AI-Generated Art and the Vestiges of Romanticism

AKA "Some Observations and Predictions on AI-Generated Art (Part 1.5/2)"

“Blick auf Dresden bei Sonnenuntergang” (1822) by Carl Gustav Carus, whom Ludwig Klages admired as a great romantic thinker

I. Introduction

In my last discussion on how AI-generated art will effect things in the future, there was one major recurring theme that perhaps I could have spelled out a bit more obviously. It is that the changes which AI-generated art will bring are not going to happen overnight. AI art will bring some to the surface and reveal them to everyone, but these will have been changes that were going on for quite some time. What seems like a rapid shift will have been borne out through a process that was going on undetected for decades. And while plenty of men are predicting that AI will bring about apocalyptic changes to everything, not just the arts, on account of its violent and chaotic Promethean fire, I see it otherwise. As Thales says to Anaximander in the “Classical Walpurgis Night” of Goethe’s Faust,

Never was Nature and her fluid power
Indentured yet to day and night and hour
She shapes each form to her controlling course,
And be the scale immense, eschews all force.

In other words, nature likes to take her time via slow processes. And even when a great cataclysmic event occurs (this certainly does happen), its effects are often only felt gradually, as secondary processes massage people into complacency, making them unaware of any real change until perhaps they find themselves in a moment of contemplation long after the fact. “Boy,” a guy says, “That event sure was important, wasn’t it?” Sometimes, in the world of science or art, an important year isn’t understood as such until decades later.

There’s no doubt that AI-generated art is causing people to question what “art” means, but it’s also worth bearing in mind that “art” is a concept that has never remained stable for long during history. And so in this discussion, I’d like to focus on how AI-generated art will modify a mentality pertaining to art that has never quite gone away from us. I’m talking about romanticism, the intellectual movement of the late 18th-19th centuries in which philosophers and artists waged war against the mechanistic outlook of the baroque period and enlightenment.

People with romantic inclinations appear to be the most opposed to AI-generated art and deeply concerned about what its effects on culture will be. And they have good reason to feel this way, since the idea of fully-automated entertainment seems a direct threat to the anthropocentric and naturalistic view of art that romanticism holds. I also should stress that while no epoch ever truly ends, we hold onto more of the romantic period than any other in how we consider the value and function of art. At least here in the west, our intellectuals, critics, and especially artists have been in the same post-romantic paradigm for over 150 years. Maybe I will justify my stance on this a bit later. But in light of this clash between the romantic sensibility and AI, which I’m seeing as a bit of a conundrum, I’d like to make a bold prediction comprising two sub-predictions, which I’ll go on to discuss:

AI art will badly discredit one aspect of romanticism while immensely strengthening another.

1. It will discredit the romantic view of imagination as the most important of man’s mental faculties.

2. It will ultimately bolster the romantic inclination to see the artist’s life as his greatest work of art.

II. Genesis of Imagination as The Important Thing

Let’s start with sub-proposition one. “Imagination” is a fairly complex topic, so it’s worth having some understanding of why the romantics got so keen on it.

Immanuel Kant was the guy who philosophically planted the seeds that allowed the imagination to begin its reign as the most important mental faculty in his Critique of Judgment, since he maintained that the aesthetic sensibility is not a cognitive process and is thus free from any cognitive or even moral determination. The idea that the aesthetic sensibility isn’t in any way cognitive might not seem too interesting to us now, but it was a big deal back then. Cognition, for Kant, is how man strives toward what’s morally righteous. But Kant also didn’t see the aesthetic as the same thing as animal satisfaction, like how a dog will want to chew on a bone. For Kant, the aesthetic was something only humans can have because it’s what lies in the nexus between animals and God.¹ And, because of the aesthetic sensibility’s distinctness as a category, imagination, its natural counterpart, can allow man to take what’s in nature and conceive of it as another, altogether different nature, unbounded by the physical laws of our earthly nature or the (essentially cognitive) moral laws of practical reason. Essentially, the imagination can allow man to create his own internal universe.

The implications of this are potentially vast, but the problem is, Kant wasn’t really into thinking much about artistic creation, and he never really pressed the idea that imagination could overcome the division between the sensuous and moral aspects of man’s experience. He just sort of left all of this open, hanging out, and didn’t return to it. In fact, there’s a good chance he wouldn’t have liked the way the romantics extracted his ideas and built upon them.

History, however, doesn’t care for the feelings of those who have left a legacy. So when early romantics Friedrich Schiller and Novalis showed up, they were quite interested in imagination as a creative force, and they picked up a lot of the loose ends that Kant had left there. To orient the imagination toward the production of art, they distinguished between a primary imagination and a secondary one. The primary imagination is at the root of consciousness and largely operates involuntarily, picking up various experiences and synthesizing them, determining our perception by ordering our engagement with the world into something understandable. The secondary imagination, then, consciously creates imaginary worlds by combining elements from past perception into new forms. For Novalis, imagination is superior to reason, since it bridges sense and thought, allowing for the possibility of consciously understood experience. And the poet, in his view, is perfectly attuned to nature’s cosmic harmonies; his imagination is free and fully self-possessed. For Schiller, the creation of a “messianic kingdom” is really an aesthetic project for humanity, realized through imagination.²

III. Imagination as More Than Combining Things

Currently, we’re at a place in which the AI soothsayers are predicting some sort of future in which entire AI-generated universes can function as surrogate realities for unproductive proles whose entire existences have been rendered worthless by automation. A virtual solution to a “dire problem,” one might call it. So if it’s true that AI-generated art will allow people to escape into its own sort of self-contained world for sustained periods of time, wouldn’t that imply that computers can have an imagination?

