Introduction

“So God created man in his own image, in the image of God created he him; male and female created he them.” – Genesis 1:27

Being Christmas this past weekend, and having attended dinner, parties, and watched numerous holiday films, I found it difficult to do the research I have normally accustomed myself to doing. However, I still managed to pick a topic to write about this week, which I think will be fun to explore and an interesting intellectual exercise. This week, I’d like to discuss the topic of AI Generated Art and Literature.

Over the past few weeks, this topic has proliferated through websites such as MIT Technology Review, Venture Beat, CNET, etc., all of whom (and many more) wrote rather broadly on the topic. The articles covered software such as DALL-E and DALL-E 2, as well as ChatGPT; MIT’s Technology Review delved into the workings of AI Art Generators, while other sites, such as CNET, divulged more of the limitations of the software. I will be covering some central ideas from these articles (and a few others) throughout the course of this paper.

Primarily, I would like to try to unravel the question of “Is AI Art actually Art” a little further. To do so, I will begin by exploring a general definition of Creativity as it is understood by Cognitive Neuroscientists. I will then be stepping into the realm of Ancient Greek philosophy by unraveling the meaning Aristotle and the Peripatetics (likely) gave to the concept of Soul, as well as the meaning Plato (through Socrates) and the Platonists (likely) gave to the soul. I say likely as I will be forming a general and working definition of the Soul from both the Aristotelian and Platonic perspectives. In the process, I will try to synthesize the two.  I will conclude the paper by reflecting on these questions: 1. Is AI “Art” Creative; 2. Does the AI have a Soul in the Classical Sense; and lastly, Depending on the Answers for 1. and 2., does the AI actually produce Art?

I hope you find my musings on this topic as entertaining as I found the topic itself.

Is AI Art Creative?

Margret Boden has defined Creativity as an idea or an artifact that is new, surprising, and valuable.

While this definition hits on aspects of the broad definition of Creativity understood by Cognitive Neuroscientists, i.e., Creativity is that which is Original and Appropriate (Abraham, 2018, pp. 7-15), the problem with Boden’s definition vis-à-vis AI Art Generators rests on whether AI Art Generators are valuable.

For example, for who is ChatGPT valuable, and what do you mean by valuable precisely? In what context is ChatGPT valuable? If I use ChatGPT to write an essay, more specifically, if I have ChatGPT write an essay for me, is ChatGPT valuable? If it gets me the grade I want, perhaps with respect to that kind of mental schema ChatGPT is valuable, but with consideration towards my future self, if I have never written an essay, I have used ChatGPT to produce all of my essays, and I have graduated (with a degree in hand) having never written an idea of my own, having never learned to think for myself, is ChatGPT genuinely valuable or is it, as a matter of fact, costly? 

One could also argue that at first glance, and only seemingly, may we call the work produced by an AI Art Generator, like DALL-E and DALL-E 2 new and surprising, but after spending time with DALL-E and DALL-E 2’s products, we may find that they very quickly lack the je ne se qua of a creative product. In part, this is the result of the AI’s programming.

In short, the AI simulates both divergent and convergent neural processes (Jones and Estes, 2015; Beaty, Benedek, Kaufman, and Silvia, 2015). For a start, AI Word and Image generators simulate the Default Mode Network (DMN), which is “associated with cognitive processes that require internally-directed or self-generated thought, such as mind-wandering, future thinking, perspective taking, and mental simulation” (Beaty et al., 2015). Essentially, this naturally occurring DMN is analogous to the process produced by DALL-E’s ability to generate multiple images from large data sets such as LAION. AI Art Software allows you to “type in a short description… [providing you] a handful of images that fit [your prompt]” (Heaven, 2022). Via the prompt, the software collates a series of images related to the input (prompt). This prompt acts as the executive network directing the pseudo-DMN – which may be a cause of imagination (Carroll, 2020). The Executive Control Network (ECN) “is engaged during cognitive tasks that require externally-directed attention, such as working memory, relation integration, response inhibition, and task-set switching” (Beaty et al., 2015). The prompt acts as a pseudo-ECN; the prompt filters relevant from non-relevant images the software has been programmed to associate with any term provided. The AI then, using the pseudo-DMN, collates, integrates, and generates products defined by the pseudo-ECN (prompt).

