My REAL website is here:

Friday, January 11, 2008

machine translation divination

For those of you who have been longtime subscribers, you may remember that a while ago, when Sloppyslurper Slurpysip was located in a different place, I spent a lot of time uploading all these experiments in machine translation (i.e. Google translation, Babelfish, etc.). (Looking back at them now, they are so beautiful: here's one about the mass mind project). Specifically, I was using machine translation not for its stated purpose of automated computer translation, but as a way of processing language, destroying language. I know that there have been poets who have looked at this process, but there are a lot of challenges that I think have yet to be addressed, perhaps because the right approach has not yet been found. I'd like to explain why this fascinates me.

I am interested in this process not as a way of creating language that is beautiful to the ear of the poet, but as a way of crossing a mechanically rational logic with a natural evolutionary logic to divine a new kind of knowledge that could never have been known before we had these tools and these methods. I think the word "divine" is correct, as by crossing the electromechanical and biological minds we are creating a synthetic third kind of mind of whose existence we are barely becoming aware of. This perhaps has the potential to become a transcendent gnostic process.

Recently programmers and linguists have put their heads together in a concerted effort to create mathematical formulae and algorithms that attempt to describe language in numbers to feed into our number-crunching machines. This astounding collaboration is one of the peaks of our present technological abilities, and one can only laugh at the fact that the main impression given by machine translated text is one of gross limitation, failure and absurdity. But this is perhaps because we are judging the text by human standards. Our cognitive demands for grammatical agreement and word order is not met, so we judge the text to be nonsense. But in this we are being unfair. We should judge the texts by the computer's standards, in which case we would see that to the perfectly functioning digital mind these texts are proof of a formula correctly completed, electrons shooting from power to ground. All the words are beautiful.

To the machines it must be our strange, organic, rule-defying and highly-redundant language that is dissatisfying. The fact that there are many different trees of language, each with its own branches of diverging dialects must be quite an ugly phenomenon to the chip-eye. Of course we know this to be untrue, that language is one of humankind's crowning achievements that has its own intrinsic worth and beauty, its own aesthetic system by which it is judged, but moreover it has its own secretly encoded knowledge which is implied in its etymology, its history, the knowledge that has been injected into the language through thousands of generations of humans who lived and died dealing with the challenges of communicating, warning each other about danger, gossiping about sex, insulting each other, forgetting words, misinterpreting subtleties, making up new words, rhyming, playing word games, speaking in tongues, teaching strict rules of grammar, etc.

If we cross these two systems of knowledge and standards of beauty two things will happen. First, there will be a decoding of knowledge that is hidden from us if we look at the text with either our human eyes or the chip-eye. Second, there will be a synthesis of the two systems of language aesthetics to create a new way of assessing the worth of the text, the eye of a greater judge. I can't explain precisely what I mean by "crossing." An analogy might be how a mathematical equation can be solved by performing the same operation on both sides of the equal sign. Mistakes are perfectly OK as long as it's balanced over the equal sign. In fact there aren't really any mistakes; some operations bring us closer to the solution, and some draw us away. In the same way, if two people are talking and they both make the same mistakes with their grammar they will not perceive any divergence from order. In fact, if they respect their mistakes a new language begins its life. This is true not only with grammar, but with agreed-upon facts. I can't tell if these analogies are useful, but I hope they point in the right direction.

There are two questions about this process. We are currently making inroads in the solution of the first, the second is perhaps one of those imponderables that drives theosophists, metaphysicians, and philosophers of science mad (and that provide them with their jobs...). The first question is; what does it really mean to "cross two systems of knowledge"? Of course we can talk about crossing two systems of knowledge from different cultures (i.e. French impressionists being influenced by Japanese ukiyo e prints), but this is surely the philosophically simplest way of approaching this problem. Consider what it means to synthetically and simultaneously understand both how the cat understands the mouse and the mouse understands the cat... Not to have both understandings, or reduce to commonalities, or to find compromises, but to naturally have both systems of knowledge. Naturally, not forced. I think this is where machine translation becomes interesting; our language is presented naturally to the machine. Then the machine performs its operations on its data and speaks back to us in its language. We then hear it and perform our operations on it with the language portions of our brain. Both minds are acting completely naturally, free and easy. Nothing is forced.

