THE POETICS OF SPAM (SPAMBEAU LIVES)

Dick Pountain/Wed 16 July 2003/10:23 am/Idealog 108

<strong>F<!ferdie!>RE<!ferdie!>E UP T<!ferdie!>HE C<!ferdie!>ASH Y<!ferdie!>OU DES<!ferdie!>ERVE N<!ferdie!>OW!<br></strong>.

Tempting offer? Well, er, no actually. I've mentioned here before that I deal with spam by diverting almost all HTML-format email straight into my Ameol Trash folder. Sometimes I skim the headers before purging the whole lot, and was recently intrigued to discover that many of the messages are now crudely encrypted like the example above, by inserting random strings as HTML comments into the middle of all their words. Web Business columnist Paul Ockenden explained to me the purpose, which is to defeat certain spam-trapping software by disguising key words that betray the message's spamboid content - words like CASH, VIAGRA and ENLARGE (the one above says FREE UP THE CASH YOU DESERVE NOW!)

This discovery depressed me and tickled my imagination in almost equal degree. After idly wondering who the hell 'ferdie' is (does he get a royalty?) I realized that this all ties up rather horribly with a recent topic of this column, namely computers and semantics. Here we have people deliberately disguising the meaning of a message, or reverse semantics if you like. We've had viruses and worms, now we have the cybernetic equivalent of cancer, the proliferation of useless cells throughout the body of a message, eating away its meaning. Then another biological analogy suggested itself - these ferdies remind me of the stretches of junk DNA that are found in the genomes of all creatures, the accumulated detritus of long-fought evolutionary battles for survival.

This last metaphor is actually rather better than it looks at first, because computer programs share with DNA the property that their semantics exist solely in their execution. Alan Turing showed that there's no way of deducing the meaning of a computer program by simply inspecting its text (unlike its syntax which, under certain conditions can be checked without executing it). You have to actually execute it on a computer, and its meaning might be altered by the exact computing environment in which you execute it.

Similarly the 'meaning' of your DNA is you (and your children), the result of trillions of ribosomes 'executing' those strings of nucleotides trillions of times over many years. And as Steven Rose pointed out in a recent book review, listing the human genome, gargantuan task as it has been, is only a tiny first step in our knowledge because the execution of this genome is acutely context dependent. Which genes get turned on and off, and when, is partly controlled by all kinds of external factors such as climate, hormone levels, nutrition, disease, so this is like self-modifying code running on a self-modifying computer. (This incidentally also means that genetic engineering is a far more hubristic enterprise than its proponents will admit - we already knew the politicians are naive about the possibility of containing GM crops in the field, but the truth is that the scientists are being equally naive about the un-anticipatable interactions that may take place during such an unimaginably complex execution.)

It's exactly this same extreme context dependence that prevents, and will in my view always prevent, computers from understanding more than a tiny, grossly simplified subset of human meanings. The context nowadays is the whole world, and the computer would have in effect to be a human being - with similar sensory inputs, memories and history - even to understand the simplest joke. In fact humans only imperfectly and with effort understand each other. It all boils down to the distributed storage of information - each human's skull contains not only their 'consciousness engine' but a database of images and memories from their whole life, though these cannot be directly shared with other humans - they have to be translated first into language and spoken or written. The phantasy of computer intelligence stems from the true fact that computers can exchange information with each other far faster and more directly that humans can - but the problem is they don't possess, and can't realistically ever possess that database of experiences that comes from living.

I've just been reading the American literary critic Helen Vendler on the poet Wallace Stevens. In one lecture she re-constructs one of my favourite Stevens poems 'The Emperor of Ice-Cream', which I've read and enjoyed for years just for its word-play and suggestive imagery. Vendler shows how by inventing a 'back-story' that's mostly absent from the text of the poem (a country wake, dead woman in the bedroom, young girls churning ice-cream in the kitchen) an extra level of meaning emerges concerning life, death, and the necessity of choosing life. I liked the poem before, now I love it. To invent that back-story though you needed to know all Stevens other poems, plus those of Keats and a few others.

That's just a rather extreme example of the problem we face every time we communicate with another human, and why computers will never be very much good at it. You might object that poetry isn't a good model for down-to-earth, hard-nosed business communication, that it's deliberately obscure and artsy-fartsy. One possible retort is that a world in which only hard-nosed business communication is allowed would be utterly intolerable, but you'll find that you have the modern business community against you too - the advertising industry increasingly poaches on the preserves of poetry and drama, because it knows that you can sell people things by manipulating their emotions. The Emperor of McFlurry perhaps.