Translation : <

Instead, it is the combination of statistical-machine-translating algorithm, database, and search-engine that the word “translator” referred to, and the three forms of labor the question rather foggily envisioned—the “training of the translator,” the “searching,” and the “translating”—were to be understood as instances of the same process, a process camped out in anachronistic metaphors as the most primitive form of extractive manual labor, mining, carried out by entirely non-human means. This is Och, in answer to the questionL: “When you train the translator, you’ve got to get so-called parallel data sets, where every document occurs in at least two languages. Where do you get all of those translations from?” Och says:

When we started, there were standard test sets provided by the Linguistic Data Consortium, which provides data for research and academic institutes. Then there are places like the United Nations, which have all their documents translated into the six official languages of the United Nations. And there’s a vast pool of documents available there in the database, which has been very useful because the translation quality has been very good. But then otherwise, it’s kind of ‘the Web’ . . . Our algorithms basically mine everything that’s out there.

To this the interviewer remarks, “So it’s sort of analogous to the way Google’s Web crawler spiders Web pages?” And Och answers:

It’s similar. While the Web crawler is mining the whole Web and indexing it, then for the translation crawler is the subset of documents that include translations [sic]. The challenge is to find which texts are translated into another language — and where to find the corresponding translation.¹⁴

Note the outrageously anachronistic use of the term “mining”—the wonderful and symptomatic verbalization of the noun “spider.” Remark the awkward anacoluthon in Och’s answer. We are watching the interview’s language strain to express a state of affairs for which there is not yet a lexicon or a syntax. Something, and it is neither logos nor glossa, flashes. We are, it appears, in the event.

Of course, when we tell the story in this way we are surrendering to a different sort of temptation. We imagine that an event has occurred or is occurring, and we stand in its shadow or at its advent. We moderns, like bold Cortez upon a peak, or Zarathustra at the opening to his cave, or Captain Kirk on the deck of the Starship Enterprise. Nothing about this new, adventitious time is or will be translatable into its forerunner; our world has changed; the linguistic function of “referring” differs, and what we refer to does as well, semantically and syntactically—a spider now “spiders,” an algorithm mines, and the search for “a corresponding translation” to a string of words is carried out neither instrumentally nor aesthetically, but mechanically. When we say “language,” or “translation,” different things, unimaginably different things, are designated than when the same words were used before Turing, and before the Cold War.

The temptation, then, is twofold. In the first place, to imagine the advent of machine translation in the breathless light of the messianic—or its proxies, the robotic, the inhuman. “The Android version of Google Translate allows the user to speak to the application, and have his or her words translated,” says the LA Times interviewer. “Is it,” he continues, “just one short step from here to real time, speech-to-speech translation, a la ‘Star Trek’s’ universal translator?” And Och answers:

To really do the integrated speech-to-speech translation, where you can have a phone call with someone and it would [be] interpreted live? I believe that based on the technology that we have, and the improvement rate we have in the core quality of M[achine] T[ranslation] and speech recognition, that it should be possible to do that in the not-too-distant future.¹⁵

The garishness of this science-fiction scenario tends to obscure the second, related mistake, which is to imagine that the advent of machine translation does indeed mark an epochal break with the Babelian paradigm, whose humanist heroics and whose elliptical circuit, defined by and mapping the fluctuation of value invested in the foci of technical and aesthetic labor, would seem untranslatable into the emerging lexicon and syntax of machine translation. Each, we like to think, is cut from the other by a historical caesura every bit as sharp, as unbridgeable and untranslatable, as any we might find in the Christological imaginary. But is this indeed so? Might there be other ways of rethinking translation, the concept, the history and the present of translation?

Which brings me to the second thing we will have overlooked in all this imagining and remarking. Let us say that Machine Translation has always and already inhabited the Babelian paradigm. You can imagine already the consequences this might have for the various maps in which we’ve settled what we call “translation”—the map of the market, of evangelization, of the long-cherished definitions of the human animal as homo faber, or as the animal that produces art. You can imagine what it might mean for the long linking of value to material production to claim that an inhuman commodity-practice has always and already spidered away in the texture of what seems to be most proper to human animals, to zoon logon ekhon.

We can imagine these consequences of claiming that machine translation has always spidered away at natural languages, at an angle to the twin foci of instrument, and aesthetic enjoyment that I’ve graphed—we can imagine the consequences of the claim, but what would it mean to claim such a thing? Not, of course, offering some sort of time-traveling marriage of Google with Heraclitus, the mere reverse of the science-fiction scenario I have been evoking, peopled with Androids and Star Trek translating devices. What we mean by a “machine” will have to change; what we mean by “translation” too; “history” as well—not to the point where the modern “machine” cannot serve us as a translation of the older one, but also in a way that does not stand or fall upon the fantasies of designation, of possible-world economies, that I have been outlining. To claim that Machine Translation has always and already inhabited, worked, or spidered away at natural languages would mean claiming that the difficulties the concept of translation entails work (silently or clamorously), or occur, not just across societies and at different historical moments, but structurally in natural languages, and in the space of single, determining descriptions of the processes of translation.

^{14. David Sarno, “Franz Josef Och, Google’s translation über-scientist, talks about Google Translate,” Los Angeles Times (March 11, 2010), http://latimesblogs.latimes.com/technology/2010/03/the-web-site-translategooglecom-was-done-in-2001-we-were-just–licensing-3rd-party-machine-translation-technologies-tha.html, accessed March 2012.↩}

^{15. David Sarno, “Franz Josef Och, Google’s translation über-scientist.”↩}