Report 5: Computational Aesthetics, Generative Art

Post Reply
Posts: 160
Joined: Wed Sep 22, 2010 12:26 pm

Report 5: Computational Aesthetics, Generative Art

Post by glegrady » Tue May 04, 2021 11:43 am

MAT 255 Techniques, History & Aesthetics of the Computational Photographic Image ... s255b.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.

Topic: Computational Aesthetics, Generative Art

Report for this topic is due by May 13, 2021 but each of your submissions can be updated throughout the length of the course.
George Legrady

Posts: 6
Joined: Thu Apr 01, 2021 5:33 pm

Re: Report 5: Computational Aesthetics, Generative Art

Post by kevinclancy » Wed May 05, 2021 6:11 pm

I enjoyed reading the series of research papers this week. The academic research paper is a totally different format than I am used to, and provides a different perspective and approach. I definitely took issue with Bo et al’s rigid, systematic approach to “beauty” and “ugliness”. I think there are many problematic aspects that could arise in these biases, and it is crucial to think through the foundations of thoughtfully and ethically training AI. I think artists/philosophers and scientists/engineers are a critical balance on each other in this regard. Michel Serres wrote very poetically about this synthesis between Art and Science in what he termed the Northwest Passage. Galanter straddles this line really well, as both an artist and academic. I particularly appreciated his delineation of not classifying Pollock as a generative artist, because such a leap would universalize the term generative art to the point of no return, where it is completely meaningless and useless as a term. I was definitely starting to go down that track thinking about physics, fluid dynamics, and velocity as generative systems, but Galanter reminds us that: “All artwork has underlying physics, and if that were the measure then all art would have to be called generative art.” I wonder if physics engines in computer modeling could still be considered generative systems, because artists often tweak the parameters of physics engines to get bizarre and uncanny effects. Galanter's paper reminds us that sometimes rigidity and classification are vital to effective classification, and as artists we can overthink it.

Dina Kelberman, I’m Google (2011-ongoing), website,

In our discussion of computational aesthetics, computer vision, and machine learning, I couldn’t help but think of artist Dina Kelberman’s long running tumblr, I’m Google (2011-ongoing). Kelberman's endlessly evolving image set appears to be AI generated, but she meticulously scoured Google Images for over a decade to string together images that cascade together in mesmerizing, bizarre, and surprising ways. The string of images are literally linear, but they also function in non-linear and abstract forms of relation, which relate to the “diachrony” and “synchrony” we would later discuss in John Baldessari’s work. In a way, she is collaborating with the Google Image search algorithm—playing a role of aesthetic mediation—but she also critiques the algorithm and has designed her own work arounds to use the search engine as a creative tool. Her process speaks to the aesthetic value of order, within the equally interesting systems of noise and chaos we have previously discussed. Her project really makes me think about how much human sensibilities and biases are either present or absent, reflected or omitted, within AI systems.

Sarah Rosalena Balbuena-Brady, ABOVE BELOW, AI-generated textile, cotton, training: Mars Reconnaissance Orbiter satellite images taken from High Resolution Imaging Science Experiment (HiRISE), 2020, 60 x 80 in

I also thought of Professor Sarah Rosalena Balbuena-Brady’s work ABOVE BELOW (2020) in relation to the discussion of the Jacquard Loom, and her project Deformation of 50,000 Letters (2017) in relation to Bo et al’s analysis of Chinese letterforms in their paper Computational aesthetics and applications. Balbuena-Brady’s most recent exhibition ABOVE BELOW, exhibited at Blum & Poe Los Angeles, features a series of AI-generated textiles that use a neural network trained on Mars Reconnaissance Orbiter satellite images taken from High Resolution Imaging Science Experiment (HiRISE). In her thoughtful essay published on Fulcrum Arts, Balbuena-Brady states:
I am engaged with computation and fiber art in relation to the Jacquard loom because of its relationship to image production. The first computer algorithm was written by Ada Lovelace while she observed its capability to weave intricate flowers and leaves. The exchange between imaging and the loom untangle contemporary understandings of mapping by materializing computation.

Similarly, woven geospatial imagery on the Jacquard loom embodies the computer’s earthly origin from cotton thread to pixel and around again—a web of the past and future geographies. AI-generated textiles reveal a tactile connection with the pixels it signifies, tracing its production, operating between reality/artificiality and material/immaterial, and back again.
Balbuena-Brady’s connection to Ada Lovelace, Joseph-Marie Jacquard, algorithms, and the pixelation of botanical forms finds interesting resonance with both Anna Ridler and Hito Steyerl’s exploration of machine learning and speculative botany. Both Ridler and Steyerl critically examine the problematic history of British statistician Ronald Fisher's ubiquitous data set. To quote Ridler:
The iris flower dataset, created by British statistician Ronald Fisher, contains 50 samples of 3 different irises and is used as an example for many statistical classification techniques in machine learning. It is included in the package Scikit-learn so that every machine learning programme that uses this package also contains within it somewhere a hidden flower dataset. This unexpected link brings the installation into the history of machine learning. But by referencing Fisher, I am also referencing the fact that he was also heavily involved in racism and eugenics (foreshadowing perhaps some of the inherent problems with machine learning, bias and datasets). Even something as simple as a flower contains within it hidden layers and narratives.
Anna Ridler, Myriad (Tulips) (2018) C-type digital prints with handwritten annotations, magnetic paint, magnets, ink