The second, imponderable question is whether it is possible for us to understand these texts. To appreciate them for their worth. If any mind can understand the workings of a mind different from it. Whether the ant can understand why it carries its load, whether humans can communicate with the mass mind. And this leads to a second question; if we can't understand the aesthetics of the higher order, how will we know when we have created something beautiful? If we don't understand the knowledge of the higher order, how will we know when we have said something true?

Regarding divination I would like to say something; while I think this process is an (admittedly feebly primitive) way of dealing with higher states of consciousness, I don't think that there is anything mystical about this. At least it should only be as mystical as the contemplation of the unfathomable mystery that is our own consciousness, our own understanding of beauty and knowledge. I would like to smooth the sand so that any small steps in the right direction will leave clear footprints.

Now, back to Earth. Please don't feel let down by the gulf between theory and practice.

I have been working on a project, which, as all the most fun projects do, has led me astray. I started off trying to create a text for a mass mind experiment of a sort, and ended up using google translation to destroy a strangely oversimplified script that came to me one night which dealt with transhumanism and gnostic crossing over. In pursuing this I strayed further and further afield. I made an interesting discovery which hinges upon a third level of misunderstanding. I tell secrets on this blog, so I'll let you know what it is, even though a secret revealed can seem like a small thing indeed.

My new method of machine translation divination uses the following algorithm. First, translate the original text from English to Japanese. Then copy that Japanese text and translate it from Chinese to English - this is where the new layer of synthesis enters; the computer translates the Japanese versions of the Chinese ideograms which are separated by almost a thousand years of divergence and cultural difference (additionally the computer assigns numerical values to the non-Chinese Japanese letters and translates some of these as well). Finally, copy this text and translate it from Japanese to English to clean up the remaining Japanese characters. This is the final text.

A word about the original text; considering our imponderable question re understanding the resultant synthetic text on the higher mind's terms, I think that the path to understanding lies through using an intuitive understanding of the original as a key to unlock the final text. Therefore, partially to show you all that I am not altogether serious in my posthuman misanthropy, I have provided below a translation of Mario Savio's famous speech from the free speech cafe here at Berkeley.

One more thing. For reasons that may be apparent if you consider what I said about the imponderable question above, I am not going to analyse the final text in an attempt to understand it for myself, for all of you, or for the benefit of the chip-eye. But I must say that "random" would be an inappropriate adjective to describe this text. It is an essential part of the process that the computer off in Mountain View (or wherever) that calculates this text has no free choice. It will always return the same result. It is this that makes the synthesis possible. It is always an accurate "reading." For example, the Wikipedia article I copied this quote from had a missing apostrophe, fill it in to witness the butterfly effect in action!

English original;

There is a time when the operation of the machine becomes so odious, makes you so sick at heart, that you cant take part, you can’t even passively take part, and you’ve got to put your bodies upon the gears and upon the wheels, upon the levers, upon all the apparatus, and you’ve got to make it stop!

Japanese translation;

時間があるときには、マシンの操作になるので、憎むべき、そんなににより心臓の病気では、カントに参加することは、受動的に参加することはできません。さえ、そして君は遺体の上に置いてくださいギアとする車輪は、レバーの上、その上のすべての装置、および停止することに就くべき!

English translation from Chinese (very similar to below...):

Time when the earth and the nation, is, Solomon treesンtheir operation Lo you willでtheir hate ever humiliating, Are you better than Lo Lo their heart is illで, KafンLot Lo participated in Able! Things is by moving Lo! Participate therein and the turn isでAnd letん. Me Him, And Jun is the body of Our Home! But most do!ギAllah! Things and the wheels are, Has the browserバ, And on earth to speak of their devices,おO mankind! Stop Able to do things on the most humiliating!

Cleaned up translation from Japanese:

Time when the earth and the nation, is the Solomon trees on their operation Lo you will in their hate ever humiliating, Are you better than Lo Lo their heart is ill, Kaf-Lot Lo participated in Able! Things is by moving Lo! Participate therein and the turn is, And let me. Me Him, And Jun is the body of Our Home! But most do! formic Allah! Things and the wheels are, Has the browser version, And on earth to speak of their devices, Please O mankind! Stop Able to do things on the most humiliating!

Amen!

PS - I would be happy to see any texts that you generate using this process!

1 comment:

Henry said...

how did that get so religious? sounds like some fishy religious hacker made the translation machine into a god machine!