Is This the End of Translation?

 

“We’d better pack our bags. Looks like all the good times are gone.” When a translator teaching at my university uttered this remark in early March over the advent of artificial intelligence, I thought to myself: “Were there ever good times for translation?” Last winter, as Open AI—the world’s largest AI research institution—was launching ChatGPT, the general belief was that it could completely replace writers and translators. Discourse on the translation apocalypse became mainstream. The translation tools that people had relied on were at most Papago and Google Translate, but now even tools like DeepL have entered the fray. In times like these, has human translation truly run its course? Are there no longer any prospects for professional translators? What should those of us in the realm of literary translation education be concerned about? It is no surprise that in the face of these questions, the number of people translating literature is dwindling.

    Even though I am a bidirectional Korean-English translator of poetry, as well as a translation researcher and educator, I didn’t think much of these questions at first. But when I saw the faces of the students in my spring semester course, glowing half with interest and half with curiosity about what literary translation was, I was struck by this thought: ‘In a translation course, the assignments that students complete outside the classroom are important. How can I stop them from using ChatGPT to do them?’ So I made a decision. “All right, how about this? Let’s try it out with AI.” In that class, we began to learn more about the theory and practice of literary translation while working symbiotically with AI.

    Over the course of the semester, we conducted various experiments. My students employed every means possible to translate their target texts. The only prerequisite was that they had to compare the AI translations with their own and write an essay about their observations in which they considered what aspects were important when it came to literary translation. Contrary to my expectations, I had students who were dissatisfied with the AI translation as well as students who were surprised at how quickly the AI translator was able to generate its output.

    Here are the conclusions we drew after that semester. First of all, AI is great because it’s fast. But it makes frequent, glaring mistakes because it doesn’t understand context. Of all the many genres of literature, the one that remains a definite weakness for AI is poetry. In many cases, AI cannot discern the context or the particular characteristics of the form. That is not to say that human translations are perfect, either. The process evoked many laments such as “AI still has far to go!” and “Just as I thought—poetry is hard!” as well as sighs of relief with remarks such as “Ah, translation isn’t dead just yet! We need literary translators. Even more so for translating poetry!” This was how we ended the semester.

 

Can an AI Author an Afterword?

 

Once I shifted my concerns about plagiarism and other issues into an active use of ChatGPT in the classroom, the semester went by without incident, but I still had not managed to solve the issue of the diminishing place of translation in the face of AI. Yet as I was grading at the end of the term, a wave of dizziness struck me while looking over one student’s answer sheet. The student had responded to a question with more questions: “Why is everyone so intimidated by AI? Can an AI author an afterword?” To write an afterword for a translation, an AI would need to have a translation philosophy and a writing style of its own, but no AI translators meet these criteria. The student then asked about reading speed in literature. A self-described old-fashioned reader who preferred “slow reading,” the student said that because the process of literary translation was intertwined with the process of reading, to claim that human translators were useless against AI was to cede the importance of both literature and human beings.

    This student’s bold questions and belief in slow reading touch on the issues addressed in this essay. As a translator of poetry, I often ask about the starting and end points of a translation. Why translate, and what for? Translations of Korean poetry into English or English poetry into Korean are not exactly welcomed in the commercial literary market. The readership for this work is almost nonexistent. There has never been a heyday for poetry in translation, nor will there be in the future. So why translate? As I confronted this student’s questions, I was able to question my own motives as a translator. I also realized that the nature of the anxiety we feel in the age of AI is a kind of contamination brought about by thinking too simplistically about translation’s starting and end points.

    I translate poems because I love them and want to relay them to others. My reasoning may be simple, but my role as a messenger is a resolute higher calling. Poetry is a genre that even native speakers struggle to understand, so translators must have an eye to interpret the works in detail. If my starting point for reading and translating poetry is a love for the source texts, then the afterword I write at the final stop of translation is the mode, the ritual, that serves as the period officially marking the end of a translation project.

    In that regard, writing an afterword is both a joy and a right for the poetry translator. It is there that I reflect on how I encountered this collection of poems, what I enjoyed in the process of ferrying from one language to the other, what I found challenging, and how I found solutions for the uncertain bits. It is a lot like how we love. The kind of high-intensity love in which you adore something with a passion but must maintain a cool distance so as not to heedlessly fall into it. The afterword is my personal love letter to the poetry collection.

