewer, and which have the potential to be expressed in different ways or not at all. Similarly, more highly automated media, such as dynamic interactive or iterative works, especially those employing machine learning, can be made to be different every time they are run and experienced.

Conclusion

Automation and procedurality can be thought of as supporting or augmenting an artist’s practice, enabling them to explore in different ways they may not have discovered on their own and to break out of entrenched habits, norms and value systems. In some circumstances, automation can be considered to take the place of artistic decisions considered to be arbitrary or which would otherwise be left to intuition. Automation may also be viewed as a way of narrowing down the search space for the production of a work, allowing artists to determine a smaller frame to explore out of potentially endless possibilities. The choice of which elements are automated or left to chance procedures is highly impactful upon the outcomes, for instance, whether employing automation for the purpose of drawing inspiration or if it is actually employed for the execution of the work. Using systems which employ randomness within artistic practices enables new modes of creative decision-making, for instance, determining components, arrangements and individual aspects of elements used in the work. In Surrealist practices, for instance, the implementation of randomness contributed to the creation of new ways of meaning-making by embracing meaninglessness and forcing new pathways to be forged. While this kind of engagement with automated processes has the potential for diverse or unpredictable outcomes, it may not necessarily result in meaningful, interesting, or aesthetically-pleasing results, as they are decided purely by chance. The radical adherence to formal procedures, in spite of potentially deriving undesirable conclusions, makes this kind of artistic methodology especially interesting as it presents a potential break with existing methodologies and aesthetics. Highly automated practices, such as unsupervised machine learning, could lend toward the production of new modalities of artistic exploration and, ultimately, toward a new kind of conceptual aesthetic.

Continuation of Research

In the continuation of this research, artistic methodologies working with algorithmic procedures will be explored in relation to the processes at work in machine learning. Rather than considering machine learning as merely a means to an end, as is the common tendency in current artistic practices, this research examines the potential implications of machine learning as a methodology for the creation of a conceptual aesthetic produced by or in cooperation with machines. Much as human logic is employed for the purpose of reasoning through problems, machine learning now often plays a role in making sense of unstructured data (Greenfield 184). This research speculates that a machinic logic may present forms of computational conceptualism and radically alien aesthetics, which may appear as randomness, error or undesirable results. The aim of this research is to conduct a critical investigation of machine learning in conceptual terms, drawing parallels between the procedural logic of machine learning algorithms and procedural artistic methodologies.

References

Broeckmann, Andreas. “Image Machine.” Machine Art in the Twentieth Century, Massachusetts Institute of Technology, 2016, pp. 123–65.

Carvalhais, Miguel. Artificial Aesthetics, Editora da Universidade do Porto, 2016.

Farocki, Harun. “Phantom Images.” PUBLIC, edited by Saara Liinamaa et al., vol. 29, 2004, pp. 12–22.

Greenfield, Adam. Radical Technologies The Design of Everyday Life. Verso, 2017.

Hoelzl, Ingrid, and Rémi Marie. Softimage: Towards a New Theory of the Digital Image. Intellect Ltd, 2015.

Kosuth, Joseph. Art After Philosophy and After: Collected Writings, 1966-1990. Edited by Gabriele Guercio, MIT Press, 1991.

Lehman, Joel, et al. The Surprising Creativity of Digital Evolution: A Collection of Anecdotes from the Evolutionary Computation and Artificial Life Research Communities. 2018, pp. 1–32.

Manovich, Lev. The Language of New Media. MIT Press, 2001.

Massey, Irving. The Neural Imagination: Aesthetic and Neuroscientific Approaches to the Arts. University of Texas Press, 2009.

Parisi, Luciana. Algorithmic Control. Critical Theory of Digital Cultures, Goldsmiths, University of London.

—. “Mediality with Ends: When Machines Start to Think.” The War of Appearances: Transparency, Opacity, Radiance, edited by Joke Brouwer et al., V2_Publishing, 2016, pp. 44–63.

Verbeek, Peter-Paul. What Things Do. Pennsylvania State University Press, 2005.

“Vitruvian Man.” Wikipedia, 2018, https://en.wikipedia.org/w/index.php?title=Vitruvian_Man&oldid=863637803.

Weibel, Peter, et al. Eigenwelt Apparate-Welt. Edited by David Dunn, The Vasulkas, 1992.

Zylinska, Joanna. Nonhuman Photography. MIT Press, 2017.