Traduire comme du monde

Notes for a talk to the Literary Translators Association of Canada

Comme du monde

It’s taken me years to get used to how people in Quebec use “comme du monde.” Je ne suis pas du monde, people say when they are out of sorts. Y’est pas du monde for someone unfriendly, uncooperative. Yet fais ça comme du monde means to do a job right, no corners cut. Literally, “like a person.” So, let’s start here as we think about what it means to translate both well and like a human, now that the latter cannot be taken for granted.

First principles

I hear a lot about why machine translation is bad, and much less about what makes human translation good. Yet much human translation lacks the natural ease of writing by native speakers in their language, and most of the criticisms I hear of machine translation/AI could be made against much human translation.

I also sometimes read translations so good they fill me with joy and awe. How has the translator done it? No two have the same approach. What follows are my own ideas and beliefs, based on my experience and reading.

1.      Refusing to use AI won’t make us better translators, any more than using it will.
2.      The comparison that matters is not between an unedited machine translation and one carefully translated by a professional; it is between two produced by the same professional spending the same amount of time, with or without the aid of machine translation.
3.      The threat posed by AI to our profession is (at least) twofold: that it will eliminate our livelihood, and that it will alter our work in a manner that sucks all the joy from it. (I fear the latter.)
4.      To keep translating in a way that is both economically viable and enjoyable to us as human beings we will have to think hard – perhaps for the first time – about what makes this a human process.
5.      The problems AI brings to the fore are structural, and solutions to them will come from communities, not individuals.

 Why would I want to?

I’m a writer because I want to write. I don’t want a machine to do it for me … what is the point of handing over the job of understanding something more deeply, seeing the pattern that underlies? Why would I want to give up that profound experience?

As translators we need to ask ourselves the same question, both as to using AI and as to doing the work we do: Why would we want to?

Assume we don’t wish to eliminate our own labour, which we depend on for pay and fulfillment. The question then becomes more subtle: Is the work of translation, which is like writing in entailing “understanding more deeply, seeing the pattern that underlies,” incompatible with giving over a portion of the task to AI? A second and more existentially challenging question is: Could that portion be an entire first draft?

I think it could. A narrow view of what it means to translate and to be a translator has caused us to focus on the knowledge of languages and on finding the right words, giving short shrift to the equally important work of arranging these words into compelling sentences and paragraphs. We privilege meaning to the detriment of form. This often yields prose that says the same thing as the original but doesn’t do the same thing.    

If we broke down the work of literary translators into multiple steps, it might look like this:

1.      Read widely in the language we translate from
2.      Find projects
3.      Pitch/negotiate
4.      Research
5.      Draft
6.      Edit
7.      Draft
8.      Polish
9.      Collaborate with other workers in the publishing chain
10.    Share/publicize

I’m surely not alone in noticing that at least 8 of these 10 steps are unpaid. They are the parts of the job universally acknowledged but occluded by our “per word” system of payment.  This misconception begets another: If translators are paid to spit out words – not craft great sentences, not bring books from one world into another – then should we be surprised when people think that providing us a mess of words already in the right language is justification to slash our pay? How long can it really take to “fix up” a machine translation?

In my practice, half-way through 2024, the answer is almost as long as starting from scratch. My own unassisted first drafts, the first of around ten, are quickly hammered out, so with or without AI the vast majority of my time is spent rewriting and redrafting, and having others help me do more of the same – basically turning over each sentence until I can’t see any way to make it better. Even after a book is printed and out in the world, no longer within my power, I’ll keep picking it up and reading random sentences, rejoicing when they measure up and feeling my heart sink when I find a loose joint or unhappy turn of phrase.

Literary translation, literary writing, is an individual and communal experience unfolding over time and carrying us toward understanding. It is a deeply human process. But what does that mean?

 Writing is a technology

Walter J. Ong’s “Writing is a Technology that Restructures Thought” is the most economical description I know of how writing is itself a technology that has so shaped our thoughts that we cannot see its power.

Writing (and especially alphabetic writing) is a technology, calling for the use of tools and other equipment, styli or brushes or pens, carefully prepared surfaces such as paper, animals skins, strips of wood, as well as inks or paints … Although we take writing so much for granted as to forget that it is a technology, writing is in a way the most drastic of the three technologies of the word. It initiated what printing and electronics only continued, the physical reduction of dynamic sound to quiescent space, the separation of the word from the living present, where alone real, spoken words exist … By contrast with natural, oral speech, writing is completely artificial. There is no way to write “naturally.”

Ong’s choice of the world artificial is tonic: it reminds us that we are staring at the invading barbrarian hordes of AI from atop the stack of earlier technologies we have internalized and grown so comfortable with that we mistake them for the natural order of the world. As translators, the technologies we accept and even love are frequently cumbersome, like paper dictionaries; some, like CAT tools, feel like working in a straightjacket.

Technologies are artificial, but – paradox again, artificiality is natural to human beings. Technology, properly interiorized, does not degrade human life but on the contrary enhances it. The modern orchestra, for example, is a result of high technology. A clarinet is an instrument, which is to say a tool. A piano is an intricate, hand-powered machine. An organ is a huge machine, with sources of power – pumps, bellows, electric generators, motors – in motion before the organist is touched by its operator.

Since I first read Ong’s short essay twenty years ago, I have been unable to be complacent about my knee-jerk rejection of new technologies (and my knee always jerks).

