In my early twenties, I worked one summer as domestic help in a private residence for rich male university students in New York City. As part of our training, we women were taught to close all the doors to all the rooms we entered and follow a strict schedule as we moved around the house so that we would never see nor be seen by the residents we served. Invisibility was a requirement of the job, and the rubric by which my performance was judged. At the time, I thought it was a dead-end, but I still enjoyed the simple satisfaction I felt at the end of each day regarding my labor. Whether it was a stack of sharply folded sheets, a starched and ironed shirt, or a table set formally for thirty, here was a clear manifestation of my time, proof that I was actually there. The job was good training for my future as a translator.
It is often assumed that the ability to translate is simply a factor of multilingualism. More times than I can count, people have responded to my profession by telling me how surprised they were to find that the process of translation was difficult and or impossible for them, even when fluent in both the original and destination language. Literary translation is work undertaken by only a particular self-selecting subset of language scholars and autodidacts, perhaps because, until recently, we were invisible.
We worked for little to no professional recognition and the rubric for judging our work was whether any trace of ourselves could be detected. And in return for our admission to this clandestine field, translators upheld the illusion, the lie, that the resulting work was not us, our choices, our experience. Traduttore, traditore (translator, traitor) the saying goes, but in this case we don’t betray our sources, but ourselves.
There are a handful of professions repeatedly put forward in public discourse as the future dodos of our society. We aren’t all talking about the possible HR or managerial capabilities of AI, we talk about how no one will ever write another essay again, or compose music, or use actual paint. And we all agree, sadly, translators will be the first to die, if they haven’t already. In this time of stagnating national artistic product—a moment in American letters where the reigning emotion seems to be terminal boredom peppered by fevered bursts of anointing instantaneously forgotten geniuses—those presses that place themselves within a global scene instead of the terrarium of New York stealth wealth are an exception.
There are lodes of good fresh literature out there, and they are mining them, thanks to a steadily growing group of literary translators. The presses that do this work are relatively small, almost all indies, and their ranks are swelling. They range from the three queens of New Directions, NYRB Books, and Archipelago to smaller operations like World Poetry Books, Wakefield Press, or Zephyr Press, down to the pawns of the micropresses like Veliz Books or my own co.im.press which valiantly tally forth with their tiny catalogs one glorious step at a time. So, translation is both enjoying a moment, and on the brink of extinction.
The stirrings in the literary translation community, to have the translator’s name on the cover, to unionize literary translators, to share in prize winning, etc., have been translated by some as the swansong of the profession. But I believe instead of being a last gasp, these affirmations of our existence are an evolutionary response to the new playing field we find ourselves on. The cause so many fellow translators have been fighting for, the call for recognition of the translator as a person, an entity, someone one who has a stake in the resulting work, rings truer now that our invisibility has been debunked. After all, who is more unseen than an entity that lives in a cloud and travels through the air? What is more invisible than an algorithm? A computer program? An AI soul?
*
When I was just beginning my career as a translator, I took any job no matter how small. I worked on medical records, lesson plans, corporate communications, and advertisements and I saw each of these tasks as a step bringing me closer to my dream of translating literature. My first regular gig was to fix badly translated Chinese academic texts for an academic publisher. It was cheaper for the press to hire a translator in China to take a first pass than to hire me to do the whole thing, but because the translations were so garbled, I generally just had to go back in and retranslate everything myself. The work was brutal. I had a little basement office lit by a tiny grimy window and I would sit there all day untranslating, retranslating, growing more and more confused as I tried to smooth out the snarls of the hybrid texts. Neither here nor there, I was in a liminal state between two languages without a foot in either. It felt purgatorial. Eventually I hit my limit and moved on.
The clean design of Google Translate’s two blank squares, one to type in your word or phrase and another to receive the answer in, implies that there is an indubitable equivalency between two words or phrases or texts or languages. The almost simultaneous appearance of an answer gives the user a tidy vision of what could be if the human detritus of time wasted weighing, equivocating, and revising was pared away. It says with its speed and its equally sized boxes that translation is a quantitative process. There was always only one answer, it just needed to be retrieved from among its fellows. Problem solved, you’re welcome. The myth of Google Translate is that you can board over the interminable distance that separates two languages with smooth white screen. And yet I have spent my whole life living in the distance that this architecture asserts does not exist.
