Thursday, April 10, 2008

Poetry in the Machine

In case you've missed it, there's a fascinating conversation about poetry going on over at Space Bar's blog, building off an experiment by Vivek which you can read about on his blog.

You'd think that having left two llooonnnggg comments on that post I'd have exhausted all I had to say on the subject (well, at least you'd think that if you were new to this blog and knew absolutely nothing about me), but obviously that isn't true.

In particular, I want to focus on Vivek's claim that he's looking to humanize mechanical processes, in partial response to the work of poets like Christian Bok who are interested in a more 'machinic' view of poetry. While I think that's interesting, I think there's a middle ground of human-machine interaction that's being ignored. I too have issues with the idea of replacing the "lyric voice of a human agent" with the "actions of a machine" but I think there's an exciting set of intermediate possibilities, through which machines (or more correctly, mechanical processes) could be set up in dialog with human agency to accentuate the lyric voice. In particular, I think greater attention to mechanical processes opens up the possibility of a whole range of new 'formal' challenges, based on outsourcing parts of the poetic process to mechanical operation in order to focus on the rest (you can tell I used to be a consultant, can't you?).

The way I see it is this: the essence of formalism is constraint - the poet voluntarily relinquishes certain choices that should, by rights, be available to him or her, both so as to make writing the poem a greater challenge and in the hope that the adoption of these limiting parameters will help to sharpen what lies within them, making for a better poem, or at the very least, a better poet. Rather than make those choices himself / herself, the poet allows them to be predetermined by invoking some established rule or template, thus eliminating human agency from parts of the poetic process.

But what if, instead of invoking some preset rule or condition, we were to randomize some of the choices behind the poem by entrusting them to a machine? We could then imagine a whole set of 'new' constraints that we could attempt to write poems within.

Consider two examples, off the top of my head. Suppose you were told that you had to write a poem but would have no control over line order. That after you wrote the poem a computer (or some other randomizing device - technology is not important here) would rearrange your lines in a completely random way. The challenge would then become to come up with a poem (presumably a short one) that would work in every possible permutation. Of course, no computer would actually be needed to rearrange your lines once you were done - the point is that the idea of mechanizing part of the poetic process (line arrangement) introduces a new kind of constraint that makes the task of writing the poem more challenging.

Or, to take a second example - imagine that you took the database of lines that Vivek currently has (or any other set of lines) and used a random number generator to assign a number to each one of them. You then took the lines corresponding to numbers in the top 5 percentiles (say) and tried to write a poem using them. Let's say (to make things more challenging) you forced yourself to follow the rank order of the lines - i.e. you started with the line that corresponded to the highest number, then got to the second highest and so forth. You'd be free to insert lines in between, of course - the challenge would be to create a credible poem that incorporated a randomly generated sequence of lines (okay, okay, so I know it's a bit like the 'Whose Line' game on Whose Line Is It Anyway?, but then I've always maintained that Whose Line is a rich source of poetic ideas; it's also - to give it a more respectable lineage - very much like what Calvino does in Castle of Crossed Destinies with Tarot cards).

The point is not that these exercises are particularly brilliant (they're not - it's one in the morning and I'm exhausted) but that what the mechanization of parts of the poetic process makes possible is the creation of a virtually infinite set of constraint conditions that poets can use to challenge themselves if they're so inclined.

Personally, I remain unconvinced that all this desperate search for constraints is worth it. Oh, I think it's fun to do and entertaining to read, and I see how it helps to hone one's skills as a poet. I'm just not convinced that what comes out of it is particularly good poetry. But that's a whole other discussion. The point of this post is only to suggest that we may not want to dismiss machinic processes so easily, but may want to embrace them and subvert them to our own uses. And to propose that we try and think up new and interesting ways of doing that.


equivocal said...

I'm still not certain if you've misunderstood me. I've got nothing against machines, right? Machines were used in the organising and collation of this poem, and moreover, a friend has been developing text processing tools that will help humans come up with arrangements of the lines that they like. Although human agency was an integral part of the process, no machines were harmed in the making of the movie.

It's precisely those intermediate possibilities that I'm fishing for when I say that Bok speaks of "machines writing" when what he is actually talking about (or should be talking about) is human-machine interactions. "What if, instead, we began to investigate the organic, contingent, even willed, nature of machines, to conceive of machines as composed of and animated by humans"?

As Bruno Latour says, it is not that we love machines too much-- the problem is that, finally, we are unable to love them *enough*.

equivocal said...

Ie., what I meant when I said that we should humanize mechanical processes and not the other way around, is that
instead of using machines to "replace" or nullify the emotional lyric voice (the end of emotion in literature is also part of Bok's polemical agenda, even if his poetic works can be full of feeling), we should interact with and through machines and mechanical processes to extend the range and possibilities of the lyric voice, and of forms of feeling. Whether people think this is possible is a different question, and the answer can only be found through practice. But this, in retrospect, in a small, tentative, low-tech way, was one of the things this performance was trying to do.

equivocal said...

