AI Art Must Not Be Art
“I am sure I could not light a fire.”
“You are an artist, are you not, Mr Dedalus?” said the dean, glancing up and blinking his pale eyes. “The object of the artist is the creation of the beautiful. What the beautiful is is another question.”
He rubbed his hands slowly and drily over the difficulty.
“Can you solve that question now?” he asked.
“Aquinas, answered Stephen, says Pulcra sunt quae visa placent.”
“This fire before us,” said the dean, “will be pleasing to the eye. Will it therefore be beautiful?”
“In so far as it is apprehended by the sight, which I suppose means here esthetic intellection, it will be beautiful. But Aquinas also says Bonum est in quod tendit appetitus. In so far as it satisfies the animal craving for warmth fire is a good. In hell however it is an evil.”
I deny the reality of the significant event which has marked the internet so solemnly in the last weeks: I deny the reality of ChatGPT.
Much like the show-trial of Hitler that Joseph Roth was describing, which ultimately did occur; ChatGPT does in fact exist. Like the trial of Hitler also, moreover, its procedure is fallacious and its resulting state-of-affairs—Hitler’s boosted popularity and eventual freedom, the misconception that a computer can create art—are dangers to society. Yes, I deny the reality that ChatGPT, or any other artificial intelligence program for that matter, is able to replicate human artistic output. Further, not only do I deny that artificial intelligence is able to make art—that AI Art is not art (as has been well-argued by Erik Hoel)—I argue that AI Art must not be art.
To the extent that art is a human endeavour, ‘AI art’ is not art; rather, the term is an impossibility. Nay, it is a possibility only by making worrying, dehumanizing concessions. For software to be able to produce art, computers must be equated to humans—man equaled to machine—by one of two, or the combination of two, broad reclassification procedures.
For AI art to be considered art, either Man must be lowered to the level of Machine, or Machine elevated to the plane of Man (or both). Those that allow AI art to be art must either believe that humans are so base to be machines, or that machines contain all the necessary constituent elements to be human. Both approaches contain distinct though equally unsettling dangers.
It bears separate discussion, though it is my opinion that modern man is mostly falling for the former folly. For, we have allowed ourselves to be reduced to the functioning of machines, of glorified automatons; we are slaves to the technology that was meant to serve us. For instance, even the Turing Test that ardent technologists deem so significant in measuring computer capability only professes to test whether machines can merely demonstrate human intelligence, and in reality succeeds, never mind often fails, in measuring a relatively small and often insignificant part of total human knowledge and experience.
I would go so far as to rate ChatGPT’s parent organization, OpenAI (so aptly classified as a non-profit—though recent acquisition rumours counter that label), among the most dangerous threats to our society. The charge is serious, though the attempted offence demands it: they have created a weapon of mass destruction, which, if successful, means to usurp humanity of its unique singularity—the creation of art—no matter by intention or by accident as a side-effect of making their devilish dreams a reality. Art is the extension of the human mind, the human hand, the human spirit, the human heart, in short: the human soul. To attempt to replicate the capacity to create art is to try annex the essence of human experience.
For, OpenAI wish to “create software that outperforms humans at most economically valuable work for benefit all of humanity.” In truth, the only way to achieve the latter—the benefit of all humanity—will be to destroy the former, humanity itself. It is like a dictator who says his mission statement is to nourish humanity but that they can only do so by conquering the world and imprison humanity only so that they can be sure to provide three meals a day to everyone. To be human is to rather starve.
And so by the deployment of artificial intelligence systems guerrilla bands such as DALL-E and GPT and Mindjourney, which are really only the beginning—the technology remains in its infancy—artificial intelligence companies attempt to neglectfully persuade that human peculiarity is irrelevant, that human enterprise does not matter, and that in the end, we are no better—nor worse—than machines.
The battle is already being waged, and humanity is losing. A few months ago, a computer won first place at the Colorado State Fair Fine Arts Competition. To make matters worse, the credit for this computer’s victory was claimed by a local resident named Jason Allen (who also goes by the fitting nickname “Sincarnate” in the shallow echo-chamber he employed to brag about his sinister achievement). In his defense, Allen argued that he is the one that came up with the prompt to tell the computer that yielded such a result, that the computer was a mere tool, just as is a brush or a pen.
Like a calculator winning an elementary math competition, whose user might claim they only “typed the numbers in,” Allen claims that he didn’t break the rules. Except, he broke the only one: he is not the creator of the work. And, the work is not art—it was not created by the human hand. Yes, a computer won an art competition and humanity lost.
What it boils down to is that computers can’t be humans, and that humans are not computers. If we allow computers to achieve a human function, we reduce the sacredness of that function. Those who see no distinction between human art, and AI art, see no ultimate distinction between humans and machines. To argue that computers can't be humans is to protect what it is to be human. I mean to protect the sanctity of humanity that AI posing as artists threatens.
An initial thesis was that AI art is not art. While I’m more than happy to defend that claim, it leads to arguments that are unrelated to the crux of the issue; the whole ‘did the bullet or gun or finger that pulled the trigger kill the man that had his head blown off by a rifle.’ Just as the law decided that it must be the man that shoots the gun that must face the consequence of its operation, so we must decide that software must be incapable of human achievement.
So it isn’t only that AI art is not art, It is that it must not be, for that AI Art to be art art is to hinder our existence; to allow the robbery of our most valuable possession, our greatest invention, our supreme and unique method of expression. And any organization or software actively campaigning for such is a threat and terror to our race. We must not treat our conquerors as saviours.
As a human, I will defend what is undeniably and undefinably human, lest we have anything left of our to defend. As Roth writes somewhere else, “Even art can have inhuman consequences.”
Only if we allow it to—only if we don’t see ourselves in Hell.James Joyce, A Portrait of the Artist as a Young Man (1916)
Joseph Roth, The Dream of a Carnival Night (Vorwarts, 2nd March 1924)
Originally: “I deny the reality of the significant event which marked Germany so solemnly this week: I deny the reality of Hitler’s trial.”
Joseph Roth, The Myth of the German Soul (Das Neue Tage-Buch, 12th March 1938)
A well-argued position, and one I'm still grappling with. We live in interesting times, and it seems that we'll need to keep our philosophical wits about us.