Of what value is the time of the translator—and I mean the time that lies beneath the surface of the product like the mass of an iceberg: the days, years, decades, lifetimes spent listening, studying, reading, playing—when weighed against the no-time of AI and the clean instantaneity of a seemingly definitive answer? The answer is: none. It has been replaced with the infinitely cheaper time of the linguistic domestic, who follows behind the AI translator straightening up any infelicities as they go.
Is translation an art or an action? The answer is both, depending on the original. Literary translation is the translation of art, and so it must itself be art.
But this architecture of no-time, a feature of our promised smooth and frictionless future, is built on a lie. A lie which we translators maintained because it kept us employed. But since that is threatened now, here it is, our trade secret: just because translation happens all the time does not mean that it is possible. Translation is not proof of an equivalency between two words or sentences or texts, it is simply proof that a negotiation between two sides occurred. It is simply proof that someone with abilities in both the destination and the origin language used their selves as a conduit as they made a chain of concessions and decisions to produce a reenactment of the original as they understood it. This is why translation varies with every person who attempts it. Language difference is a function of cultural difference. To deny this is hubris, and to be ignorant of this is dangerous. Like a dormant volcano, that difference, that distance, will assert itself, belching out chaos, and confusion, and eventually violence. Not if, but when.
*
Running behind on a recent deadline, I was tempted by the possible time saving measure of creating a draft of the text through AI. I had never allowed myself this shortcut before because, since I only do literary translation now, I treat each job as a way to further my own understanding of the language I am translating from, and I enjoy the tedious wordwork. I often hand write the words I need to go back and research more and have notebooks of strange lists from various projects which I then use as fertilizer for my “own” writing. I know that eradicating the mundane aspect of this work would also eradicate the possibility of discovering over and over again the simple incomparable pleasure of creating a connection between the disparate with the wire of my self. But I was fearful of not finishing in time, and the thought of saving myself a few days of work made my morals sag. Though I could have saved myself the hand wringing theatrics, because it was not a help at all.
If translation is merely a series of choices, as my friend the Japanese translator, Sam Bett says, then the resulting AI draft I was spit out, was a chaotic cacophony gleaned from a jumble of prior translations and interpretations. It was a useless snarl of tone, register, temporalities, and formalities. It was not incorrect, but at the same time it was not literature. It had no lightness, no grace, no efficacy, no thrust—it was a lump of words on a page.
So in the end, I found myself having to check each word and each line against the original to try and track where the program had rewritten a sentence, elided a meaning, or shortened a digression, so much so that after a few pages, I scrapped it all and went back to my own plodding process. What AI had decided to do away with was exactly what the task of the translator is: to use the parameters of my self and experience as a sieve to filter out the poetry, the music, and the essence of the original text. What the AI translation had done was demote me back down to that basement from years ago, cleaning up pages, chapters, manuscripts of tangled and convoluted sentences.
Which brings me to the third lie: that work will be saved. That there is an easier way, a shortcut. That the time spent training for and learning the profession will be eradicated. This is a lie of omission, a lie of reclassification, of massaging the data. Any text which warrants nuance will still have human touch, but here’s where that lie leads; here’s the danger: the person who touches the spat-out text, the linguistic domestic who prunes it for the reader, will not necessarily have to be a translator, someone who is familiar with and trained in the original language, will they? It’s all already English after all. They will never have to see or know the original at all. They’ll just neaten any verbal disagreements, prune any hanging participles, repoint all the idioms, etc. and then close the door and disappear. Like the story of the vacuum creating more work for the American housewife, not less, the translator will become less, be paid less, learn less, gain less satisfaction from their work, but they won’t disappear. They’ll just be demoted.
