Media Arts and Technology

MAT594GL Techniques, History & Aesthetics of the Computational Photographic Image
https://www.mat.ucsb.edu/~g.legrady/aca ... f594b.html

Please provide a response to any of the material covered in this week's two presentations by clicking on "Post Reply". Consider this to be a journal to be viewed by class members. The idea is to share thoughts, other information through links, anything that may be of interest to you and the topic at hand.


Report for this topic is due by November 17, 2020 but each of your submissions can be updated throughout the length of the course.

When considering the foundational goals of computational aesthetics according to Anselm Brachmann and Christoph Redies’s Computational and Experimental Approaches to Visual Aesthetics (2017),

In 1876, the founder of experimental aesthetics, Gustav Fechner, published his seminal book entitled “Vorschule der Ästhetik” (Fechner, 1876). He believed that the aesthetic appeal of physical objects manifests itself in stimulus properties that can be measured in an objective (formalistic) way. Specifically, he attempted to show that rectangles with an aspect ratio equal to the golden ratio are more appealing to human observers than rectangles having other aspect ratios.

The intent of computational aesthetics is to mechanically reproduce a generalized, innate human aesthetic understanding. In class we went to the notion of human evolution as the source of innate understanding, which I believe holds some very valid footing. However, it is always a bit murky to get into generalized human experience, and I certainly don’t agree that “computational methods are objective in nature” (Brachmann and Redies). The key advantage of computational aesthetics is that it has the possibility to represent a multitude of subjective experiences, but it would be misrepresentational to consider the outcome objective in itself.

My Work
In my own studio practice, I am drawn to the influence of unconscious biological functions as representing an “innate'' embodied human experience, and how these processes override how I would understand a computational process would currently understand external space.
For my work, I incorporate research that suggests the human brain possesses plasticity to overcome and artificially introduce information to make sense of an individual's surroundings. Because the brain has the ability to impose artificial information onto external perception, human biological consciousness is inseparable from misinformation. However, an interesting study on hemispatial neglect (a condition where the brain compensates for unilaterally missing visual information by cognitively filling in the gaps) shows the subconscious body's ability to override misinformation with gesture. This image below shows the unconscious influence of the gesture of drawing a circle and how this exists beyond the conscious understanding of the external environment. Cy Twombly via Roland Barthes
I believe gesture is essential when looking at the possibility of computationally replicating innate aesthetic understanding. This is why looking at Cy Twombly is particularly intriguing. Roland Barthes isolates the gesture of Twombly in Works on Paper (1979) with

Hence let us distinguish the message, which seeks to produce information, and the sign, which seeks to produce an intellection, from the gesture, which produces all the rest (the surplus) without necessarily seeking to produce anything…TW, contrary to the venture of so many present-day painters, shows the gesture. We are not asked to see, to conceive, to savor the product, but to review, to identify, and, so to speak, to enjoy the movement which has ended up here.

The gesture is the compilation of practiced movements held in the subconscious that can be isolated from iconographic influences.
Phenomenology: Merleau- Ponty
To examine the embodiment of artist creation and understanding, it seems essential to make reference to the writings on the phenomenological aesthetic experience by french philosopher Maurice Merleau-Ponty. Merleau-Ponty offers a physical understanding of art as productions of a carnal interpretation of the artist’s external world. In this model, perception is thought of as a creative piecing together of activities to draw meaning rather than having a pre-existent meaning of the world. On that note, the aesthetic experience is then thought of as a mode of extension to the understanding of the world through a semantic interpretation.

Merleau- Ponty’s writings focus on the primary embodiment of consciousness and consequently, the fundamental contact of this body to the outside world through tactile practical interactions, such as looking, handling, and using objects. Merleau-Ponty’s work on the physicality of perception then becomes an interesting conversation about the limitation of computational aesthetics that exist without human embodiment. I think that there is a case to be made about computational gesture, however, it is fundamentally different from embodied human gesture so would be a misrepresentation to conflate the two. It would also be quite difficult, and probably impossible, to isolate a generalize representation of gesture as gesture is so closely tied to cultural/social habits and movements.

Our initial discussion of generative art led me to ponder the possibilities of other analog works could fit within the definition given by Galanter:

“Generative art refers to any art practice where the artist uses a system, such as a set of natural language rules, a computer program, a machine, or other procedural invention, which is set into motion with some degree of autonomy contributing to or resulting in a completed work of art.”

