Week 5: Stefanos & Diaa

Stefanos

​In a previous entry, I mentioned some studies that aim to classify paintings based on quantitative measures. I wanted to follow up on that in some detail, because, even though this is an exciting nascent field of scientific inquiry, one must be mindful of the scope of each study to avoid misinterpreting the results. This is dually important in this context, as there can be misunderstandings on both the artistic and scientific sides. I outline a simple example below.

One of the studies I mentioned was the recent AI-assisted modern art classification of Sigaki et al. The authors use two information theoretic measures - entropy (in its information-theoretic sense) and complexity - that they calculate for each painting. They also use a machine learning algorithm to automatically assign predetermined labels, corresponding to artistic movements, to each painting in a large database. The authors then show that paintings belonging to a certain movement tend to aggregate in a certain region of the entropy-complexity plain. Interestingly, they were able to infer pivotal points in history that led to important stylistic shifts.

The complexity-entropy classification of art movements from the article by Sigaki et al.

This classification scheme applies to large collections of paintings, and not to individual paintings themselves. Consider, for example, the data points the periods 1031-1570 and 1939-1952. These have almost the same entropy and very similar complexity measures, even though the artworks produced in these two periods can look markedly different - see example below. It becomes clear that individual paintings may evade or trick this method.

The Healing of the Lame Man, Raphael, ca. 1515. Right: Composition X, Wassily Kandinsky, 1939

One can ask whether this classification protocol can be improved to detect differences in style on a finer level of individual paintings or small collections, such as periods of a painter’s life. I am intrigued by this question and have been thinking about ways to answer it affirmatively. One could, for example, try to assign to each painting an entire spectrum of values of a characterizing quantity, instead of just two (or a handful of) numbers. I think there is a straightforward way to do this. To outline what I have in mind, I first have to explain the tools used in the aforementioned work in a bit more detail.

The definitions of complexity and entropy in the diagram above involve the calculation of a distribution of probabilities of “minimal structures” appearing in the digital copies of the paintings studied. The minimal structures used were all the 2x2-pixel squares in each painting. Imagine, for example, that paintings are restricted to consist of only black and white pixels. Then, each pixel can take one of two values, black or white. The number of 2x2 minimal structures is then 2x2x2x2=16. The number of occurrences of each minimal structure is calculated by scanning the image, yielding a probability distribution. This distribution is then summed in two different ways, yielding the two numbers relevant to the analysis, i.e., entropy and complexity. Of course the analysis allows for more than just black and white pixels – the authors in fact use grayscale with 256 levels of intensity.

What I claim is that, instead of the summed values of complexity and entropy, one could use directly the distribution of probabilities of minimal structures to identify paintings. One can then use the notion of distance between probability distributions to define a finer distinction between styles. I hope to get the chance to try this out soon. Funnily, my collaborators and I used a similar notion of distance in recent work that characterizes the time evolution in certain quantum systems. Talk about science providing inspiration!​

---

Diaa

I would like to expand Stefanos’ explain regarding my interactive neural holographic puzzle.
In fact, this neural artwork has its conceptual roots in Marshall McLuhan’s writings, when he Claimed the media as an extension of our body. However, these kinds of neural interactive setups have exceeded what did McLuhan imagined when he talked about our body totally that can perform a wide range of physical activities, which could be expanded by using media.

But now after the human body is fragmented into its original systems and it became possible to use our neural or biological system alone in order to integrate them into an artificial system, the extension of our body, in this case, became stronger and effective than its source (the human body itself) to the extent that there is no longer a need to the human body as a unified physical form in the artwork (as an artist or as a participant) as long as there is a part of it (the neural or biological system) that can perform its functions, more effectively, in the artistic environment.

While the human body was a main source of the actions in the beginning of the interactive techniques, the neural or biological system of our body can govern the new generation of the interactive techniques as neural or biological actions. In this case, the consequences of the action (the neural or biological responses of the human body) are more magnified and expanded than the action itself inside the interactive field in its physical environment.

By this way the human body is no longer a physical part of the artwork itself, rather, its basic systems (neural or biological) are expanded as a part of the artwork, which can represent a form of a post-body actions in the interactive arts, in which the artwork system can build a direct, mutual contact with our body systems, excluding the physical activities, in order to integrate the core system of our bodies’ functions.

Similarly, in the first assumed neural-holographic system, the holographic puzzle must be considered as processes that can be governed by the participant's neural responses. By this way, the holographic processes are in direct connection with our brain. This contact, in fact, became the essence of the artwork, in which all information, derived from the entire artistic processes, has been decoded neurologically to help the participants to reshape their aesthetic experience and conclude an operational definition of their interactive actions neurologically.

Although the New-Media-art, as a separate artistic discipline, has achieved several paradigm shifts in contemporary visual arts, the most important one of them raises a challenging question about the nature of the aesthetic pleasure of artworks, not only in contemporary arts, but also in traditional arts.

In fact, the aesthetic pleasure exists in our feeling, not in the artwork itself. It is produced by a chemical reaction between the artwork (as a stimulator) and the viewer’s brain (as a receiver) to affect our feeling to make us feel pleasure. in media art, any artwork is a visual equation that can stimulate a kind of an aesthetic pleasure inside the viewers. Of course, every artist, traditional or contemporary, has his special way to create his visual equations. While the traditional artists prefer to write their visual equations in a complete visual/pictorial form, the interactive artists and media artists prefer to invite the viewers to participate in writing their equations.
​
Since the beginning of the processes art in the mid 1960s, several artists began to think about the core of the artwork, is it in the final product? or is it in the processes carried out to achieve it in the physical world?

In fact, in the contemporary, interactive media arts, there is no physical existence in the artwork itself. What we have already done is; we removed the external visual stimulator (the traditional artwork), and instead, we designed systems that can deal with the main sources of the aesthetic pleasure directly, seeking to prove that the artwork inside us, the artwork is what we feel, the artwork is what we think.