This does generate a product that may appear new, but as users of the software may have quickly realized, the newness or surprise of the software will become outworn. This is because the software creates products or responses based on what has been fed to it. ChatGPT, for instance, simply “recognize[s] patterns in vast swaths of text, harvested from the internet” (Shankland, 2022). Then “with human assistance” provides more “useful [and] better” responses. Of course, what is useful and what is better are subjective. According to Maria Theresa Llano, “perhaps machine learning will only ever produce images that imitate what it’s been exposed to in the past” (MIT Source). In other words, Boden’s definitional standard is not met. The work is not new and surprising, in fact, it would be completely accurate to call it derivative.

AI generators, such as DALL-E, have also been compared to a Film Crew (Dujmovic, 2022); i.e., they are a product comparable to a movie. However, I think this analogy is flawed. In so far as the comparison goes, AI Generators are like Movie Crews, while the prompt is like a movie director or script. The writer is like the scriptwriter and the director. The director gives his instructions to his crew, and they assemble the art piece. However, the director also has creative oversight, can micromanage his crew’s work, and the crew themselves are also capable of engaging in the creative enterprise by contributing to the movie director’s instructions by altering them or suggesting changes. The AI is incapable of doing this; it can only collate image sets that the AI draws from a set of images it has been taught to match to the prompt(s) given. In other words, the analogy fails to acknowledge that the film crew are also a part of the creative process and not simply robots that do what they’re told vis-à-vis their programming. For these AI to truly be creative, then, they must at least be “creative partners… autonomous, [responsible, and able to] curate as well as create” (Heaven, 2022). In other words, the AI needs to have a soul.

Soul

The Soul, as I will be exploring it, exists in two different contexts: the Aristotelian or Platonic, with many offshoots from each. In Michael Davis’ book The Soul of the Ancient Greeks, the soul is defined as follows via the proceeding argument:

“If life is movement, and if whatever moves itself by its own nature could never stop moving itself, then whatever moves itself would be deathless. Now if there is anything that moves itself, it is soul. But there must be something self-moving. If there is motion, it must have a beginning. A beginning cannot be moved by anything else, or it would not be a beginning. Since there is motion, there must be a beginning of motion, which is therefore self-moved. This is what we call soul. As its very nature is self-motion, it cannot cease moving or die; hence it is immortal.” (Davis, pp. 194-5)

This self-moving force can most succinctly be described as Eros, whose antithesis is Strife. The Phaedrus presents an image for us depicting a rider, and two horses, one a White Horse, representing Beauty or the concept we are drawn to (any one of us), and a Black Horse, representing our imperfections, our failures, or what others deride us for. As we look at the latter, we are drawn to the former and thus are inclined to goad the Black Horse towards the White Horse; hence, we unify ourselves as intelligent beings, capable of control, with our imperfections vis-à-vis our desires (what we find to be beautiful) so that we reduce our imperfections or control them; Eros then is the animating and unifying force of the Soul.   

But is anyone truly capable of grasping the Beautiful? Doing so would destroy its animating potential, for if one were to grasp that which animates it, its raison d'etre would cease to be, and thus so too would its becoming. Davis describes this quite well:

“One might say that we experience beautiful things as objects within the whole that point to the order of the whole. They are images of the perfection of the whole. But if the whole is a perfect order with its parts functioning so as each to fit perfectly into the whole and performs its function with no waste or excess, and if a part of the whole that images the whole can do so only by being apart from and not limited by the other parts of the whole, then this part of the whole can be beautiful only by being disruptive of the perfect order of the whole. This disruption of the wholeness of the whole is the same as the introduction of the seeming, and so falsity, into the being of the whole.” (p. 224)

Thus, we can conclude because Beauty cannot be grasped by its very nature without unraveling the whole, the Beautiful, as an animating force, is ungraspable and perpetually self-animating. In the Symposium, two forms of an animating force are identified: Eros and Thumos. Davis identifies Thumos as the self-animating passion or really a self-animating Eros. However, Eros does not exist with respect to the self; it animates a being irrespective of its passions or desires; Beauty (as a seemingly intelligible object) is the goal of Eros, while the Erotic (as an obtainable object of one’s own making) leads to Strife for its attainment is a hollow victory. Fame, as in the goal of Achilles, was one such erotic desire produced by Thumos. Thus, for Davis, the perfectly lovable, as an attainable, cannot be a genuinely animating force for a being. Once grasped, once the dove of our desire is held in our hands, it melts away and we lose all passion necessary for our becoming nature. Thus, the perfectly lovable cannot be the beginning of the perpetual becoming around us, for if ever any part of it or if it as a whole attained its object of desire, all would cease to exist. Since it does not cease to exist, the perfectly lovable is not the beginning of the perpetual becoming around us.

