The mangled, uncanny visions of AI generated images have lately become difficult to avoid, and often difficult to spot. I think this speaks more to the intelligence of the viewer than it does to the machine, as I will explore. The human hand continues to mystify even sophisticated models, however. The horrid, surreal masses of ill-defined tendrils that appear where hands ought to be have become an indelible hallmark of AI image generation. As unsettling as these pictures are, I think the case of the incomprehensible hands is a symptom of a more subtle apparatus at play. 

I would suggest that the shortcomings of these models can be viewed productively through the lens of ontology. I was struck recently by the profundity of an otherwise increasingly familiar sensation: the cognitive frustration induced by ambiguous objects, specifically those peculiar to AI images. I stared long and hard at a particular AI image in a vain attempt to discern a single determinate object within the frame. I would even have settled for an indication of a flawed object that referred somehow to a real one, but was left disappointed here as well. 

Critically, an AI cannot be said to know anything, at least not in the same sense that the human brain knows. Any objects that may appear in these images are incidental, the result of cognitive projections onto algorithmic amalgams. That is to say that the image contains nothing but calculatedly random splatters. The final product is Darwinian, selected by particular criteria from a slew of equally random but less fit competitors. The mind of the viewer constructs the concept of the object from this more-or-less convincingly rendered facsimile. 

And it is exactly that — a facsimile. The same incongruities that characterize the now infamous alien hands permeate these images. Lines that have no clear beginning or end, flecks of pigment with no clear purpose, lights and shadows from nowhere, otherwise apparently solid bodies with gradations into nothingness, eerie repetitions and incoherent details all are symptoms of the nature of this process. There are no objects being depicted, only those being guessed at by random flailing of pixels. The pixel is the fundamental object of the image to the “mind” of the machine, it is the atomic particle of digital visuals. Any objects beyond this cannot exceed the status of a mathematically determinable pattern of pixels, and these patterns must necessarily transcend the machine. It is, again, ultimately the mind that finds the objects where there are only specters of such. We can think of these as “pseudo-objects.”

With regard to this phenomenon in comparison to art, a comparison that has been, in my opinion, often made erroneously, I posit that what is often termed “AI art” is really “pseudo-art.” Not only does it depict pseudo-objects, but this fact points to failure to meet a critical criteria of art (previously touched on in an earlier post). The audience bears the sole responsibility in constructing meaning from the primordial data soup of these images. There is no creator but the viewer. I am hardly the first to propose this view of these visual works and I will certainly not be the last. 

These propositions rest on the premise that the artist depicts objects, which abstract concepts can be categorized as as well, that those objects are real in some sense, and the viewer interprets those objects. The machine generates pseudo-objects, which the viewer must ontologically define, bring into real being, by their own power. An obvious implication of this is the self-indulgent nature of AI image generation, in which the assignment of prompts becomes the means by which the viewer generates phony reflections of his own psyche, perverting his role. This is incredibly obvious in the tragically inevitable abuse of these programs for generation of warped, hyper-sexual artifacts. 

But, as that sense of frustration often demonstrates, the mind of the viewer cannot on its own make the objects real. In viewing a work of art, the viewer, speaking Platonically, grasps at the forms via their instantiation in the work. He views the art as symbolic participation in some discrete subject, whether that subject, again, be concrete or abstract. This is made possible because the subject is also present in the mind of the artist. The artwork is the nexus of subjective participation in cognizance of the object through the mediation of its objectively real, subjectively manifested representation. That is, the subject signifies an object that transcends either participant. Because an AI image generation model cannot participate in that shared cognizance, it cannot appeal to a real object. (The unguided paint splashes of artists like Pollock can be considered closely related to this.)

These facts give an impression of the sinister aspect of this species of AI model. The AI generation of any pseudo-artistic media, audible, textual or visual, is inherently a perpetuation of postmodernism into a hyperactive mutation of itself. It delegates all power to the viewer, eradicates the possibility of transcendent objects and thus meaning, and usurps the transcendental values with empty veneers which we might closely associate in varied senses with what Solomon called “vanity” or “vapor.” Beauty is distilled through vapid chaos and consolidated exclusively into the eye of the beholder. Truth and goodness are brought to the piece in mutilated, relativistic form by the viewer as well, if they appear at all. 

Form appears fundamentally visually undermined in these images because objects have been made unattainable by the nature of the images’ creation. These pseudo-art pieces are drenched with the incoherence of unconsciousness that Surrealism perhaps was able to comment on coherently. In this case, the commentary is the incoherence. They are the dreams of unconscious nothingness, books plucked off the shelves of the Library of Babel. I do not mean to say by any of this that AI images are clearly evil, should be outlawed, or are entirely useless — though these propositions are not out of the question — only that they cannot rightly be called art. 

It is in this line of thought that I struggle to perceive any possible noble intention behind such models. What artistic desire or inspiration could drive their development? Pseudo-art of this kind epitomizes within, we might say, the metanarrative of artistic endeavor all of the degenerative effects of individualistic atomization, liturgical reliance on industry, secularism and postmodernism in general. It seems to follow that the misuse of science as a philosophy so prominent in the West would lead to the misuse of engineering as art. These models hold industry and innovation, materialistic progress, as the blade with which they may gut culture of real meaning and value, which is to say, deprive her of reality. Calling their products what they are, pseudo-art, seems to me the best way to immediately dull that potentially deadly edge. 

Every good gift and every perfect gift is from above, and cometh down from the Father of lights, with whom is no variableness, neither shadow of turning.

— James 1:17 (KJV)