Hito Steyerl, installation view of This is the Future / Power Plants, 2019, in “I Will Survive” at Kunstsammlung Nordrhein-Westfalen, 2020. © VG Bild-Kunst, Bonn, 2020. Photo by Achim Kukulies. Courtesy of Kunstsammlung Nordrhein-Westfalen.

Dina Kelberman “I’m Google” :
Sarah Rosalena Brady : ... 00-letters
Ada Lovelace:
Anna Ridler “Myriad (Tulips)” :
Hito Steyerl discusses “Power Plants” :
Hito Steyerl’s lecture at MIT Wasserman Forum (to be uploaded soon) :

Additional Links:
Oliver Laric “Versions” :
Rhizome Net Art Anthology:

Meant to share this when we were discussing pixels:
Daniel Rozin :
Last edited by kevinclancy on Tue May 18, 2021 11:25 am, edited 1 time in total.

Posts: 6
Joined: Thu Apr 01, 2021 5:31 pm

Re: Report 5: Computational Aesthetics, Generative Art

Post by jungahson » Wed May 12, 2021 11:51 pm

While reading Bo et al.'s paper, I focused on the description about what research has been done in modeling abstract painting of well-known styles. For instance, Kandinsky's paintings have been modelled with rule-based approach. As I have also worked on reconstructing Kandinsky's paintings, I agree that Kandinsky proposed certain rules and theoretical principles, and applied them to his artwork. However, I think what was important in his work was not the mere form but the inner content. Although abstract paintings did not represent what is in this world, they do have subject matter, which was called "hidden" reality.

Kandinsky's teaching on analytical drawing aimed in looking, precise observation, and the precise representation of constructive elements, which is closely connected with this hidden reality. Below is Bella Ullmann-Broner's example of student exercises in his class. In this class, students developed their drawings starting from still life as still life was perceived as an artistic medium with an important transitional role in the evolution of abstraction.
Analytical drawing was a process in three stages: simplification, analysis, and transformation. As you can see from Bella's example above, the first stage, simplification, required the students to subordinate the whole complex to one simple overall form. Next, the second stage was to make clear the tensions discovered in the structure. The last stage, transformation, advances the aspect of the second toward more radical, freer abstract solutions. You can see that the rightmost image turned into abstract painting that is hardly recognizable from the first image. I found it is important to understand these steps as it promoted students' ability to perceive the abstract, the essential form.

Inspired by Bella's exercise, I created a GUI that translates a photographic image into analytical drawing. Some image processing methods such as edge detection, corner detection, and network generation have been applied. Although I am using dataset which has mappings between objects in photographs and 3d models, I hope to automatize the whole process by working on object detection and poze detection for the 3d models. As I believe the whole process is important, I have displayed how the photograph is being transformed into the drawing, rather than showing only the result. I think that the understanding of this process should be included in modeling abstract painting of Kandinsky.
Last edited by jungahson on Thu May 13, 2021 10:24 am, edited 37 times in total.

Posts: 6
Joined: Thu Apr 01, 2021 5:32 pm

Re: Report 5: Computational Aesthetics, Generative Art

Post by alexiskrasnoff » Thu May 13, 2021 1:36 am

We touched on Mario Klingemann's work very briefly, so I thought I would link a few things about him and his work if anyone is interested! Here is a pretty good summary of his body of work: and here is a video about his Memories of Passersby I piece that goes a little more in depth into the work and the process of its creation.

I also put this in the chat on Tuesday but I wanted to link it here as well so it's a little more permanent. Here is the Google Colab for Fabian's feature visualization demo so you can try it out and see some of the feature vis concepts in action! ... tion.ipynb

The Galanter article was very eye-opening for me about the way generative art is defined. I never really thought about the fact that 3D modeling/CG and using game engines are forms of generative art, but that definitely makes sense. A cool example of a tool for digital 3D generative art that also incorporates AI is this Colab demo called PIFuHD, which uses a neural net to generate a 3D model of a person from a 2D image. I've used it a lot in previous works and it's really fun to play around with so check it out! ... 5G5jmh2Wgt

Another form of generative art that was mentioned in the other paper was fractal art, which uses mathematical equations/fractals to generate really interesting images. A tool that I've played with to create fractal art that wasn't mentioned was Chaotica, found here: To be honest, I don't understand a lot of the math or functions, but by just messing with parameters you can generate some pretty cool images, here are a few I've made:

Post Reply