At Latour's website you can find the first chapter of the novel (it is indeed a novel of sorts) Aramis, Or the Love of Technology, but unfortunately only in French. But here is a description translated into bad English:

"Aramis is a very high tech automated subway that was developped in France during the 80s; after its sudden demise, an investigation has been requested in the reasons of this failure; the book is the scenography of this enquiry that aims at understanding what happened to Aramis, at training readers in the booming field of technology studies and at experimenting in the many new literary forms that are necessary to handle mechanisms and automatisms without using the belief that they are mechanical or automatic.

[Emphasis mine.]

And here as compensation, on another somewhat related topic, to get a sense of his writing style when well translated, is the delightful and profound first chapter of his book of essays, Pandora's Hope, entitled, "Do You Believe in Reality?":

"To put it even more bluntly, science studies has become a hostage in a huge shift from Science to what we could call Research (or Science No. 2, as I will call it in Chapter 8). While Science had certainty, coldness, aloofness, objectivity, distance, and necessity, Research appears to have all the opposite characteristics: it is uncertain; open-ended; immersed in many lowly problems of money, instruments, and know-how; unable to differentiate as yet between hot and cold, subjective and objective, human and nonhuman. If Science thrived by behaving as if it were totally disconnected from the collective, Research is best seen as a collective experimentation about what humans and nonhumans together are able to swallow or to withstand. It seems to me that the second model is wiser than the former."

Which is a definition of research that echoes very nicely with this commentary on what bpNichol was up to:

"The Toronto Research Group (which was bpNichol and Steve McCaffery) gave me what I currently find to be the best term for the less traditional artistic endeavors that occupy my time. I have always disliked the term “work” (if your art is work, find a different art), and cumbersome compound phrases (like the just-used “artistic endeavors”), and felt there was always something marginalizing about the terms “experimental” and “avant garde”. But the term “research” carries with it a much more accurate set of connotations. Experiments often blow up in Beaker’s face, but research is an exploratory mapping of a knowable territory."

Sorry to bore. I'll say no more.

Falstaff said...

equivocal: Oh, I didn't think you had anything against machines (though using machines in the way you describe is irrelevant to what I'm talking about here) though it certainly wasn't clear to me from your post that when you talked about humanizing machinic processes you were interested in these intermediate processes. I'm still not sure I would call this humanizing the machine or conceiving of it as composed of by humans - to me it's about letting machines be machines but 'collaborating' with them by breaking up the poetic process into its component parts and entrusting some of them to the machine. At any rate, it sounds like we're in agreement on the specifics, so let's not quibble about how we describe it.

I suppose I should eventually get around to reading Latour. I must confess I find his description of Research puzzling - it certainly doesn't accord with my view of what Research is about. I suspect the real distinction is between 'academic' research, which is interested in developing theory and
research in practice, which is interested in developing products / innovations. I'm willing to admit that Latour may be right about the latter being wiser in developing innovation per se, but it does suggest that trying to theorize from the latter kind of innovation is a little pointless. So on the whole, contrary to what Dan Waber says, I think 'experimentation' is a far more accurate term than 'research'.

Space Bar said...

About randomising.

This is with regard to a film I'd read about a long time ago - it came out of South East Asia as so many of these films do. I have never seen the film and I may honestly be imaging the whole thing because it has such an apocryphal quality to it. The film could apparently be played in any order at all. That is, no reel was numbered. It could be played in any number of permutations and what the result would be just boggles the mind.

Does this mean that the film was shot at random? Or edited reel by reel at random? That no thought was given to what the shape of the whole might take? Or did it mean that every permutation was considered? Did the director and editor just decide to make each reel sufficient in itself and the film was merely a matter of juxtaposition - rather like deciding which poem should follow which one in a collection?

At the Institute, one director made a film that he worked on for so long, editing and re-editing (a film that others found so obscure) that the joke was that you could play either of the two reels first and the film would still make sense - whatever that was.

Is it possible to get so entangled in possibilities that the result is really only a refusal to choose rather than something coherent? In other words, isn't it possible to confuse incoherence with depth?

equivocal said...

Ah, now we're really getting into Borgesian territory!

Cheshire Cat said...

I like the anecdote (anecdote?) about the Toronto Research Group. Poetry is research, more than research is research.

I might have said this before at some point, but Raymond Queneau's "One Hundred Thousand Billion Sonnets" implements something close to the "invariance under permutation" idea. Queneau was a combinatoricist and interested in the combinatorics of poetry; of course, poetry involves the combinatorics of language, and language is useful because of combinatorial possibility.

Maybe randomization isn't the best example of an algorithmic process, but the point stands. I'm skeptical that algorithmics can aid a lyric poem, but as to whether they can produce interesting poetry, why not? I have no idea what sort of poetry I might find interesting, though once I do find it interesting, I'm sure to be able to say something in justification.

Some might find this interesting.

It's certainly possible to confuse incoherence with ambiguity. But then again the ambiguity of a failed poem surely has something to do with our fishy friend Intentionality. William Empson was a master of ambiguity, and he computed the meanings of all his poems in advance. He's not a bad poet, for all that, which is mysterious and satisfying.