Overweening Generalist

Tuesday, June 26, 2012

More On Translations (NOT: "Moron Translations")

Searle and Future AI Robots
In the previous installment here, I briefly alluded to Prof. John Searle, and it was because his philosophical invention of the "Chinese Room" still is, for me, a Thought Experiment that I've found exceedingly stimulating. As I see it, as of today's date, I think he makes a strong case for the impossibility, in principle, of the Machine Language translation to have anything like a human mind behind it: it only executes algorithms; no matter how accurate and seemingly human-like its responses, it doesn't truly understand.

                             One of my favorite living academics, John Searle of Berkeley
                             Photo from, as much as I can glean, 2005, Mexico. Not sure
                             who took it, but I'd guess Dagmar?

Nevertheless, I'm also drawn toward the eerie converse: maybe all we're doing is executing mental logical algorithm-like strings of thought-stuff when we're operating, on all our levels of Being. One strand of thought about AI that extrapolates to some future Singularity sees the AI robots (probably fairly attractive and somewhat "graceful" in bodily movement?) as very adept intelligences in some ways that are far beyond ours, while we Humans are still able to do things (with humor, irony, and empathy?) that the AIs don't - and won't - quite get. Intuitively, this line of prognostication makes sense to me.

So, the experiment I ran using Michael Chabon's line from a recent essay in the New York Times on Finnegans Wake and H.P. Lovecraft was AI "machine mind" attempting a series of translations. I wonder what it would look like if a series of humans who were skilled enough to translate English to Chinese, one who could translate Chinese to Arabic, another person Arabic to Spanish, one rare wily one for the Spanish to Ukrainian, a Ukrainian to Dutch person, and then someone taking from that entire series of translations, with Dutch in hand, translating it back to English? At present I lack the funding, the research grant, or lots of Dad's money to conduct such a whimsical experiment, so I must be happy with the ease of something like Babel Fish. 

What Was the Damned Point?
I've not blogged on this stuff before, but for the past three and a half years, I've been enmeshed in something I think of as the Neo-Whorfian Model, which derives from the classic Sapir-Whorf Hypothesis, and which seemed to devolved to a rather robust debunking by anthropological linguists and others roughly in the cognitive sciences of Hard Whorfianism: that each language creates a different worldview. What has been labeled Soft Whorfianism was given its due (despite some rather overweening Chomskyites): the language we speak shapes the cognitive "realities" we find ourselves in. 

                                Lera Boroditsky, one of the most fascinating researchers
                                      in the revived field of linguistic relativity 

The Neo-Whorfian stuff seems to be waxing now, and I'm currently in thrall with people like Lera Boroditsky and George Lakoff. I think language seems a strong part of a total environment, and neurologically it's instantiated/"mapped" on neural circuits that other parts of our brain seem to think "are" the "things" that are wired in as circuits there, in brain-matter. But maps deceive. They also give different, abstracted-from-existential/phenomenological "reality" perspectives, depending on the intentions of their makers. In my playful experiment, I imputed intentionality and mind to the Machine. This seems analogous to trying to communicate in a foreign country and not quite having the adequate language, a situation I've been in many times, one which seems always sobering if bracing. With all the pantomiming, pointing, mispronunciations, and frenzied attempts to use the Phrase Book in real time, I don't know if the future AI-bots will be better than us, or we'll have augmented ourselves to keep up with them. Maybe they'll just kick our asses. (I just now recalled a brilliant book I read that speaks to this. I read Slaves To The Machine right after it came out, and it had a drug-like effect.)

Finnegans Wake and Translations: Strange But True
Simply put: I find the very idea confounding. And yet there it is: Finnegans Wake was translated into French, Dutch, German, Polish, and wildly (and absurdly?) enough, Japanese and Finnish! I have no comment, but wonder who, how, why, to what end, what do readers get out of these...?

I must surmise that this proves there is no limit to what a certain sort of linguistic freak will attempt. Which I find totally wonderful.

Next thing you're gonna try and tell me is that they're going to get a Chinese version of it. Yea, right...

Possibly the Wake represents a Third situation, juxtaposed with the two alluded to above (humans and AI using language to translate/communicate). To briefly play with this idea, let us see Joyce's book as the apotheosis of extremely dense wordplay in the Novel. Many devices the Novel (and its writers) has/have developed are used therein. It requires above all Time, but pays off in overflowing information flow-through in the nervous system and its wetware architecture; the Novel's ability to create Other Worlds in the minds of its readers may still be unparalleled. So might the Novel's ability to develop empathic circuitry be nonpareil. Whether any one given reader of the Wake sees this limbic tropism occur within themselves seems peculiar to that reader; for many others Joyce's "claybook" appeals to the musical sense, and the frontal cortex and its abilities to Solve Problems...Not to mention something variously labeled "The Creative Self" or "Artistic Mind."

