Kendall Heitzman’s introduction: “Project Babel” was published in Japanese in the Tokyo Shimbun on 26 November 2022. It first appeared in English in my translation at a reading on 22 September 2023 at the University of Minnesota. An abridged version was performed on 5 October 2023 as part of the theatrical production Global Express at the University of Iowa. It is appearing in print for the first time here.
Echoing her novels Higanbana ga saku shima (The Island Where Red Spider Lilies Bloom) and Sei o iwau (Celebrating Life), Li Kotomi’s “Project Babel” is set in a near future where things have gone (more) awry, and as in those works a linguistic shift is at the heart of the problem. Although this is a lighthearted piece originally published for a general audience, the real-world issues it taps into are quite serious ones. The notion that language should be made to conform to the parameters of machine learning, rather than the other way around, doesn’t feel all that distant in an era of “deep learning” and ChatGPT.
I have gotten into the habit of gazing out the window after I wake up in the morning. And then I say with a sigh, ah, still there, is it?
There are towers just like this one, looming over Tokyo, in all the major cities of the world. New York, London, São Paulo, Johannesburg, Beijing, Mumbai . . . every one of them with a spire disappearing into the clouds, and covered with solar panels that catch the light and blind anyone looking directly at it.
I work in the tower. It is called, appropriately enough, a Tower of Babel.
I get ready for work, unable to shake the feeling, like a dull throb, that I really don’t want to go. Every morning, as I make my way in, I start to think I am personally responsible for contributing to the decay of something important. But when I catch sight of the motto inscribed above the entrance to the tower—“For a wisdom that reaches to the heavens”—the feeling disappears completely.
Ten years ago, the push to develop machine translation ran out of steam.
At first, the scientists working in the field had thought it could be perfected. If they could just create a universal translator, they believed, humanity would no longer be divided by the walls of language, and individuals and entire cultures would know a freedom they had never known before. People would no longer be required to spend such immense periods of time on language learning; they could devote all of those hours to the pursuit of science instead. The scientists envisioned a new hope for humanity in that world without linguistic barriers. Thanks to deep learning and other innovations, the algorithms were making great strides.
But once it reached a certain level, machine translation plateaued; it simply couldn’t rise above a certain level of accuracy. In the end, the researchers were forced to admit that it was impossible for a machine to accurately translate the languages of humans.
“There are all kinds of problems. Ambiguity, high-context sentence patterns, individual idiosyncrasies, shifts in language over time, and so on,” the leader of the development team explained at a press conference. “The languages we speak are ambiguous; they rely on context to be understandable. Any given language has all kinds of expressions and formulations that are unique to it, and it will change over time. On top of all that, there are simply too many languages on the planet.”
Project Babel rose from that technological dead end. Its basic premise was simplicity itself: if there wasn’t any way to get machines to perfectly understand human language, then the solution would have to come from the human side. Humans would have to create languages that machines could understand.
“It’s really not as outlandish as you might think. When microwave ovens were invented, microwave-safe dishware needed to be developed. When induction stovetops spread, so did cookware designed for induction heating. And now, we’re doing the same thing with language. This tower will develop human language designed to handle today’s machine translation. To us, there will be no sacred cows!”
Thanks in large part to this passionate speech from the scientist who initiated Project Babel, it soon found favour in every corner of the world. To create a language that even a machine could understand perfectly, one that would never change over time—this was the dream of humanity. And by extension, it was a social justice project, designed to contribute to peace and human development. It was the dawn of a new era for language, and a resolution was adopted by the United Nations to the effect that anyone intentionally using outdated language that couldn’t be understood by machines was obstructing human progress and guilty of crimes against humanity.
The project began by trying to limit the number of languages in the world. Only the top ten, as ranked by the number of native speakers, were allowed to survive—Chinese was first, English second, and Japanese barely made it through, in ninth place. Everything else was eliminated. Needless to say, there was a brief period of furious resistance from groups advocating for the preservation of languages and diversity in general, but their voices were quelled with the right amount of pressure.
“The new human society we are creating doesn’t require thousands of languages to communicate, any more than it needs thousands of varieties of knives and forks to eat. There will always be esoterically inclined people who say that language is an art, language is culture, what have you, but the fact is, language is nothing more than a tool,” the project leader said. “I have a question for these language enthusiasts: in all the times you’ve used Google Translate, have you ever once given even a passing thought to the languages that aren’t there?”
But even with only ten languages remaining, that would still mean that a total of forty-five possible language pairings would have to be handled by the new translation system. And so, the project established a few rules:
First, a representative language was designated for each region of the world. In Asia, Chinese was chosen; in the Americas and Europe, English; and in the Middle East and Africa, Arabic. Next, each of the regions would establish translation systems between the representative language and the other languages in the region. In Asia, for example, systems of translation would need to be established between Chinese and Japanese, Chinese and Russian, and Chinese and Hindi. All that remained from there would be the work of creating a means of translating between the three representative languages.
By doing it this way, even if there wouldn’t be a translation system set up between Japanese and English, for example, one could translate from Japanese to Chinese and then from Chinese to English. With this hub-and-spoke model, the number of necessary translation combinations was reduced to exactly nine.
Now, the Towers of Babel constructed in major cities around the world are racing to produce the systems of translation within each region. The branch tower in Tokyo is responsible for translation between Japanese and Chinese, and I am a member of that team.
