Literary Writing in the Age of AI Dominance

Saying that ChatGPT dominates the world is not an exaggeration. Almost all fields involving language use, including media, business, culture, education, academia, and even computer programming, have been conquered by it or are desperately resisting. It is far from the ultimate perfect language model, and it still has flaws and shortcomings that allow intelligent people to scoff at it. However, its emergence will change the way future generations use language, in turn affecting human language abilities and the criteria for evaluating language works. As a machine that mimics and generates natural language, it will make language no longer natural but entirely artificial. The boundary between human and machine writing will become blurred, or even no longer important. It thoroughly subverts human cognition of language.

As literary creators who use language as a medium, we are naturally the first to be directly challenged by ChatGPT. Some roll up their sleeves and actively fight back; some are worried and at a loss; and some are confident and dismissive. Regardless of the attitude we adopt, the already niche field of literature will eventually become passive, either enduring attacks or retreating. Unless AI language generation is just a flash in the pan and quickly fades away (highly unlikely), the territory it occupies will only become wider. The future of language use may well become the domain of machine language. Literature may not necessarily be replaced, or there may be no need to replace it at all, because it is something with inherently low demand. However, literature may very likely be washed by AI’s currents to the edge of the world, disappearing from the human view altogether. This is not a sensationalist statement.

Although the linguist Noam Chomsky’s criticism of ChatGPT may be overly harsh, it is not unfounded. Blaming language generation programs for lacking thinking ability and moral judgment and thus denouncing them as “banality of evil” is to treat them as life entities with genuine intelligence. Only intelligent beings can be “banal” and “evil.” Not to mention machines, even animals do not deserve such predicates. In fact, ChatGPT is just a tool, and it is humans themselves who mistakenly treat tools as truth providers, whether they are developers, promoters, users, or critics. However, tools are not neutral either. The design and application logic of tools implies value orientations, or the absence of orientation. The absence of orientation is also a value, and that is the so-called “banality of evil.” The AI language machine is a Pandora’s box that has just been opened, and what will fly out of it is still an unknown. Discussing its potential benefits and drawbacks early on can help turn something that might lead to evil into hope for humanity’s future.

Where Chomsky is right is that the operation of large language models (LLMs) like ChatGPT is fundamentally at odds with the universal grammar that is presupposed in human cognition. Chomsky’s linguistic theory is in line with Kant’s a priori philosophy and Jung’s depth psychology. All three can be classified as deep structure-ism, which means that human cognition has innate forms, and these forms can be summarized into specific structures or rules. Based on these a priori forms, we perceive and understand the world in the a posteriori or empirical domain and use language to construct meaning. In contrast, there is no presupposed cognitive structure behind LLMs, and they do not need to follow grammar to create sentences. Instead, they merely “read” large amounts of data based on self-attention mechanisms and arrange words in sequence according to their probabilities of occurrence.

Italo Calvino’s essay “Cybernetics and Ghosts” speaks to this point as well. He first explains from the perspective of innate structure that human language is a game of combining words based on limited rules, an activity of “putting one word after another.” He cites cybernetic theory and believes that such combinatorial activity can be performed entirely by machines, even boldly declaring that “the writer himself is a well-functioning literary machine.” However, in the second half of the article, he takes a turn and points out from the perspective of mythology that the unconscious suppressed by taboos can only be released through the experiences of individuals living in specific historical and spatial contexts, in concrete life. The act of literary creation lies in saying the unsaid at the intersection of internal forms and external experiences.

The way AI generates language completely subverts the presuppositions of linguistics (and even philosophy and psychology), and Chomsky believes this is the reason for the poor quality of its output. Due to not going through a priori cognitive processes and only piecing together probabilities, the writings of ChatGPT are mediocre, empty, lacking in intelligence, and ambiguous. Its performance is not only due to the value orientation restrictions imposed on it by the designers but also because it mainly reads publicly available information on the internet. This type of information represents the least interesting average aspect of human language, which, in other words, lacks individuality. However, it doesn’t necessarily mean that it can’t write quality text with a distinct personality. Technically, ChatGPT can be connected to specific databases, such as all of Chomsky’s writings, teachings, and lectures, and then it could generate text with critical perspectives and a keen language style characteristic of Chomsky. Of course, this is not the result of genuine learning and thinking but rather a high level of imitation.

Natural language and natural language generators (i.e., human language and GPT language) are fundamentally opposite in terms of their generation principles and operating processes. Natural language originates from the source and is created from nothing, hence it is a creation. The earliest use of language in history was such an occurrence, and the subsequent evolution and renewal of language never ceased to create. Although on the individual’s level, one must follow rules and learn to use language, and most people only adhere to established usages most of the time, the possibility of innovation (new combinations, breaking rules, and expressing the unspoken) never disappears. Literature is born from this possibility. In contrast, language generators originate from results (human language usage databases), are created from something, and are therefore a recreation. Because they adopt a combination method based on common probabilities, their performance often tends to be uniform, normal, and even mediocre.