Immediately, I was inspired to create a non-digital system, where viewers participated in the creation of the work. Yet, I was not able to find any examples of this—where the completion of the work was entirely up to the people who set off to see the work in the first place (when it did not yet exist). How would the showing be advertised? I wondered. One tentative project that entered my mind was coming up with a grid of squares of identical sizes, and having participants roll dice to determine which color they will fill their portion. From this basic example, things took on a life of their own.

This led me to further research. Besides Hans Haacke’s “Condensation Cube,” what other analog forms of generative art already existed. According to this article: (http://www.artificial.dk/articles/galanter.htm), Galanter makes the claim that tile work fits within the framework of his definition. And if that was the case, quilts, masonry and countless other forms as well.

Another aspect that struck me was the aesthetics of the overwhelming majority of generative works. Identical to conceptual art, where the end product serves more of a token of the overall idea than something you’d hang in your home, the majority of generative works (especially when computer driven) seem to serve as a talking point for the observer to inquire how the piece was produced, than the true end product, with all roads pointing back to the process. That said, I was determined to find some pieces that spoke to me, and would also perhaps look nice in my own home one day.

The two most impressive examples I came across were Michael Hansmeyer, an architect using generative rule-based design to create innovative installations, and Argentinian creative coder, Manolo Gamboa Naon.

Hansmeyer’s columns and vaulted ceilings are especially impressive. (http://www.michael-hansmeyer.com/projects) and Manolo shares all of his work (https://superrare.co/manoloide) and code (http://manoloide.com) for free.

Excavating AI
______



I knew at some point in this class we would dive into the technical aspects of machine learning, and I knew that most of this information would be in a language that I either do not speak at all (C++, Python) or understand in a limited sense (javascript, html, etc.) In other words, my programming skills are rudimentary at best. This is despite my having a background in working with computers since their inception.

However Weihao’s presentation was remarkably concise and well articulated, to the point that I was able to understand the basic underlying structure of CNN’s, without necessarily having the skillset to create/modify/operate them myself.

Weihao’s demonstration also reminded me that while we want science to be exact, there are many instances in the development of machine intelligence where computer functions are often, if not always, initiated by - or subject to - human interpretation. And because we know humans are far from infallible, we must understand that our machines, as being essentially human creations, will inevitably fall victim to similar shortcomings.

This led me to revisit and share research Trevor Paglen & Kate Crawford present in an essay titled Excavating AI, The Politics of Images in Machine Learning Sets. As Weihao pointed out, CNN’s require an initial dataset which is used to train these networks how to recognize (in images) and correctly label (in text) specific objects, as well as facial expressions.





The central publicly available dataset used for such training purposes is known as ImageNet, a creation of researchers from Stanford and Princeton. Paglen and Crawford, in their own research, did an extensive survey of this dataset and discovered it’s foundations to be a “bedrock composed of shaky and skewed assumptions.”

In creating ImageNet, Individuals have been tasked with classifying images with text-based descriptors (all nouns up to this point) and have not only injected bias into this process, but highlighted the very “slippery” task of image interpretation, and the subsequent difficulties in pairing image and text in an attempt to create a truthful expression.





As Paglen & Crawford point out, “Entire subfields of philosophy, art history, and media theory are dedicated to teasing out all the nuances of the unstable relationship between images and meanings.” It is for precisely these reasons that “the automated interpretation of images is an inherently social and political project, rather than a purely technical one.”

It is also a project that I think we all understand to be open to artistic interpretation.


Links:
https://www.excavating.ai

Roth's Archival Art or Throwing Your Hands In the Air Because Nothing is Good
The giving up of (varying degrees of) control in generative art, or, "negative intentionality", reminded me a lot of Dieter Roth's artwork Flat Waste (Begun in 1973) [1]. A long term project during which Roth meticulously collected, flattened, and catalogued food packaging and other found garbage, it fits the description of an artwork driven by a program/algorithm.

While I do not know the motivation for early generative art artists, Roth's motivation for projects such as Poetry Machine (which consist of a black paper with some holes punched in it) seem to be an anxiety that came from his self-criticism and doubts on what art can do[2]:

The great judges of European literature and art are located somewhere in my head and play the judge there and do me in. I have to keep watching out that I don’t offend them.

I would call myself an inventor of machines that are meant to entertain (or inspire) feelings (or thoughts) that help to digest this Central European civilization wading in junk.