Well, not necessarily. What the romantics contributed to our understanding of imagination is that it isn’t just the part of the mind that combines things. As of now, AI works much like how the British empiricists (e.g. Locke, Hume) understood the human mind to work. It takes in information like an inert block of wax onto which various things imprint themselves, then spits back out some combination of them depending on the content one requests from it. The romantics saw the imagination differently. Schiller argued that a true work of art isn’t subordinate to nature but in fact overcomes it. Otherwise, it will merely be a mask. Probably the most representative commentator on this subject was Samuel Taylor Coleridge, who regularly compared the imagination to a living, growing natural organism.³ In the same way that plants grow via sunlight and water, and in turn give back to the environment by producing oxygen, the imagination for Coleridge works analogously. It isn’t enough to say the imagination analyzes information and then re-presents it. It instead goes a step beyond: synthesizing information and presenting it anew, revivifying it, restoring it in such a manner that feels as if we’re experiencing it for the first time. It therefore cannot be a product of anything that lacks the mysterious spark of life we find in living beings. As Meyer H. Abrams explains in his The Mirror And The Lamp,

It is astonishing how much of Coleridge’s critical writing is couched in terms that are metaphorical for art and literal for a plant; if Plato’s dialectic is a wilderness of mirrors, Coleridge’s is a very jungle of vegetation. Only let the vehicles of his metaphors come alive, and you see all the objects of criticism writhe surrealistically into plants or parts of plants, growing in tropical profusion. Authors, characters, poetic genres, poetic passages, words, meter, logic become seeds, trees, flowers, blossoms, fruit, bark, and sap.

But the romantics run into a problem. What AI can produce is convincing enough to the average person to make it seem like a living organism produced it. If people are given a test, most will not know the difference between an authentic, human-made poem and an AI prompt imitating one. Same goes for paintings now, provided that we’re looking at the painting through a computer screen (and in this lifetime, the vast majority of paintings we’ve seen have been this way). We’ve reached the point at which when we look at an unfamiliar work of art accessible through a computer, with no additional context there to tell us anything about the artist, we really can’t tell if some lines of code created it independent of the coder’s knowledge.

The romantics could not have foreseen such a situation, and it shows in the vagueness of the language they use to describe how imagination actually shows itself via artwork. If you consider poetry, for instance, it’s hard to say, “Well, name some characteristics of a poem that will show the mastery of imagination over nature, not merely the combination of various elements in it,” since just about any poem can be interpreted as a process of combining things. Fortunately, we have some specific ideas from the contemporary thinkers. Johann Jakob Breitinger argued that the poet can create “entirely new beings” through the power of his imagination, which might include personifying abstractions, animating inanimate things, humanizing animals, or materializing the metaphysical world of spirits – all his examples. And as it happens, artificial intelligence can now do pretty much all these things, both poetically and visually. All you’ve gotta do is type in some prompts.

IV. The Reliance on Extrinsic Info to Determine Imagination

The problem with romantic literary and aesthetic theory is that inevitably, the critic would have to assess the imaginative life of a poem by taking extrinsic information into account. The critic would know that a poem was deeply imaginative by way of how it corresponded to the poet’s life. And one would know the poet had a valuable life by way of how he composed his poems. One can see the circularity. Perceiving imaginative presence in an unfamiliar artist, then, becomes a bit like Meno’s paradox: I’m looking for signs of imagination in a work of art, but I don’t really know what I’m looking for because I know nothing outside the text.

Let me put it this way. Let’s forget about all this AI stuff for a minute, and I hand you what I’m presenting as a 13th century English chivalric romance poem written by an anonymous author (as most were). I ask you, “How imaginative was this guy?” And you, poor sap that you are, can’t rely on any biographical details to back up your judgment. It would seem that all you could really do, at that point, is form a judgment by making comparisons to other imaginative artworks from earlier on — other chivalric romances, if you’re lucky enough to know any. You can thus see where the artist stands within the tradition that preceded him and suss out what parts of his work imply some real ingenuity and which don’t. Is this a solution? I don’t think it’s a solution, because we’re now in T.S. Eliot territory. We’re now talking about the individual in relation to a tradition, totally ignoring the artist’s requirement of being unique and authentic and in doing so creating an alternative universe. We’re essentially ignoring all the stuff the romantics really cared about. Is the artist expressing his true emotions, as Wordsworth was really concerned about? We can’t tell. Is he engaging in some sort of creativity that inherits a tradition while subtly modifying it? Maybe we can tell that, but the signals are weaker than we’d like them to be, because what might seem like “ingenuity” could be a lucky accident, that is if we’re just looking at one work. And if you’re having this kind of trouble with an actual thing that was written, imagine if I then told you, “Oh, a computer wrote it.” You would have no choice but to conclude that the imagination is just really, really hard to perceive, enough to where it can’t be the dividing line between man and machine.