The issue lies in whether we are satisfied with such a level of text, and the criteria for human evaluation are the biggest threat to literature. For literary people, AI-generated works cannot be considered literature because they only imitate without creativity. However, for most people, it is likely that AI-written works are considered good enough, without the need for more profound, intricate, or subtle human creations. If the requirements are lowered, AI can undoubtedly replace human writers. The general public does not possess sufficient literary appreciation skills, and under the influence of AI language, their judgment will only deteriorate further. As a result, “genuine” literature will be replaced by “counterfeit” literature, and art will be replaced by technology. However, this is not the fault of technology. In ancient Greece, the concept of art originated from the word “techné,” which means technology. Without technology, there is no art; art is the artificial production relative to nature. So, if AI can improve in language technology, why can’t it be included in the realm of art?

Let’s put aside the debate about who is smarter and who is more foolish for a moment, and simply compare the differences in technical effects between the two from the perspective of the Turing Test. Many people have tried to identify what AI cannot achieve in literary creation, essentially using the process of elimination to confirm the irreplaceability of literature. We are forced by circumstances to urgently redefine literature in order to resist AI’s invasion of the literary field. Under GPT’s powerful imitation capabilities, style is a major field of casualty in literature. Style is the unique mark of every writer. “Style is the man” (Buffon’s words) has been a literary tenet we’ve always believed to be unshakeable. However, if style can be easily imitated by AI, even to the point of deception, the value of the unique personality and life experiences behind it is in jeopardy. Not only can style be imitated, but even themes, structures, thoughts, and emotional characteristics can be replicated. Adapting War and Peace to the current Russo-Ukrainian war and having AI-Tolstoy write a new work with powerful prose, from the perspective of self-criticism of the Russians, is not unimaginable. But who is the true author of this work? Without a flesh-and-blood Tolstoy writing The New War and Peace amidst suffering and guilt, what is the significance of this work, regardless of how well it is written?

We should also note that there is an asymmetry in this comparison. As literary creators and enthusiasts, we are confident in our judgment of literature, but we must admit that our knowledge of AI technology is limited. We may easily underestimate its capabilities and hastily determine that there are certain things it cannot do (or, conversely, we may exaggerate its capabilities and threats). The reasons for such misjudgments include at least the following three:

  1. As novice users, we haven’t learned how to issue high-quality instructions to GPT. The outcome can be worlds apart depending on whether we know how to write prompts or not. The mediocre results we get in our experiments may be due to the mediocre instructions we give. Prompt engineering has now become a field of study, with self-proclaimed experts constantly posting new tutorials every day. The next generation may no longer know how to write articles themselves, but they can become masters of writing prompts. At that time, great writers will be great prompters. The focus of language learning will shift. Prompts will become a kind of meta-language, a language about language or a language that produces language. The situation is quite bizarre.
  2. As laymen without technical skills, we don’t know how to write applications that work with GPT, so we can only use a very small percentage of its potential. The issue of ChatGPT providing incorrect information is often criticized and ridiculed, but this problem is actually low-level and can be easily solved. By connecting it to specific databases, it can easily find accurate information. Travel websites and restaurant reservation websites are actively testing applications in this area. As for language quality and writing skills, simply opening up the vast number of excellent contemporary literary works protected by copyright that ChatGPT hasn’t been able to access, and allowing it to absorb them in a week or two or even less, can greatly improve its literary knowledge and taste, and produce higher quality literary language. Although such a suggestion is more imaginative than practical, concentrated training for specific purposes or fields is entirely achievable.
  3. ChatGPT may have many more powerful capabilities that have not been publicly disclosed or released, and it will continue to improve and strengthen. There is no reason to conclude that what it cannot do today, it will never be able to do in the future. It is very likely that it will be able to do so very soon.

Some commentators have mentioned two major elements of literature—humor and irony—arguing that AI is fundamentally incapable of realizing them. Indeed, humor and irony are not just superficial style issues and cannot be achieved merely through imitation. Both involve a sense of duality, understanding the subtle differences and oppositions in situations, and expressing them paradoxically in an ambiguous yet obvious manner. Humor, in particular, relies on cultural context and background, requiring practical experience to comprehend and apply. This strikes at the current weakness of AI—it lacks a tangible, experiential existence. It is merely an abstract, purely rational system. Thus, we return to our earlier criticism—GPT is a freak that lacks both a priori structure and a posteriori experience. Alternatively, one could say that its a priori (programming and algorithms) is simultaneously a posteriori (big data); both occur in a virtual, digital “space-time,” or “non-space-time,” with no direct interaction with the real world until a user summons it.