Roth's giving up of control resonated with me since I became interested in generative art while feeling overwhelmed by the cumulative process of painting, as each step seem to involve conscious and unconscious choices that make up a somewhat formulaic process.

(Roth's Flat Waste seems more interesting to me in that his collection and making meaning of debris/waste is very similar to the way in which machine learning tools is said to be based on the "mining" of "excess" or "useless" data; see Thylstrup).

Emergence
Generative art seems to resemble simulation in that the "emergent properties" can be seen as result of a human spectator's failure to "comprehend the consequences of an algorithm or process"[2]. Viewing the emergent properties in this light, it seems the reason that Roth's Flat Waste does not seem "generative" enough is that it is an aggregation, the result of which (or at least the visual appearance of the result) can be predicted and comprehended by a human mind.

If we change add the requirement to be incomprehensible(/not compilable) to human mind to definition of generative art, it could also mean that Guido Segni's Demand Full Laziness: A Five Year Plan (2018-2013) (2018-present), during the first year of which the artist trained a GAN on footage of him resting and laying in bed, is more aggregative than generative (in a visual or formal sense).



Notes
[1] Inke Arns, qtd. in Soderman and Howe
[2] https://www.poetryfoundation.org/harrie ... eter-roth-
[3] https://old.guidosegni.com/work/demand- ... 2018-2023/
Bibliography
Hamilton, James F. "Drone Journalism as Visual Aggregation: Toward a Critical History." Media and Communication, vol. 8, no. 3, 27 July 2020, pp. 64-74, www.cogitatiopress.com/mediaandcommunic ... /3117/3117.
Nanna Thylstrup, "Data out of place: Toxic traces and the politics of recycling"

As Galanter puts it in his canonical publication What is Generative Art? Complexity Theory as a Context for Art Theory, the generative art is an "art practice where the artist uses a system, such as a set of natural language rules, a computer program, a machine, or other procedural invention, which is set into motion with some degree of autonomy contributing to or resulting in a completed work of art." This definition emphasizes an 'algorithm', or a procedure that is employed in the creation of the resulting artifact. From the examples in our lecture, Sol LeWitt's Wall Drawings stood out to me as an analog way of creating a generative work. Here, LeWitt does not provide the work itself to the galleries, but he gives the algorithm that recreates the work. This way, even though the work has a fixed form, it could be reborn on-site where the audience would enjoy it.

The digital generative artworks, by the nature of their medium, could employ the interplay between simplicity and complexity in the same work. Screen-based works employ various control parameters to control the variations occurring in the piece. One mathematical construct that could be particularly useful for such control is Shannon's entropy. It is presented as a measure of information, surprise, or uncertainty in a variable's set of possible outcomes. In this regard, I wonder if it's possible to embed higher-level descriptions in a generative system that employs information theory concepts borrowed from Shannon's discoveries.

Although Generative art is often referred to algorithmic art, or computer art, I think it dates back to the age before computer is invented. Marcel Duchamp's idea of ready-made liberates the artist from the manufacturing of artifacts. The conceptual artists in Sol Levitt changed people's traditional thought that artist is the manufacturer of the artifacts. On contrary, artists can make explicit rules that can then be executed by anyone without any knowledge of art. The rules themselves are the art.

How similar they are close generative art! Artists make rules with programming, then have the computers, in replacement of the past museum assistants, to execute the rules and generate images. Programming becomes a medium for artists to work with, and computers becomes a free labor to produce art. After the production, another role of the artists is to curate the results and present them.

Empowered by computer, the rules of art making become increasingly complex, especially when involving data processing. Once the rules is designed by the artist, the computation power enables the rules to scale up with the expansion of data volume. The flight information over the world can be display on one image in real time. Years of library check-out records can be visualized on one image. DNA data, weather data, astronomic data can all be used and mixed in the art creation, as long as a rule is programmed by the artist. This opens great space of exploration.

-new media artist manfred mohr

This is one of the first generative art designed by his systems he created over 50 years ago. He sent out the instructions to the system let it to draw rectangles in various shapes and sizes in specific positions. He utilized random function to create this composition Shohei Fujimoto - use algorithms to create geometrical space (I really like his works)
links:
http://sfjmt.org/
https://vimeo.com/shoheifujimoto
https://www.instagram.com/shohei_fujimoto/