But the problem goes away if we’re discussing an artist with a distinctive, inimitable style. And one can only perceive a style by comparing an artist’s various works within an oeuvre over a distention of time. If we know who the artist is and how he lived his life, we can look upon every work as a reflection of what he was going through at that exact point. And of course, once we get a sense of who he was, then it is all the more difficult for AI to fool us. A couple years ago, a guy made a YouTube video quizzing everyone on if they’d be able to recognize the difference between a J.S. Bach piece and an AI imitation. Most in the comments guessed right, as did I. The software has surely gotten more accurate since then, and to be fair, there are plenty of boring Bach four-part chorales for organ that I suspect AI would have no trouble imitating. But I’m not holding my breath waiting for AI to do anything as brilliant as The Well-Tempered Clavier or Art of Fugue anytime soon.

But my point here is that the romantic conception of the imagination, at least applied to art criticism, ultimately rested upon what the critics could gather from what they knew of the artist. This is why Coleridge had no problem at all with making frequent reference to the biographical details of all the poets he discussed, approaching everything through a psychological lens. Other romantic critics like John Keble made conjectures about poetic activity that in many ways predicted Freudian psychoanalysis. Occasionally, when romantic critics ran into authors about which we know almost nothing, they’d simply make things up. Keble made plenty of assumptions about Homer based entirely on his poetry, ignoring that he was just one bard transmitting an oral poem that survived through the Greek Dark Age, thus making it impossible to determine what from him was an original contribution.

Shakespeare, by the way, was a huge problem for the romantics, because so little is known of his life. There were two camps that emerged: one thought that his incredible skill was in keeping his personal self totally aloof from his poetry, overseeing everything with Godlike objectivity (this was Coleridge’s view). The other thought that he was deeply, intensely personal in everything he did, suffusing every single line with some aspect of his inner being. This is sort of like immanentism vs. transcendentalism debates that metaphysical monists have with one another, like the Trika Shaivas and the Advaita Vedantists. Either all the poetry was Shakespeare in its essence, or it was maya, veiling the hidden cosmic truth that is Shakespeare. This is the sort of thinking you’re left with when you’ve got nothing outside the text to go on. Eventually in the 1850s, various writers squared the circle by deciding that Shakespeare was actually Francis Bacon. Ah! What a relief. Now we’ve got something to chew on. And it should go without saying that pretty much all anti-Stratfordian theorizing today is in the post-romantic tradition. Why can’t Shakespeare have a nice, thick biography? Why can’t he be like that nice boy Milton?

V. Wrapping Up

This should be my cue to switch to a discussion on the romantic tendency to see the artist’s life as his ultimate artistic achievement. But I’ll just finish up quickly with some concluding words on the imagination. In order for the romantic view of the imagination as man’s core mental faculty to prevail, it would have to be reflected in the art everyone appreciates. When modernism came along, quite a few artists deliberately tried to demystify their own aura within society. The pop art movement is but one example. Another is the use of randomness in music composition, as with John Cage and others. Artists were now into the idea of suppressing their own voice, and their imaginative contributions along with it. Now, all of modernism truthfully can’t be reduced to just one or even a few things (which is partly why I think we’re collectively still essentially stuck in a post-romantic way of thinking), but there’s no question that it helped form the perception that the artist is not always the psychopomp of his own self-conceived otherworld. Sometimes he’s just a dice-roller, sometimes he strikes the pose of a neutral observer, sometimes he’s a trickster whose “art” is frankly whatever he can get away with.

Of course, one can still see imagination (primary imagination) as the thing that allows for consciously understood experience to occur, and in fact one can theorize independently all sorts of things about it. You’re free to do whatever you like! And besides, faculty psychology is underrated anyway. But I’m talking about the hoity toity types. You know, the smart set. The people for whom the category of the aesthetic is no small matter, even though they’ve mostly alloyed their concern with other ones. At some point, AI is going to compel those people to confront the question of what separates the artwork of a genius from that of a computer. And while imagination will play some role, certainly, I don’t think it will enjoy the same emphasis that it did during the romantic period.

I gotta run.

Next time I’ll talk about why I think we’re still stuck in a post-romantic paradigm and why AI-generated art will bolster the romantic tendency to see the artist’s life as his ultimate work of art.

1. I’m mostly getting this stuff on Kant from a couple books. One is Christian Wenzel’s An Introduction to Kant’s Aesthetics: Core Concepts and Problems (2005), the other is Jane Kneller’s Kant and The Power of Imagination (2007), which I only sort of skimmed.

2. Check out this conference paper for further details.

3. Pretty much everything on Coleridge, Keble, and Breitinger is taken from the excellent The Mirror and the Lamp by Meyer H. Abrams (1953).