Selected Writing

Below is a collection of my writing specifically related to new media art and technology. If you are interested in staying up to date or seeing all my writing, you can follow me on Substack:

Can you Create Art with Artificial Intelligence?

Of course, but that doesn’t make it any easier.

Society today tends to polarize concepts into binary camps: in-group vs out-group, right vs left, art vs non-art. Us vs Them. 

Rather than support the development of a nuanced, textured viewpoint with room for everyone, these discussions devolve to shouting matches without any chance at progress.

This cultural amplification and resulting loss of nuance is pervasive, and rapidly becoming so normal it is easy to forget the nuance ever existed.

I am frequently asked my thoughts on art in the age of artificial intelligence, often phrased something like "So, do you think AI-generated images are art?" For a while, I struggled to answer, feeling that both *Yes* and *No* were inadequate.

Both options felt inadequate because they were the only two "correct" answers as defined by societal norms. To arrive at my true feelings on the subject, I had to reject the question and the constraints its form placed on my answer, which is the subject of this essay.

What is Art?

Art is a realm of nuance and subjectivity, a heady concept that consistently defies attempts at definition.

Definitions are the mindkiller. What can we say about art that does not require us to define it? I propose a simple axiom:

Art is agnostic to medium

Humanity has demonstrated an impressive capacity to produce and consume art of all types throughout history. Sculpture, dance, music, writing, painting, etc. Artists have been creating in these mediums for centuries, and many in more than one medium. Media have no embedded heirarchy: a song is not more art than painting just because it is a song, and visa versa.

Thus, I do not believe the medium of an artwork correlates at all to its artistic merit. The term medium actually originates from the Latin medius, which literally translates as "something in the middle."

What can we derive from this axiom?

The medium is interface, establishing a shared context between the artist and viewer within which a message may be transmitted.

A strong message is a cohesive whole, a unique and emotional experience in which every component has both form and function. There are no extra pieces, or missing pieces. It is this cohesion that creates the illusion of inevitability, that feeling you get that this thing must exist exactly in this way to exist at all.

That message (or concept) is the artwork. 

The medium is fungible, as is the viewer. The artist themselves is fungible, ideas have no owners. Like the unspoiled lands of a new continent, ideas are discovered, not created. They are also easily spoiled by overuse.

Here are a few implications from this line of reasoning:
    1. Medium itself does not determine whether or not something is art
    2. Originality matters
    3. The message should dictate the medium, not the artist

Now that we have a bit of a framework in place, and can at least talk about what does not make (or break) something's "art-ness," we can return artificial intelligence.

AI is simply another medium

As such, whether or not AI-based tools were used in a work's production process does not impact its artistic merit.

What determines if a painting is art? Certainly not just the fact that it is created with paint on canvas. We consider the painter, why they painted, how they painted, and who they painted for. These are all essential to the judgement of the painting. It is the message, not the medium, we are evaluating.

So, is the output of a large generative model art?

Maybe. Maybe not. What is the message?

Data Visualization in New Media Art

New media projects labeled data visualizations tend to be quite aesthetic but hardly informative, with the aesthetics bearing virtually no conceptual or graphical connection to the data being displayed or the narrative embedded within it. This a disservice to the medium, as it treats the viewer as unsophisticated and squanders an opportunity to create a dialogue with them.

Edward Tufte, considered the father of graphical data representation, believed truly excellent data visualization induced the viewer to ponder the substance of the data rather than the technology that produces it. However, if a data visualization does not communicate with the viewer in an actionable and relevant way for them (i.e. convey meaning), then the technology producing it is the only subject left for the viewer's attention.

Data is an Observer

Interaction design is a critical component of experiential and new media art. In their essay Architecting Experience, Patten Studio outlines their framework for desiging impactful interactions. It is a fantastic read and one I strongly recommend in full, but for this essay we just need one concept:

Many graphical interfaces are designed to be used by only one person at a time. As interactivity moves beyond screens, it is increasingly important to design systems for use by multiple people. This might mean face-to-face collaboration or play, or it might mean one person is interacting while others watch. In either case, it is important that whoever is watching the interaction is able to follow what is going on: to understand cause and effect, and to surmise the general rules of engagement. We call this interaction legibility (italicized original).

The quote above talks about the need for legible interactions when there are multiple human users, but I am going to tweak it a bit: Legible interactions are critical whenever a system has multiple observers. An observer is anything that can interact with a system and alter its state. This includes:
  • Human viewers
  • Physical computation sensors
  • Data of any sort (including digitally native agents, such as other softwares)

As there must be at least one human viewer in the environment for any discussion of design to be practically relevant, we will further assume there is going to be someone looking at the system, i.e. every system will have at minimum two observers.

Extending the concept from Patten Studio, this means that legible interactions are just as critical in a data visualization installation as they are in a purely motion-tracking installation. A viewer is the second observer, and must be able to correlate interactions from other observers (data, in this case) with changes in the system's visible state.

Framework for Data Visualization in New Media Art

Or at least some questions I ask myself:

    1. Is a data integration necessary for the concept? What story does the data tell that we could not impart without it?
    2. Does the data legibly change the visual result, in a manner correlated with the data's semantic payload?
    3. Could I replace this data with a procedural noise without loss of conceptual quality? If I were to use noise instead, would my choice of noise be listed in the description?
    4. Any data used should be clearly documented, with mappings to visual impacts as well. Do not hide weak concepts behind imprecise descriptions

New media artist Alex Czetwertynski has a thoughful essay on this topic as well, which helped structure some of my own thoughts.

Data without the Viz

None of the above is to say that new media works cannot integrate data in a purely conceptual / aesthetic sense that does not establish a direct communication with the viewer. These works can be extraordinary, and the artists who produce them very talented. They are not data visualizations.

Most often, these works use data of some sort as a replacement for a procedural noise, which is a clever technique and certainly has a place in some artworks. However, the artists producing these works should correctly describe the usage of data within their works as a replacment for low-level noise, not as a main driver of the visual effect. This would, of course, be solved by clear documentation as well.