The final vector by which time generates symbolic value in a work is potentially the simplest. Prior to the proliferation of generative AI tools, it was a natural assumption to make that any work of art was the product of hours of labor. It was technically necessary for the artist to spend considerable quantities of their precious time to produce a work, and the fact that someone did so would at least suggest that the work they produced may be of interest, merit, or value. With generative artworks, this assumption is no longer reasonable. Time is no longer required in the production of an artwork, and therefore someone creating an artifact and sharing it for your consumption no longer carries any implicit suggestion that this must be a work of some value. This is not merely a marketing concern for generative artworks; it poses a very real barrier to their exegesis.

The film concept of mise-en-scène is one that has counterparts in all other artistic mediums. There is a certain maxim of artistic production that could be simplified as “if something is in the frame, it has a reason for being there.” This is true to varying degrees in different works, but as an assumption, it is the basis of how we can have conversations about the symbolic content of an artwork. The fact that every word on a page was chosen by a person, that every brush stroke was a micro-decision made by a painter, and that a person was there in their studio spending time to produce a work is what makes it a worthwhile pursuit to analyze an artwork. The artist spent at least some time on every detail; they spent time to produce the work that is now in front of you. In the absence of that, the space for critical appreciation would seem to dissolve. One needs to know that their effort in consuming a work has been (p)repaid in the form of investment by the work’s creator, and that the details of its appearance are intentional.

Like most technologies, generative AI serves to speed up workflows, in this case the production of art. However, the scale of change that it inaugurates is singular. It reduces process to the extreme, and “liberates” artists from the direct acts of artistic production, nearly completely. This modified relationship with time carries formative implications for the potential of generative artwork to communicate symbolically. This is not a complete foreclosure of the possibility for AI art that is meaningful. Some artists have emerged who use generative AI to create truly novel and aesthetically provocative work. Instagram creators like @kingcon2k11, @napulensuonno, and @xeocho produce work that is bold, maximalist, and (over)stimulating. They create aesthetics that can only be produced by generative AI, and the works are highly compelling. I recall the examples of early GAN image generation in the late 2010s as being some of the most aesthetically interesting images I had seen in some time.[5] They displayed an eerie alterity that arose specifically from their proximity to photographic reality.

But a lack of aesthetic novelty is not the problem here.[6] These images resulted in powerful but ephemeral aesthetic experiences. The degraded relationship with time made it difficult to identify with these works on any deeper level or to feel as if they succeeded on the level of symbolic communication. The aesthetic provocation was there but not the deeper significance.

A natural criticism of this temporal critique would be that AI art is able to retain a relationship with time through iteration. Iteration is deeply embedded in traditional artistic process and it is a profoundly productive aspect of art-making, so shouldn’t its prominence in AI workflows give the medium a form of temporality? For an artist using an image generation model, the process of prompting an image may progress over time as an effect of the prompter refining, re-prompting, and tweaking towards a final output. This iterative process has the potential to be productive in two ways: the artist can adjust the model’s outputs closer to their pre-existing concept for the artwork, wrangling them to fit their imagination, or they can adjust their own concept of the work through a process of experimentation. The latter is closer to a meaningful temporality of AI art, but it is still distinctly different from what can be found in previous mediums. In the latter case it is the concept of the work undergoing a change, not the artifact itself. In the former, it is a matter of mass generation, and then choosing between the outputs. The act becomes like that of a curator: a potentially creative role, but not the same as that of the artist. In neither case does the work really undergo any change in and of itself. The work, as a final image, has no history to speak of, and appears in a moment. It may have some lineage in the mind of the artist, or as the final stop in a series of trials and errors, but these histories are abstract. They don’t situate the work in time. The Benjaminian aura of the work itself is degraded beyond recognition.

Or maybe one could advocate for an AI temporality of a different sort, like in the work of artists such as Hito Steyerl and Trevor Paglen.[7] Both artists materially adopt the workflows of AI models as a way to get inside of them and to critique them. They use AI in a manner that has a process to it, but it is the process of the machine. The symbolic content of the work relies on the instantaneity of AI tools, or alternately a distention of their processes, stretching and flattening them to make them legible. Here the temporality of traditional workflows is not faithfully mapped on to an AI process, but rather AI process is fully inhabited by the artists, along with its alternative temporalities. These are rich works, but not because they maintain traditional temporality in production of an AI artifact. They are not even primarily concerned with the artifact, but instead the processes that generated it, and the possibility of their critique.