The goal then, for the Platonists, according to Davis, is to let the object of your desire draw you “out of yourself but not hold on to the [object] that draws you and threatens to trap you in a new version of yourself” (p. 228). When we are drawn out and grasp our desire, we reconstitute ourselves, we recognize our imperfections once more, and we once more shift our desires to a new object of desire; our old erotic desire, an object to be discarded. Thus, the essence of a Soul cannot be considered to be purely old, trite, or unsurprising. If an object is considered old, trite, or unsurprising, or lacks the kind of self-animating force necessary to the essence of Soul, which would perpetually allow the object to be reconsidered, granting it the animating force necessary for it to have a soul, then it lacks Soul, as the concept would be understood Platonically. Aristotle held a different conception of the Soul.

For Aristotle, the Soul divides into a Rational and a non-Rational part. The non-rational part consists of the “nutritive, generative, or plantlike and the desiring or appetitive.” The rational part of the soul consists of two parts: a practical and a theoretical part (Davis, pp. 56-7). Each of these parts consists of a structure not from within the being but from without. For Aristotle, this creates a conundrum, the qualities of the Soul are imposed from without; if that is so, then how is it that, with the use of the rational part (logos), one may impose structure from within? Is that structure not merely a continuation of the structure imposed from without? More importantly, if the soul’s rational aim is the Good, and one who is ensouled nears the Good, may we praise them for having neared the Good; i.e., may we praise the virtuous?

For Aristotle, the answer, in both instances, means that if we were to believe the proliferation of the Good, Virtuous, and Beautiful were really the product of forces that imposed a structure or habit onto a person, then we must not praise them, for they did nothing to produce such a product. Therefore, because we do praise people when they do Good, Virtuous or Beautiful things, such products cannot merely be from without the person who produced them but must also come from within the person who produced them. To continue to praise them for their virtues, for example, even though you believe they did nothing to achieve their virtue, leads to pure absurdity. Hence, virtue, goodness, beauty, all may come from within, not merely without.

The animating force for Aristotle is the Good, and the Beautiful is what draws one towards the Good, i.e., what animates a being. Aristotle’s main discussion then on the soul seems to be rooted in the problem posed by Plato, specifically: what should the aim of an individual be if, in seeking to achieve the Good, he will never fully realize it. For Aristotle, achieving the Good is like a dance. I.e., Joy or Happiness is the aim of the dance, but until such an end is known, the dance in itself is good. We are drawn to beautiful dancers, and we desire that we dance beautifully as an end in itself first because dancing is beautiful, and thus something is good about it. Yet then, once we know of the dance, once we know that it is good because it causes happiness, then happiness, or eudemonia, is the aim of the dance. Unfortunately, this then turns it into an art, a graspable thing, and likely causes it to – over time – lose its animating potential. One day, the dancer will find that he no longer feels happiness from dancing.

What Aristotle then proceeds to, if we are to lead a Good life, if our soul is to be animated, is the architectonic art. The architectonic art is not the art itself, the thing done, but the symbol of the art; “[it] exists solely as posited and not as experienced” (Davis, p. 63). Politics is one such form of the architectonic art, and for Aristotle, the highest; its desire is the Good, and thus produces the Beautiful and Just. But what exactly is the Good that Aristotle, through Politics, aims at?

The Good, as far as I can discern from Davis’ work, that Aristotle is aiming at is both product and action; i.e., a producer as a product whose genuine allure is its process. I.e., the Good is a desired object and action simultaneously. From Aristotle’s discussion on Politics its aim being the Good achieved through the beautiful and just, “the structure of the just comes to be the object of internal longing” (Davis, p. 71). This helps to solve the problem of fading happiness, acquired as an activity as a product. The individual thus knows its aim (the beautiful and just); this aim becomes both a product of the individual and a process of becoming by the individual. It is the activity of producing the beautiful and just for the sake of the polis, according to Aristotle, that is desired. The process becomes the product, and the product is the process, the thing to be desired in the product. To live virtuously is to produce a virtuous thing through one’s self, and one desires the virtuous thing because it has been acquired virtuously.  

