Art vs. non-art
Exploring the boundary
I recently found myself musing about one aspect of my recent work, and various ways in which it was or wasn’t artful. One surprising result of this process is some thoughts about the role of AI as an art critic.
(I have previously speculated about the nature of art and the role of being an artist: Art and Purpose, Normal Art vs. Revolutions, Multiculturalism, Universality, and Audiences, part 1 and part 2, What Makes Art?, and Bad Art.)
As part of my current job, I construct a weekly report. That report requires attention to detail and is carefully crafted to be a concise, accurate, and useful summary for its readers. It’s clear to me and to everyone else that this report is not a work of art, but it does feel as though it is artfully done.
If we want to think about what art is, one possible starting point is to consider what it would take for something like my weekly report to become art.
Obviously not art
There are three immediately obvious problems that would need to be fixed.
The first problem is that my weekly summary requires a great deal of background in some esoteric company-specific issues, and so for most people it is simply not interesting. That’s not a concern for me as its author, because the currently intended audience does find it interesting – and, in fact, its audience has grown considerably inside the company. But its inaccessibility is a factor that keeps it from being art.
A second, related, problem is that the summary does not address anything that is of interest or value to most people. Indeed, unless you were a technical leader or a business executive at my company, it is hard to see how you would get anything from my weekly report. So even if you happen to have the technical background to understand it and perhaps appreciate it, the summary is still not likely to be of interest.
The third problem is that the summary is of only transient interest. Somewhat like a newspaper, it is of great significance when it is first distributed, with a sharply declining value thereafter. It becomes almost irrelevant when the next week’s summary is sent out.
So we’ve identified a kind of “barrier to entry,” a kind of “lack of value,” and a kind of “short shelf life” with respect to my weekly report. To put it differently, the immediate problems are that it is accessible only to a very small audience of specialists, it is interesting only to that same small audience of specialists, and its effective lifetime is quite short.
Declaring the problem away (not)
It is tempting to take a romantic attitude and say that what really matters for art is the feeling or intention of the artist. But I don’t think it makes any difference how I, as the purported “artist,” feel about this weekly report. My feeling or intention is not, in itself, sufficient to shift this clearly non-art entity into the category of art. (There is an entertaining mirror image of this kind of declaration in the work of a Boston street artist who uses the logo “NOT ART” to create art.)
Likewise, it is tempting to take an iconoclastic attitude and simply declare items to be art, in the spirit of Marcel Duchamp and his iconic Fountain. Notwithstanding the success that Duchamp and similar artists had with their effort to redefine art, I don’t think it makes any difference for me or others to simply declare that my weekly report is an instance of art. Or if it is successful, it does so at the expense of destroying the utility of the category “art” – if everything is art, then why bother using the word?
Hypothetical transformation
If we look past those problems, there is something vaguely artistic going on in my weekly report: each element of the summary is carefully crafted, with a goal of being clear and impactful while also being concise. There is a voice to the summary: even when I want to incorporate some information from another person’s writeup of a problem, I rewrite it so that it matches the usual voice. If the summaries were only about providing raw information, there would be other ways to accomplish it that would be easier for me as the writer but harder for the audience. Samuel Johnson’s aphorism that “what is written without effort is in general read without pleasure” seems appropriate here, and relevant to my claim that the summaries are at least adjacent to art… even though they are not art.
We could imagine that there is an initial step of assembling the facts, followed by the step of writing the summary, then finally the step of distributing the summary. (My actual process is more complex, but not in ways that seem germane to this analysis.) The process of transforming the mere facts of each incident into a pithy but accurate piece of text seems artistic in nature, even if the actual result is clearly not art. In principle, I could take those same writing skills and produce art – but at a bare minimum, I would clearly need to be writing about something that is more accessible, as well as writing something that more people will find interesting for a longer time.
Fixing the evident problems
Suppose now that I have somehow made these changes to the report: I have produced something that is more accessible, as well as being interesting to more people for a longer time. A first question is, have I necessarily produced art?
The answer seems to be pretty clearly, “no.” The transition from non-art to art requires an increase in accessibility, interest, and lifetime, but the degree of such change is unclear. In addition, it’s quite possible that other ingredients are also required to produce art. We can see that improved accessibility, interest, and lifetime appear to be necessary, but they may not be sufficient.
Further improvements?
Let us now adjust the situation again, so that we now agree that I am indeed producing art. Will I make that art somehow better if I continue to make it more accessible? Will I make that art somehow better if I make it even more interesting or interesting to even more people? Will I make that art somehow better if I make its lifetime longer?
I think the answer to all these questions is, “it depends.” There are circumstances in which making my work more accessible will make it better, but there are also circumstances in which making my work more accessible will not make it better – and might even make it worse. Likewise, making the work interesting for more people may make it better, but there are circumstances in which making it interesting for a wider audience does not improve its quality.
Artistic merit or artistic quality is not a matter of majority vote or market share. “Clickbait” is the relevant term of abuse for online content that attracts interest but lacks integrity, and anything that has turned into clickbait has failed as art.
Assessing artistic merit via pattern languages
The previous sections treated art vs. non-art as a simple yes/no assessment, but an alternative approach would be to think of a spectrum of artistic merit. We could perhaps say that all entities fall somewhere on this spectrum, and that we are simply considering ways to nudge a given entity higher on the spectrum. If we take that position, we might then be interested in ways of assessing or measuring this quantity.
To consider the question of artistic quality, it’s tempting to think in terms of Alexander’s concept of a pattern language. One attraction of a pattern language is that it can be used for both analysis and synthesis – for both assessing the merit of a design and for generating a design. Another attraction of a pattern language is that it offers a crude measure of comparative merit for two designs: all else being equal, one design is superior to another if it includes more distinct patterns.
We might reasonably ask whether this paradigm could be applied to the question of artistic merit. We would not be using Alexander’s original pattern language for towns and buildings, nor would we be using the many Alexander-influenced pattern languages for software engineering. Instead, we would be creating (or, for the moment, imagining) an artistic pattern language. This language might be as crude as one of the many surveys of art that identifies tropes, themes, or categories across a wide spectrum of art history. Could we realistically assess the merit of a work within such a framework?
My current answer is that I think the work could be performed, but it would primarily serve to confirm that the approach is not an accurate measure of artistic merit. We don’t necessarily have much in the way of first principles or an underlying structure for what people do when they judge art.
AI and artistic merit
An intriguing aspect of the brave new world of AI is the possibility of developing machinery that assesses artistic merit. We can certainly see the outline of such an approach, and it’s entirely possible that people are already working to realize it. The first step would involve training a model on a large body of existing art history, which would allow it to then behave as a kind of embodiment of conventional wisdom. It would then be able to express opinions on a new artwork.
A crucial question, here as in other areas of AI, is what level of authority to grant to those judgments. A clearly erroneous approach would be to treat this art machine as the final word, some kind of superior and objective assessment. It would be far wiser to treat it as an additional observer/critic, comparable to one of the authors of encyclopedic surveys – an interesting and idiosyncratic perspective, but by no means infallible. After all, it “knows” only what it was trained on, and the basic method of a generative AI system has been usefully (if disdainfully) compared to “stochastic parrots” – the output is consistently plausible given the training data, but is not necessarily true.
Such an AI critic might make for interesting new creative processes, if it were asked to evaluate the current state of a work being constructed. Amusingly, we can anticipate that different artists would likely use such a tool in very different ways. Some artists might take the AI suggestions. Others would intentionally do the opposite. Still others might toss a coin to decide whether to take a suggestion or do the opposite. Some artists might choose to incorporate the AI critic into a larger creative process of controlled chance, such as oblique strategies.