 

“Slow and Quick, Sharp”

 

Of course, not every literary translator encounters every work with this exceptional feeling of love. There are cases where translators are approached with requests from publishers. I translated my first poems from Korean to English in the early 2000s, when Korean poetry was virtually unknown in the world market. I also heard many slights about my research, asking why I was doing such untenable work. For the most part, all translations must endure delays. For example, it took twelve years from when I began translating Lee Seong-Bok’s Ah, Mouthless Things in 2005 until the collection was published in English. The poetry must first catch the translator’s eye, but it must also then stand out in the eyes of the target language editor. That kind of luck is not so easy to come by, which is why it takes such a long time for translations to be published.

    However, “slowness” here does not refer only to the difficulty of publishing translations. More than that, when we accept translation as an act of reading, this slowness becomes a necessary exercise. Since speed alone does not tell us everything about human reading practices, the much-lauded speed of AI translation does not make our reading more efficient. Thus, even when we talk about translation speed, we need to look at slowness in a new light. When it comes to translating and reading, what are the benefits of high speed? Human translators cannot translate a text they do not properly understand. AI translation tools manage to regardless. Like a human brain, when a user inputs a sentence into AI, the AI then translates it by converting it into vectors (coordinates) that contain information such as vocabulary, sentence structure, word order, etc. Thus, the more data a language has amassed, the more accurate and detailed the translation will be. For languages that have not amassed much data, the AI will make egregious errors and feign ignorance. The strength of ChatGPT is its unabashed confidence even when it has made a mistake. The output, untouched by the translator’s anguish, is crude in so many ways.

    What effect does speed have on us when we read works of literature, particularly poetry? Do we absorb what we read? In times like these, when speed and efficiency have become of the utmost importance, I think this slowness I describe can be applied as a new value to all reading and translation practices. Reading a poem is not encountering a single, unchanging interpretation. Likewise, translating a poem is not handing down a single, final verdict. Nonetheless, after all their agonizing, translators must make certain choices. Impossible choices. This process involves the dual rhythm of slowness and swiftness.

    “Impossibility” is a concept that always comes up when we discuss translating poetry. Many doubt whether translating poetry is truly possible. Let’s look at a poem that is often mentioned in these discussions—Kim Sowol’s “Azaleas.” A set of lines reads: “가시는 걸음걸음 / 놓인 그 꽃을 / 사뿐히 즈려밟고 가시옵소서.” Some critics argue that the word “즈려밟고” here should be interpreted as “to trample or stomp on,” while other critics interpret the word as meaning “to take the first step or go ahead of someone.” Depending on which side the translators are on, their translations will differ. Each translator has to make a decision. It is a mysterious and often painful process. Until the last period is placed at the end of a translation, the translator is starting from a basic reading within their own experience and knowledge and spanning out to high-level criticism. This is why I often say, “A translator is a work’s first reader and its final critic.”

    That is when creativity matters. Because a literary translation (which reveals the stance, personality, intelligence, empathy, reading ability, and language ability of the translator) involves all manner of reading and interpretation as well as criticism, it is impossible to carry out without creativity. One must read the semi-implicit meanings, omissions, and even spaces between the lines, then weigh the number of cases to make a final choice from among all the possible words with overlapping meanings. In his poem “A Sort of a Song,” the American poet William Carlos Williams advised writers to be “slow and quick, sharp” in choosing their words, likening poets to snakes waiting under the weeds. Translation, too, asks the same creative methodology of translators.

    No AI technology can achieve this. ChatGPT has neither the patience nor the sharp resolve. The task of choosing one of the many possibilities contained in a single word, that creative intervention, distinguishes poetry translation from literal translation, which simply replicates and reproduces. That slow and quick, sharp judgment. Creativity in translation can be thought of in a new way along with certain rhythms, different speeds, that come up during the creative process.

    In this way, the creativity of translation encompasses the creative reading of both translator and reader. Herein lies the greatest pleasure and comfort that translation can offer. Translators who fret over how to show a work of literature in the best light need the ability to regulate their own pace. To walk the taut line between two languages, staying focused while suppressing the desire to rush across. AI does not know this precarious, thrilling experience. AI does not know hesitation. It does not know the torment that persists until one solution is chosen out of all the different options, nor does it know the dual rhythm of slowness and swiftness involved in making that choice. The reason why AI cannot author an afterword is because it cannot recall any memories of that process.