The fact is that by using the mechanical contrivance a clarinettist or pianist or an organist can express something poignantly human that cannot be expressed without mechanical contrivance. To achieve such expression effectively, of course, the musician has to have interiorized the technology, made the tool or machine a second nature, a psychological part of himself or herself … We are not born with art but add it to ourselves. Mastering a musical tool, making it one’s own, calls for years of mechanical “practice,” learning how we can make the tool do mechanically all that it can do. Little girls and boys know how boring it can be. Yet such shaping of the tool to one’s self, learning a technological skill, is hardly dehumanizing. The use of technology can enrich the human psyche, enlarge human spirit, set it free, intensify its interior life.

The sentences alone do the convincing

Everyone needs to find their own sense of what it is they actually do. Mine is that my job is to write sentences. Yes, I must read and understand the original language, and I must carry over the full breadth of its multiple meanings, but many people can do this, and AI is getting close. If all we can do is convey the meaning of another text, then we are indeed replaceable by machines. But there is something in good writing, in great sentences, that like other human relations and productions is very hard to fake. When we applaud the “lifelikeness” of AI-generated prose – it is eerily almost convincing – we also catch a whiff of insincerity, of unreality; an absence of conviction. (The same is true of much human-generated prose, like corporate communications).

I love Charles D’Ambrosio’s explanation of his method for writing essays, and believe we translators have much to learn from his idea that the music, mood, and feel of prose can precede and produce meaning, rather than being its poor cousin, an afterthought.

The subjects of these pieces mattered, of course, but it was important to me that the sentences alone do the convincing. What this meant in actual practice was that I often had no idea what I was doing, no plan or sense of purpose, until I started putting words on paper. I relied on my ear to a ridiculous extent, trusting that if I got the sound right – the music, the mood, the feel of things – then sense might eventually make an appearance. Sometimes sense showed up, and other times, in the tussle of trial and error, error won the day …

I relied on my ear to a ridiculous extent, writes D’Ambrosio, who says elsewhere that “one of my earliest ideas about writing was that the rhythms of prose came from the body, and although I still believe that, I still don’t know what I mean,” and again elsewhere that “the unit of measure that best suits prose is the human breath.”

To apply this to translation, we might suggest that the rhythm of our prose must come not from that of our original text – even while it must somehow reflect and honour it – but from somewhere in our bodies, in our human experience of reading and writing and speaking and listening, our haphazard bike rides and walks through crowded cities, the sounds of nature and the music in our ears and the feel of the wind and the texture of sand – all these deeply human things that we feel with our senses, a means of understanding unavailable to AI.

Then, the prose must be sculpted by spending “a stupid amount of time on it.” My own experience tells me the same: only time and effort, honest toil, gives our prose integrity, like a bicycle wheel spinning true, something we can easily recognize but not easily fake.

The promise held out by the people bringing AI to market – endless shortcuts, effortlessness, an end to toil – is not one I can believe, nor is it the world I want to see.

Since I do not want to dwell on the worst case here, let me lay out the best case: AI may free us to spend more time rewriting. It is then up to us to work harder to write better – and also to build the communities that will sustain our practice.

 Community, communities

My argument is that community is a vital but oft-neglected sibling of those rarefied entities that keep one away from isolation and despair; it rests right up there with cherished friends, a partner who loves you, a family of some stripe who love you back, a passion or commitment that gives you juice through the days.

First let’s admire this sentence. It distills complex ideas into a unified statement with an admirable parallel structure, while resisting simplification (of some stripe); what is more, as Plett does throughout her essay, it slips in an entirely unexpected but somehow perfect word – juice! – that make the discussion of an abstract idea feel immediate, organic, carnal, human. This is not because Casey Plett is a person; most human prose fails to achieve this. It is because of the great lengths she has gone to, the “stupid amount of time” I can tell she spent on this short essay and on that one banger of a sentence. But I can also feel the unstinting efforts of the book’s designer, editor, copy editor, and typesetter (who are credited), along with the countless others who are not, from high school English teachers to librarians right through to the booksellers who chose On Community as one of the few English titles in my Quebec City bookstore, where it waited for me on a shelf. All these people together form a community (or communities) that we as translators are part of.

If a community is a group of people bound by shared interests who respect each other to a minimal degree, it follows that we recognize each others’ value. We want each other to survive. If I appreciate my local bookseller, I will not (and do not) order books more cheaply from Amazon. I go to the bookstore because I like and respect the people who work there, and recognize that we are working on the same collective project to bring books to readers, and also because, despite what tech companies want us to believe, it's nice to leave my house once in a while.

As a translator I work for agencies that use technology to depersonalize relations to the point where no community can form. In the past I worked for an agency that possessed a very strong sense of community, which was later purchased by a hedge fund that does not.

But the publishers I work with, and the booksellers, and the editors, and the book designers, do not spend their time trying to eliminate all labour and expenditure. They go to great lengths to keep the machine turning, often at a high personal cost. To survive as translators we need to throw in our lot with clients who we know, who we speak to and see in person. The dominant assumption of capitalism today is that we have no choice but to do whatever we can to pay less, to pay nothing, to embrace tech that eliminates all human contact. This system tells people that they want translations dropped off at their front door by unseen and exploited hands, just like their meals. And what will we do once all labour has been eliminated (or hidden from our view)? Stare at an almost-convincing simulated world through giant goggles? Why would I want to?

The problem is not AI but the assumptions and systems of those who create and deploy it. Against these powerful forces, all we have is our conviction and our communities. And what I see wherever I go is that no matter how bad it gets, and it is getting very bad, the bands keep putting on shows and the poets gather for readings and the gardeners garden, and as translators we too can be a community, small but resilient, resistant to eradication.