I am not sure if I would be interested in machine-generated poetry, or poetry at all, if it wasn’t for reading the work of Racter.
Short for “Raconteur,” Racter was touted to be the first computer program program to write a book. Called The Policeman's Beard is Half Constructed, Racter’s literary debut and swan song was published by United Artists Books in 1984 and went out of print shortly thereafter. New copies fetch a good price on Alibris and Amazon.
I used to keep a copy of The Policeman’s Beard is Half Constructed in a desk I had when I worked nights at the Rutgers-Camden's Paul Robeson Library. I would read passages,
At all events my own essays and dissertations about love
and its endless pain and perpetual pleasure will be
known and understood by all of you who read this and
talk or sing or chant about it to your worried friends
or nervous enemies. Love is the question and the subject
of this essay. We will commence with a question: does
steak love lettuce? This quesion is implacably
hard and inevitably difficult to answer. Here is
a question: does an electron love a proton,
or does it love a neutron? Here is a question: does
a man love a woman or, to be specific and to be
precise, does Bill love Diane? The interesting
and critical response to this question is: no! He
is obsessed and infatuated with her. He is loony
and crazy about her. That is not the love
of steak and lettuce, of electron and proton and
neutron. This dissertation will show that the
love of a man and a woman is not the love of
steak and lettuce. Love is interesting to me
and fascinating to you but it is painful
to Bill and Diane. That is love!
then read something my creative writing professor recommended to me. John Ashbery’s Self-Portrait in a Convex Mirror stands out a as good example, or maybe Wallace Stevens. I would then marvel at how, at least on the surface, this computer program, “written in compiled BASIC on a Z80 with 64k of RAM,” could be more inventive than my own work.
It was, to be honest, vastly depressing.
Back then, even with all the science fiction reading under my belt, back issues of Omni strewn across my room and progressive rock posters taped to my walls, it never occurred to me that there was a human somewhere along the line in Racter’s compositions, either keying in the program, coming up with the idea in the first place, or suggesting words for the machine what words to spit out.
I think I wanted to believe a machine could write these poems, if for no other reason than to believe I could build such a machine myself.
“A poem is a small (or large) Machine made of words,” William Carlos Williams writes, after all.
But Racter was a program, not a machine, and one authored, or written, by William Chamberlain and Thomas Etter. When I tried to write about Racter a few years ago, I grew interested in how folks in the Artificial Intelligence communities were quick to point out that Racter’s program parameters were “artificially tweaked” to generate more “literary literature.” Of course it had to be tweaked, I would say to myself. What’s even more interesting perhaps is that those initial critics of Racter couldn’t get over the fact that the success of the book, or its continued fame, is really more about how those words that were put in the program were better or more creative than their own. Here’s the most famous criticism of Racter, from the Web site Robot Wisdom--
None of the long pieces in the book could have been produced except by using elaborate boilerplate templates that are not included in the commercially available release of Racter. Nor does [Racter] include any sort of 'syntax directive' powerful enough to string words together into a form like the published stories.
Racter, the writer says, is ‘writing degeneration more than text generation.’ This truth is further disguised by using templates that are themselves 'wacky,' leading one to attribute to Racter a style that's really Chamberlain's.
Well, what style were we aiming at? The tip of the finger of God has to appear somewhere, right?
Right now, on another open window in my computer, I’m typing into a chat bot, which basically answers inputs with versions of what you wrote or the programming team. You could type in “Your mom wear combat boots,” as I just did, and Claude, one of my favorite chatbots, replies, “Hey there, now I would never want to do anything to harm my mom. They are just so gosh darn purposeful!”
The interest or novelty or sensibility or appeal of artificial intelligence collaboration has to do with all of the random texts coming at us everywhere—the Fox News ticker, Google search results, Twitter feeds, instant messages. When one considers this, the notion of sole poetic authorship, computer program or human, may seem almost ludicrous. I am of course barking up the same trees of others who are far more learned and eloquent about this—Kenneth Goldsmith, for example.
On the other hand, and I’ll be the first to admit that I also hold on to a perhaps contrary viewpoint, one that a European-American, authority-seeking, reformed New Critic would make—is that there has to be an agent of order, the mark of a style. Perhaps that is where authorship, as tenuous as it is, remains.
A section from OuLiPo: A Primer of Potential Literature sticks in my mind. Raymond Queneau, in his section explaining the N+7 method, which consists in taking and replacing each noun with the seventh following it in a dictionary, writes that the results from these experiments are “always interesting” and “sometimes astonishing.” And I agree. But Queneau says something else, and I used to think it was just a joke, but in my own practice of using N+7, I have found it to be true.
“It seems,” Queneau wirite, “that only good texts give good results.”
One correlative to this I might add is that a writer who produces those good text has to be willing to submit his or her text to a manipulation, be it artificial intelligence or a BASIC program or the OuLiPo’s “prop for inspiration.” The writer must be willing to collobarate with the machine, to take the leap over to a new text. At this point, I think not all writers will integrate or submit their text to a machine or program as part of their process, but to me, it’s no different than James Merrill and David Jackson looking for answers in a Ouija Board.
Writing in 1950, Alan Turing proposed a test, now called the Turing Test, as a way to determine if a computer program could pass for a human being. He proposed that a judge could talk to a computer and man over telephone lines, and if by conversation alone, he could choose the human. If the judge picked the machine, then that machine would be considered to be artificially intelligent.
Alan Turing predicted in 1950 that in about 50 years “an average interrogator will not have more than a 70% chance of making the right identification after five minutes of questioning” in an imitation game.
On November 8, 1991, eight years early before Turing’s prediction, the first Turing Test in the world was held at the Boston Computer Museum. Eight computer programs vied for the The Loebner Prize, and communicated through modems, along with two human controls also at remote locations.
One computer won.
“PC Therapist III,” programmed by Joseph Weintraub from Thinking Software in Woodside, New York, was the first computer program to pass the Turing Test. “The judges had only a terminal number (1 to 10) and a topic printed on a card affixed to the terminal,” Weintraub’s writes in his account. “Some of the topics were: Whimsical Conversation, a dry martini, Shakespeare, romantic affairs.”
Five of the ten judges decided that PC Therapist III, running on a 396PC and engaging in “whimsical conversation,” was no different than a human.
Whimsical conversation on a computer one mistakes for a human, to my mind, places itself in the same family tree as the anthropological notion of letting the rascal take over, something Lewis Hyde talks about in Trickster Makes This World. DuChamp, Allen Ginsberg and Bugs Bunny are all coyotes whose works stand in for other things; they trick an audience that is willing to be tricked, who want to believe they are tricked.
With Racter, I wanted a machine to trick me into being a poet. I wanted to be part of the whimsical conversation. And it worked.
 Ubuweb offers a PDF scan of the Racter book at http://ubu.artmob.ca/text/racter/racter_policemansbeard.pdf.