For Aristotle, this doesn’t produce completeness, nor can it produce completeness; i.e., the happiness acquired through living virtuously, acquired by seeking the structure of the just as an object of internal longing – to produce justice and to be just – is never ultimately achieved. Specifically, we run into the problem of the fading dancer. Davis presents the argument for us:

“In reflecting on what it means to be happy, Aristotle sees that happiness belongs to a whole life. Priam would look to be the happiest of human beings before the last ten years of his life, when the king of a great and prosperous city with many successful and talented children lived to see his city captured, burned and pillaged; his sons killed; his wife and daughters carried off; and himself slain by the son of Achilles.” (Davis, 71)

If this was the fate of Priam, what is our fate? If happiness is ultimately achieved when we are pursuing the structure of the just as an object of our internal longing, can we ever genuinely and literally be happy? Our lives then, as they stand separate from us, ultimately serve as the object of the moralists: did Priam lead a happy life because he was virtuous or not?

Concerning Aristotle's conception of the Soul then, the animating force is happiness, which is the acquisition of the beautiful. But as Plato noted, once the beautiful as a tangible thing is acquired, its animating force is liable to fade. Even Biblically, man, as a corrupt being, if he were ever to hold onto a genuinely divine and beautiful thing here on earth, it would either destroy him or he would find some way to corrupt it. Then for Aristotle, through the rational faculty of the soul (logos), the aim needs to be both an object and a process. Eudemonia, happiness, excellence are acquired then through one’s desire for a just and beautiful life. The product, finally complete at life’s end, then serves as a paradigmatic model of the Good, as an object to be desired in itself both as a process that achieved virtue and a thing that is virtuous. We can then say, looking at this object, whether one was happy, for it would likely bring happiness to others if it served as a model (a product of a process whose product is a process) of a Good life.

Discussion and Conclusion

“We live in a world where there is more and more information, and less and less meaning.”

― Jean Baudrillard, Simulacra and Simulation

The first question I would like to address to conclude this paper is whether or not AI Art and Essays are “Creative.” Our definition of Creative consisted of two overlapping criteria: Appropriate and Original. Some appropriate things are original, and vice versa. What counts as “appropriate,” or as Boden identified it, valuable, is highly subjective. DALL-E and ChatGPT may be valuable from one perspective or mental schema, but it also may not be. For artists who use DALL-E, for example, its utility may run its course very quickly, given the fact that the newness or surprising quality of the images generated by the AI art loses sway due to the very nature of the software.

The trite quality of the answers provided by software like ChatGPT and the images created by DALL-E and DALL-E 2 is rooted in its programming. The nature of the pseudo-ECN that mediates the pseudo-DMN of such virtual interface (VI) software limits the output of the software to what it has been taught to associate with the prompts it has been given. To put this simply, neither ChatGPT nor DALL-E or any of their like can understand the semantics of the prompts it has been given; it can only understand the structure of the prompts posed to it or the prompt's correlates. For example, if the prompt is “Van Gogh Pineapple,” the AI will combine only images that have the tags Van Gogh and Pineapple. It will not add more to it than what it has been given, nor will it loosely construe the term “Pineapple” to mean something like a pine cone combined with an apple, a pine tree with apples, or an apple with pine needles coming out of it (nor will it think of an apple as anything other than what it’s been shown is an apple).  

Writing prompts suffer from the same issue, if not more. Stephan Shankland had ChatGPT write him a poem, but he didn't find it very "exciting." Upon prompting the chatbot to write something more “exciting,” the bot simply used terms it was taught to associate with excitement. However, I am not sure that merely using words associated with “excitement” makes a poem more exciting actually.

Responses by such software, incapable of autonomously acquiring and generating information for itself (due to the nature of its programming), are liable to provide incorrect answers (Shankland, 2022). The information ChatGPT uses to provide answers, for example, is contingent upon “human-made” information already online. In short, as more and more ChatGPT responses are generated and proliferated, the more it will simply use its own information. As people become reliant on software like ChatGPT for art, writing, etc., the more it will draw from VI-generated materials, the more it will simply be copying copies of artwork that were already derivative.

Without belaboring the point too much further, it is obvious the newness or the surprise created by the VI products will be exceedingly short.  