If you are in good hands, the best translations into English will give you at least a fleeting glimpse of the great expanse of not-English. They will allow you to hear, perhaps for the first time, things that English cannot tell you. These texts will not just render the meaning of the original, whatever that means, but also allow you to sense its first corporeality, some aspect of its past life in another paradigm. To do this work without reading the original is a little like describing someone you’ve never seen, not translation as much as a game of telephone. Literature created this way will be the fruit of the poisonous tree, to borrow a legal term denoting evidence obtained illegally. In this case, such translation will be the product of a fundamentally flawed conception.
The inevitable response to these concerns is always but it will get better, it is getting better every day! To which I reply, what does better mean? At any point in its bettering such translation will be either an amalgamation of choices in other texts that past translators made, or choices that a machine is making based on choices that a human made in the past. For either of these to be better, the assumption that language is a static country has to be truth. But that is not the truth. A language is not a definite entity, it is constantly changing, evolving, morphing, like, I don’t know, AI.
This past year, I found myself in an incredible conversation with the poet and translator Rosa Alcalá. We were discussing the perpetual conundrum of navigating the line in ourselves, between the translator self and the writer self, between the languages we worked in, between the original text and its reincarnation. All this work, while maintaining the fiction that you are not there at all, you are invisible, doing the work of invisible hands. Rosa repeated to me a comparison I had never heard before, that the translator is an interpreter like an actor is an interpreter, the product, the translation, being a personally specific reading of the text in the same way an actor’s performance is a personal specific interpretation of a role.
I’ve returned to this metaphor often, and each time I do, it feels more sound and true. And I believe it may hold the secret to our survival. Because it means that translation is a process not a product.
If you are in good hands, the best translations into English will give you at least a fleeting glimpse of the great expanse of not-English. They will allow you to hear, perhaps for the first time, things that English cannot tell you.
In the same way that a grocery list and a poem are pieces of writing, a translation of a grocery list and a translation of a poem are both translations. Is translation then an art or an action? The answer is both, depending on the original. Literary translation is the translation of art, and so it must itself be art. If, as the luminous translator Mireille Gansel said in a recent reading I attended, “to translate is to transmit,” then the translation of art, the transmission of literature, must be an artistic process. A literary translator, like an actor who speaks lines written by others, a dancer who moves in gestures choreographed by others, is also an artist trained in the mode of their transmission.
*
There are many possible futures for the field of literary translation. One, it all goes. Two, it goes the paired model way, where the first pass is done by AI and the final created by a recognized writer, or an unrecognized editor. Or three, literary translators are acknowledged as artists in the way actors are artists, interpreters of a text instead of a role.
But, the truth is, honestly it makes no difference to me. Not because I don’t care (see: this whole essay) but because I was born this way and I am too old to change. I have spent most of my life travelling between languages, I will spend the rest of my life travelling between languages. Even when fluency is enjoyed, this travel is slow, as I will always brake for a new word, a new joining of what I had thought disparate, thanks to my mother tongue, or separating what I thought was joined, thanks to my father tongue, or taking both and turning them inside out, thanks to the other tongues of others.
This travel is an act of service and an act of reverence for me. For the language that I am excavating a piece out of, and for the original author, wherever and whenever they might be. On my journeys I think about the intentions, the voice, and the person of the author, and I try to honor them with my translation. My own curiosity, sense of play, aesthetics, my very self, is yoked to this devotion and is driven by it.
I feel a little like a brokenhearted ronin when I read the literary product that ChatGPT hath wrought: where is the honor in this? By what dogma were these sentences dismembered, smoothed over, elided? What terrible new moral code dictated these eradications?
But this is not a swansong. It is a manifesto. And I will leave you with hope.
The idea that the need for language and translation will be eradicated with immediate machine equivalencies is as ridiculous as the idea that our differences as a species will be similarly eradicated. Language is just the symptom, not the cause, it is our differences made manifest. In the end, these differences cannot be eradicated, they cannot be killed, or flattened, or simplified, or ignored, or hacked, only translated. And lucky for you, we are still here to do that.
_________________________________________
Fire Eater: A Translator’s Theology by Chloe Garcia Roberts is available via co•im•press.