The moment I step foot into the office, I can feel there is tension in the air. Trying not to draw attention to myself, I slip into my seat. My colleague at the next desk leans over and whispers in my ear. “Our boy Tada screwed up again. The team leader really went after him.”
Apparently, when he spoke during a meeting, Tada hadn’t clearly stated the subject of the sentence. It has always been the case in Japanese that, to a certain extent, you don’t need to state the subject of the sentence for your listener to understand who or what you’re talking about. But without the subject clearly stated, it is hard for the machine to follow the Japanese, and thus hard for the machine to translate the statement into Chinese. Sun, the Chinese team leader, likes to trot out a favorite example: “When you say Ai shite iru, without a subject, all that we know is somebody loves somebody else. You love me? He loves her? Without a subject, we can’t possibly know.”
It was established early on that in order to create a language capable of being translated by machine-learning, it has to be pared of ambiguity and high-context constructions. In order for this to happen, speakers have to clearly state the subject, object, and everything else in their sentences. Sun has insisted everyone in the office get used to doing this as a matter of course in their daily communications. Tada’s mistake was to say in a meeting “As was kindly noted,” when what he should have said for the sake of the machines was, “As you said.”
When Project Babel is completed, will people who utter sentences without a subject really be considered guilty of crimes against humanity? A dismal thought indeed. The morning meeting is about to begin, and the thirty team members move toward the conference room.
“Going back to our conversation about foreign loanwords the other day… the regional headquarters in Beijing is aiming to significantly scale them back.” Sun’s announcement is met by a chorus of disbelief in the room: Whaaat? and Not again! Because the machine found them difficult to understand, and because they were difficult to translate into Chinese, in previous rounds of retrenchment we were forced to all but eliminate honorific language, as well as the various forms of onomatopoeia that appeared in Japanese.
Perhaps having anticipated the resistance from the staff, Sun presses on. “There are simply too many loanwords in Japanese; they get in the way of machine translation. Why are you calling a beverage a ‘do-reen-ku’ when you have this perfectly good Japanese word, nomimono? Why do you need to buy a bottle of ‘wa-een’ when budōshu will suffice? And if you can place a chūmon at a restaurant without any trouble, why are you going out of your way to call it an ‘ohh-daa’? And don’t get me started on all of the out-of-date English loanwords in Japanese. It’s bad enough that when women walk down the aisle you’re still calling it the ‘virgin road,’ but you really need to stop calling men with retrograde senses of chivalry ‘feminists’! The global standard these days is a little higher, and we need to update the Japanese language to bring it into line with the rest of the world. I know change can be painful, but I need everyone to understand how important this is.”
After the meeting is over, Sun stops me on my way back to my desk. “I’m glad to see your work on idioms is coming along, but I have a few questions for you.” Sun points to some notes on a tablet. “What does this Japanese phrase mean, ‘to get the rope’?”
“We use it when a criminal is arrested.”
“Then just say that someone is arrested. You’re not binding people with ropes these days, are you? How about this one: ‘the line is narrow’?”
“Well, it means someone is skinny.”
“How does one get from ‘the line is narrow’ to ‘he is skinny’? No one’s going to get that, are they?”
Sun impatiently raps on the tablet. “We are paying you a lot of money. And you have been hired for your expertise in Japanese. A lot of people are relying on you to take your task seriously. You need to be thinking, at every step, is this phrase absolutely necessary for the Japanese language of the future, which is designed first and foremost to be understood by the machines?”
As I leave the conference room, it is probably clear from my face what has happened; a number of my colleagues cast sympathetic glances in my direction. I return to my desk, log in to the system, and delete from the corpus of Japanese the phrases I have been instructed to excise. It’s OK, I tell myself, the pain is only momentary. When the machine-ready Japanese language has been completed, humanity will be one step closer to a glorious future. We will have in our grasp that “wisdom that reaches to the heavens” and they’ll put us in the history books. In the meantime, you won’t catch me stabbing the boss’s shield. Better to let them roll you up than snap you in half.
But then I realise, oh, we deleted those idioms just last week.
The main image and the cover were taken at the Kusama Yayoi Museum in Tokyo.
How to cite: Heitzman, Kendall and Li Kotomi. “Project Babel.” Cha: An Asian Literary Journal, 15 Dec. 2023, chajournal.blog/2023/12/15/babel.
Li Kotomi (author), originally from Taiwan, moved to Japan as an adult and quickly established herself as one of Japanese literature’s most exciting young novelists and a crucial voice on LGBTQ issues in Japan. In 2021, she was awarded the Akutagawa Prize for Higanbana ga saku shima (The Island Where Red Spider Lilies Bloom). Her 2018 novel Hitorimai was translated into English as Solo Dance (World Editions). In 2023, Li participated in the International Writing Program at the University of Iowa. Li writes in Japanese, but has translated her own work into Chinese.
Kendall Heitzman (translator) teaches Japanese literature and culture at the University of Iowa. His translation of Fujino Kaori’s Nails and Eyes (Pushkin Press, 2023) was awarded the Japan-United States Friendship Commission Prize for the Translation of Japanese Literature. His translation of Nakagami Nori’s “The Phone Call” appeared in Cha in 2018, and his translations of work by Furukawa Hideo appear in recent issues of the literary journal Monkey.