Art, I think it is safe to say, must have something like a soul. It must reflect the beautiful, and in reflecting the beautiful, reflect and imitate a good life; at its most virtuous, it should inspire justice and beauty in its observer, not as mere products or objects to be acquired but as a product generated by a process wherein the process is the perpetual product that produces the process which itself produces happiness, excellence, and eudemonia.

The question then is whether the AI is animated by the desire to imitate the beautiful and to frame the good life; i.e., to inspire virtue in its observer and be capable of inspiring justice and beauty as both a product to be acquired and a generative process capable of producing happiness, at least. I think it is very safe to say that the VI in question is not capable of achieving this.

This primarily rests in the fact that it is not aware of itself. It is only aware of itself in so far as its programmers have provided it with answers about itself. It may also search online to acquire information about itself, but again, this does not generate knowledge about itself; it provides an answer elicited by a prompting question.

This also rests, once again, on the fact that it does not understand the meaning of words. It has simply picked up on patterns and structures that occur with some frequency across a given domain, especially with respect to what it’s associated with a given prompt. This is why when you prompt ChatGPT with a very specific and fact-based question, you’re likely to get an accurate answer (although an accurate answer may only be one that has been proliferated the most – is such an answer actually accurate?). However, if your question is an oddball, if it’s unique, or no authoritative answer has been produced on the matter, ChatGPT will likely be unable to provide you an answer – you’ll only get nonsense. Hence, it is not animate, self-animating, and if it has a soul, it has more of a plant soul – i.e., it is primarily nutritive and appetitive; it consumes what you give it and provides fruits for your consumption; it is not animated, and it has a poor understanding of logic (Heaven, 2022); it doesn’t have an animal nor human soul.

Ultimately, I will end with this question: does AI actually produce art?

No, I do not think it produces art, nor do I even think it produces an essay; it produces what it has been programmed to associate with an essay. What it produces has some utility for artists, but if I’m correct, the work such artists will produce will become extremely derivative very rapidly; it all will be relying on the same reference material; and then it will rely on reference material it generated. If it’s not relying on reference material it generated, but works derived from the material generated from it, it will still be generating works from materials that it at least tacitly generated. Either way -- it will be very derivative. 

While the goal of artists, for a time, is to demonstrate they can produce art in the style of contemporary or past artists, making their work derivative, they must also ultimately take the skills they've acquired from such ventures and be capable of recombining them to produce novel works that generate value in some way. Hence artists are capable of more than simply derivative works. I do not think the VIs in question can achieve this kind of task.

Because the VI lacks a soul, is incapable of being continuously creative, and is incapable of being a model for a process and product (simultaneously), I do not think it would be accurate to call what the VI produces art. At most, it might finally reveal the reality or usher in the horror of Baudrillard’s Simulacra and Simulation.

Bibliography

Abraham, A., 2018. The neuroscience of creativity. Cambridge University Press.

Beaty, R.E., Benedek, M., Barry Kaufman, S. and Silvia, P.J., 2015. Default and executive network coupling supports creative idea production. Scientific reports, 5(1), pp.1-14.

Carroll, J., 2020. Imagination, the Brain’s Default Mode Network, and Imaginative Verbal Artifacts. In Evolutionary Perspectives on Imaginative Culture (pp. 31-52). Springer, Cham.

Davis, M., 2011. The Soul of the Greeks. In The Soul of the Greeks. University of Chicago Press.

Dujmovic, Jurica. “These artificial-intelligence models are disrupting the art world – and will change how movie are made.” MSN, MarketWatch. 24, Dec. 2022.

Heaven, Will Douglas. “Generative AI is changing everything. But what’s left when the hype is gone?.” MIT Technology Review. 16, Dec. 2022.

Jones, L.L. and Estes, Z., 2015. Convergent and divergent thinking in verbal analogy. Thinking & Reasoning, 21(4), pp.473-500.

Katz, Leslie. “Here’s What It Sounds Like When AI Writes Christmas Lyrics.” CNET. 22, Dec. 2022.

Khan, Imad. “ChatGPT Caused ‘Code Red’ at Google, Report Says.” CNET. 22, Dec. 2022.

Shankland, Stephen. “Why Everyone’s Obsessed With ChatGPT, a Mind-Blowing AI Chatbot.” CNET. 23, Dec. 2022.

Tennakoon, Taraka. “AI art: Death of creative industry, or its savior?.” Venturebeat, 17, Dec. 2022.