Short Commentary on the Wiki for Books Most Translated
I take it the Jehovah's Witnesses see translation as a major strategy. Pinocchio comes in at number three? It still needs to be translated into another 152 languages to surpass the Witness tome. No one's catching The Bible. I checked the Vegas odds on Pinocchio  or any of the others in the Top 10 surpassing The Bible by 2050, and they'll give you 300 to 1 odds for The Little Prince, but I really don't have a dog in this fight. Verne's 20KLeagues is a truly fantastic book, but what makes it so much better than any of his other stuff, much less anything by H.G. Wells? I can't account for those who translate for the rest of the world, and looking at this list makes me feel awfully parochial for some odd reason. The various translators themselves? Worldwide? Bless them with fine drugs, fantastic sleep, transcendent sex, and much laughter, for what they do is an act of love towards humankind...(Yes, even the army of Jehovah's Witnesses translators. I mean what the hell, right?)

Finally, I must confess that Paulo Coelho's 1988 book The Alchemist, tied with Harry Potter (!) at 67 languages? I'd never heard of it. Now I feel I must get to that book before the newly minted summer's up.

Re: Tom Robbins
He said Robert Anton Wilson turned him on to the idea of reading Finnegans Wake. Robbins has kept a copy bedside for years, has only read a few pages before nodding off, but for him, the Wake engenders wonderfully weird dreams. He said something like this in a documentary film on Robert Anton Wilson titled Maybe Logic

In the previous blog on translation, I ended with Robbins. In his book Wild Ducks Flying Backward: The Short Writings of Tom Robbins there's a short piece called "Lost In Translation" in which he gives a passage from his 1984 book Jitterbug Perfume in its original (it's hilariously, tomrobbinsy surrealistic in its anthropomorphizing of vegetables), then a "back-translation" from the Czech (something analogous to my weirdo experiment with Babel Fish), and Robbins seems quite okay with it. Some of the "meaning" from the original surrealistic vegetables' attributes and aspirations (especially the beets!) has been altered. But what of it? It still swings!

Robbins thinks some translations are "more precise" than others. It seems no one can judge this more accurately than the writer of the original, although I've seen some fascinating arguments that question this. Here's the end of TR's piece on "literal back-translation":

I've been told by bilingual readers that until recently, when the insensitive publisher unduly hurried the translator, I've been reproduced in Italian with scant loss of meaning or intent; but that the Mexican version of Even Cowgirls Get The Blues is quite "sleazy," a description that probably doesn't displease me as much as it ought.

Incidentally, Jitterbug Perfume in Czech back-translates into Perfume of the Insane Dance. I'm not sure but that I don't prefer that title to my own. So, if much is lost in translation, something on rare occasions may also be gained. (pp.209-211)


Eric Wagner said...

Great stuff, as usual. It reminds me of the story that Stanley Kubrick would have translations of his films translated back into English before he would approve their release.

michael said...

I didn't know that about Kubrick. I wonder why he did that. If you remember where you read it I'd be much obliged.

Today in the shower I was wondering about The Great Books of the Western World set, and what it would be like to read all those that were originally written not in English in our best machine translation. (I understand Google Translate is better than Babel Fish, for example.)

The semantic meaning of a word within its own language, over a mere 200 years, can drift..."significantly"? (Ha!)...much less a word closer to 2000 years old. So a machine translator that would do a passing job for Dante (800 or so years of Italian) would have to be programmed by humans with databases for 800 year old Italian words before the Italian-to-English button would be pushed. (Then, no matter how "good" the translation, if you listened or watched Lera Boroditsky, depending on the language, there might STILL be some big gaps...)

Otherwise, the trans wd probably read as fairly absurd...which might still yield interesting results?

The idea of a large enough group of programmers who'd concentrated on historical linguistics and archaic lexicography seems doable. I can foresee an edition of Lucretius that sombunall classical scholars might accept as a passable and legit and an at-times intriguing version, come out.

Then again, even if The Machine comes up with a duplicate of some Adept (say, a 21st century version in English of Cervantes that matches word for word the latest published translation "for modern readers")), then is it still a novelty? Borges's story about Pierre Menard suggests that, yes, it is novel. But because Menard was a human, and decided to rediscover the Catholic faith, fight the Moors, forget European history - not to forget knowing Spanish VERY WELL...he writes an exact copy of Don Quixote...without knowing the original?

Wha? It's more complicated. The latter text is somehow RICHER than Cervantes's.

Borges, in that piece: "The text of Cervantes and that of Menard are verbally identical, but the second is almost infinitely richer." Why? As Andre Maurois writes, "The Quixote that we read is not that of Cervantes, any more than our Madame Bovary is that of Flaubert. Every twentieth century reader involuntarily re-writes in his own way the masterpieces of past centuries."

I hope I'm not destabilizing any of my Readers' cherished notions of the "sanctity of the text." If so, I apologize for any damages caused, and hope to make it up to you in "the future."

This idea, when extrapolated, renders all translations - machine or human - as Something Else, eh?

Eric Wagner said...

I do not remember where I read that story about Kubrick. I used to read a lot about him, and the books have blended together in my mind.