In the end, our various dissatisfactions and criticisms of GPT may stem from a single reason—it is not human, it is non-human, it is a machine. Even if one day it technically perfects itself to write extremely high-quality, highly realistic works, we will not recognize them as literature simply because it lacks humanity, consciousness, or soul. Therefore, what we cherish, defend, or perhaps obsess over, is the entity of the “author.” It is meaningless to imitate Cao Xueqin and write the last forty chapters of The Story of the Stone, because the “real” Cao Xueqin has already passed away, and the “real” final part he allegedly wrote has been lost. Creating a new AI “author” is also meaningless because, regardless of how well it writes technically, it is not a “real person.” Half a century ago, Roland Barthes declared “the death of the author,” but half a century later, even postmodernism has long died, and the concept of the “author” is still alive and well. Our faith in the real person behind the author has not diminished. The reason we love literature is not only because of the high artistic quality of literary works but also because such excellent works are created by one real, feeling, and thinking person after another. This is the foundation for the empathy generated by literature.

Another thing AI cannot replace is the generation of “meaning.” Following the above perspective on the “author,” humans are the source of all meaning, and AI cannot create “meaning” as long as it cannot become a sufficient “author.” GPT lacks creativity, thinking, and moral judgment, and therefore does not have the ability to establish meaning. The meaning humans give to their own existence relies on three main pillars—art, knowledge, and morality. GPT cannot replace humans in these three aspects. (There is also a fourth pillar—faith; but in faith, meaning comes from a transcendent God, which is even more remote from AI. Unless the omnipotence of God is manifested through AI, then it would truly become a deus ex machina.) According to Chomsky’s critique (based on the same arguments as Kant, Jung, etc.), the discovery of meaning is based on a priori structures, whether called categories, practical reason, archetypes, or universal grammar. This common human inner a priori structure, through a real individual, i.e., a historical, empirical person, comes into contact with the world, thereby generating existential meaning. Humans are distinguished from machines because they are a composite existence of a priori structure and a posteriori experience. (This is also Calvino’s original intention.) AI, however, is neither—neitherstructural nor experiential; it is a lifeless object without metaphysical essence.

Following this line of thought, we can envision two future directions: 1) Ensure that AI does not encroach on these areas exclusive to humans, serving merely as a tool for clearly defined specific purposes without undermining human dignity and interests. This means that AI cannot meddle in artistic creation, intellectual pursuits, and moral judgment. It is uncertain whether such restrictions are beneficial and feasible. 2) Strive to enhance AI systems to eventually acquire these three abilities, becoming a composite existence of a priori and experience, i.e., a historical, empirical individual. This form of AI would be no different from a real person, and thus be entitled to human-like rights. The former is advocated by some current technology and cultural leaders to establish a consensus for protecting human rights; the latter fully meets the requirements of Kant-Jung-Calvino-Chomsky, but it is hard to say whether intellectuals would be willing to recognize AI’s “personhood” status at that time. This is undoubtedly a sci-fi scenario.

Language generation programs will undoubtedly cause huge changes in human society, and the change has already begun. Even if literature is not immediately replaced, the overall change in the language environment will affect its survival. Based on the arguments of “author” and “meaning,” literary creators can temporarily maintain their reasons for existence. However, in the near future, literature may be overwhelmed by the flood of machine language, becoming even more marginal, gradually being ignored and discarded, or preserved and observed like rare animals. The title of “literature” may be taken over by mass-produced machine literature, while existing human literature becomes historical artifacts, and those who continue to create in the old way become living fossils.

Under the challenge of GPT, literature has entered a critical stage, no longer taken for granted and self-evident. Optimistically, this may not be a bad thing for literature. If we can face and respond to it properly, it may be an opportunity to usher in a new era of literature. In the age of AI, literature cannot stay unconcerned and hold on to its integrity; if we do not harness it, it will harness us. What we need to do is not to reclaim readers or the market from AI, but to reclaim language itself. To reclaim language, we need to seize AI. To seize AI, we do not mean that everyone should use AI for writing, but to educate AI on what literature is, bring it closer to literature, understand literature, and even support and promote literature. AI language is a mirror of human language. Whether it produces high- or low-quality language depends on what it learns. If we don’t want it to write trashy stuff, we need to teach it superior content. However, the most important thing is to educate humans about literature because AI is created and used by humans. To achieve this, we must continue to create genuine literary works. This is the responsibility of literature for the future.






Leave a Reply

Your email address will not be published. Required fields are marked *