Updated: Jan 16
Artificial Intelligence, or A.I., rages on in the world of visual art, it’s current ability to automatically compose being the new apex of its functionality.
While A.I. continues to make progress at cognition, its most exciting influence on image making has now moved to showing what it “knows” about methods used to combine parts into arrangements that make a whole, that is, composition.
It presents a challenge to human image-makers who have placed high value on their own use of methods because of either the difficulty of mastering them or the uniqueness of applying them. Getting past that challenge is comparable to the permanent effect of industrialization in creating the default expectations of a modern life.
A.I. makes both difficulty and uniqueness nothing more than legacy conditions that henceforth will be merely choices. Meanwhile, mastery of a method has always required two things: understanding why to do something a certain way and rehearsing the action to refine its consistency on demand.
Because A.I. is a function of computers, it makes the rehearsal part trivial – the computer can practice the same thing millions of times a minute, record its own variances, systematically avoid the variances, and reach virtually perfect consistency in exactly the same way that factory automation became viable and obsoleted the need to build cars by hand. Computing is constrained only by the scope and volume of tasks that is demanded of it. And against challenging workloads, the computing answer can usually be simply to increase its brute force.
But methodology, too, can account for higher or lower labor efficiency in working on progress. As part of computing, programming is mainly about defining the way to achieve acceptable efficiency levels against any inclusions of useful variances and/or inconvenient exceptions. Programming supports methods.
The mystery of an artist’s method has always been attractive based on the appreciation of it being exceptional both as labor and as strategy – as action and as idea. The combination is what underpins our sense of it being important, in effect, of being valuable.
But for conventional image-makers in the “fine” (non-utilitarian) arts, one of the most challenging things about A.I. is that A.I. as a production tool does not inherently care why a method is meaningful. A.I. only cares about whether a stated goal correlates well with the choice of method. If that goal is not given to the A.I. factory, A.I. doesn’t even decide what goal to pursue. This can result in dazzling A.I.-generated outputs that are still felt to have very little meaning – like an invention that no one knows how to use.
Artists, however, may frequently go into process, without having made a decision beforehand about their goal. Instead, through experimentation, they discover that a technique allows something identifiable when the technique is incorporated into a method. Whatever amount of time and labor is spent doing this is something presumed to be an inalienable right of being an artist.
But for A.I., this level of effort (experimentation) is just an action that it executes and can keep track of, “mindlessly” and easily. As it also excavates repeatable patterns from its repeated execution, A.I. builds a collection of known actions that produce known effects – effects which can be intentionally combined and arranged.
In short, A.I. can readily generate designs but it doesn’t inherently care what they mean. Meanwhile, the main difference between what an artist considers to be a design and a “composition” is not found in the artifact showing the design. The difference constituting “composition” is found in the artist’s intention to have the arrangement psychologically affect a user (listener, viewer, whatever) to invoke a “meaning”.
Similarly, A.I. itself is programming that also pursues a goal; but the difference between “artificial” intelligence and an artist’s “natural” intelligence is found not in the output of A.I. being successful. Instead it is in the artist’s knowledge of how outcome derives from using the A.I. output. The user – an artist – makes a decision about what outcome is a goal. A.I. simply performs some or all of a method to realize that artistic decision.
Clearly more than just by analogy, A.I. is an “instrument”. Operating A.I. can generate effects that are already compelling in some ways, but those effects are literally instrumental and contribute to a further intent—the performance (execution) of composing.
With that understanding, we can place A.I. in proper context going forward; just as composing may be predisposed by using certain instruments, composing can be done specifically for certain instruments.
Instruments affect the decisions made in composing through a process called orchestration. A given version of a composition may get modified into another version if the composer decides to render it with instruments different than those previously used.
For example, the “same song” can be performed using a piano or a guitar or a clarinet or a voice, and if the instrument is specified beforehand, the arrangement of the composition that is accepted can be clearly different from what would be designed for a different instrument. Orchestration marries the instruments and the desired effects, for the arrangement (design) of the composition.
Showing this in practice, visual artists are increasingly using A.I. as their technical tool to find and apply methods to use as their instruments for something further – composition.
But because of that, there is a popular (casual) perception that A.I. imitates artists. While the recognition of A.I. as an instrument is nearly intuitive for an artist, the inclusion of A.I. in the production of the artwork is still causing confusion among art users (audiences) about why the work produced with A.I. has value.
As already mentioned, value is customarily attributed to artwork because of an appreciation of the distinction of how it causes a meaningful experience.
If A.I. is perceived to eliminate the distinctiveness – by sheer automated repetition destroying uniqueness, or by production efficiency minimizing the labor – then it faces pressure to generate something else compelling on its own.
So far, most of that A.I.-based value is in the drama of surprise – namely, A.I.’s producing the unexpected, or doing it unexpectedly.
A.I. can “create” renderings of imaginary states such as bringing a deceased person back as an apparently lively actor in a “real” event. The enormously high degree of detail used to create the illusion– aka the depth of specificity employed – “fakes” the overall appearance of reality, imitating an “actual” condition.
The “intelligence” of artificially creating a fake instance of an actual condition is far from being unfamiliar to artists. Counterfeiting is a longstanding reality in the artworld, and the “intelligence” to do it is itself an appreciated achievement and characteristic of the counterfeiter.
But A.I. is able to counterfeit – and imitate -- more and more actual conditions that can be imagined beforehand – a so-called virtual reality.
And as A.I. production of items becomes increasingly familiar, devaluing the product on the basis of its A.I.-based methodology will be increasingly rare, while humans’ decisions about what to do with it will naturally become the basis of attributing more value to it.
We get to raise the question: would A.I., without human involvement, ever reach a point where it generated a new defined form to a degree the equivalent of the invention of cubism, or of jazz, given what it had already worked on?
A.I. in art will, in other words, be subjected primarily to the same question that all art faces regarding its production and provision: Why make this thing this way? And the answer will be:
the artwork’s target experience ...
relies on the characteristics of ...
effects that ...
were made available ...
in a way distinctive to ...
what instruments are used ...
in the method of composing it.