    When I am stuck on a translation, I go for a walk or else I cook. After fretting for a while over the inverted order of the words, I will think of an exquisite turn of phrase while driving and scribble it on the back of my hand. That slow and quick, sharp judgment—that long wait—is a blessing, not a weakness, as well as an essential virtue for the critical work that translation accomplishes. It is an effort to rely on my heart and my mind to leap over the gap between two languages while feeling around for the historical, cultural contexts contained in the poems. I am slower than AI, but after waiting like a snake under my weed, I aim to be a messenger snatching up a language all my own.

 

The Subject and Community of Translation

 

Let’s consider the translator’s style. Translators are the first and last readers of a target text, and translations are bound to turn out differently depending on the translator. I’ve experienced this countless times in the translation practicum I teach at LTI Korea’s Translation Academy. Four students translating a text will yield four different versions, and six students translating a text will yield six. No two translations are identical. The subject of a translation is simultaneously a matter of the translator and their interpretations. We read these differences not from the angle of right or wrong but as a kind of gradation. The work becomes richer and more profound. Translation cannot be reduced to merely humans versus AI. Translating poetry is not a game of baduk where the winner is determined by expanding the reach of one’s territory in either black or white.

    Poetry creates newness. Poetry, which comes from the Greek word poiesis, meaning ‘to make,’ contains the most linguistic experimentation of all the many genres of literature. In making a familiar language appear unfamiliar, poetry is a genre that opens up new ways of seeing. Translation is relatively transparent about the production process of its subject. In that regard, translation is keener to consider the position of the subject. A closer look at the meanings of the word “subject” reveals several possible concepts, including the grammatical subject, the thematic subject, subjectivity, the self as subject, the citizen as subject, and so on. What I want to emphasize here is that, unlike AI technology, the subject of literary translation is a self-fashioning technology that can form different configurations according to its own interpretations. The interpretations that emerge via each of these different methods are a characteristic of literature that precedes other art forms. And poetry is there as well. By the power of its own subject-production, its making—poiesis itself.

    Let’s look the poet Cheon Sang-byeong’s collection 귀천 as an example. Brother Anthony of Taizé, the translator, rendered the title of this work in English as Back to Heaven. Not sky or paradise, but heaven. Here, “heaven” merges the sites of birth and death into one. It suggests that, in death, we return to the place where we were born. The principle of life that ties life and death together remains alive in the translation. In interpreting this work, the translator depicts the world the poet imagined. He pulls the late poet into an embrace. It is a beautiful and warm reception.

    In this way, translation enables new births that transcend different languages, time lags, life and death. The sympathetic community that shares translated poems creates a world that is different yet similar to, and similar yet different from, the world that has passed. A place of community and warmth created by translation. The creativity of translation is the seed that grows this community, these “commons.” And the subject of translation is the one who sows that seed. In that slow but sharp, fervent wait. In the era of AI, when I think again about the precarious place of the literary translator today, I think about “the one who sows the seed.” The sower does not linger only on fertile land. They may sow the seed even in barren soil. This seed sprouts with the help of all sorts of forces. No one knows exactly what kind of tree the seed will grow up to be. But what is certain is that it will become something. For me, as someone who reads and translates and teaches poetry, the joy and hope of translation resemble the higher calling of “the one who sows the seed.”

 



Translated by Paige Aniyah Morris

 



 

Korean Works Mentioned:

• Ah, Mouthless Things (tr. Chung Eun-Gwi, Myung Mi Kim, Brother Anthony of Taizé, Green Integer, 2017)

『아, 입이 없는 것들』 (문학과 지성사, 2003)

• Back to Heaven (tr. Brother Anthony of Taizé, Young-Moo Kim, Cornell East Asia Series, 2010)

『귀천』 (살림, 1989)

• “Azaleas,” Azaleas: A Book of Poems (tr. David McCann, Columbia University Press, 2007)

「진달래꽃」, 『진달래꽃』 (숭문사, 1951)