Wednesday, March 27, 2013

Guest Blogger, Erik Richardson: Computers As Poets: Could We Build a Better Byron?

Today, Erik Richardson gives us more food for thought as we continue to examine whether poetry can be governed--or created--by rules. What do you think? Right now, I am so wishing I understood the rules of HTML. I am sorry that I cannot for the life of me make this post look more uniform in terms of line spacing and such. I am not a designer. I am a poet. The two are not always mutually exclusive, but in my case, they are...

We used to say computers could never beat a human at chess. Then we thought they could never beat us at Jeopardy. Those ideas turned out to be wrong. Naturally enough, the generalization from there leads me to wonder: will computers ever be able to beat us at writing poetry? The answer is no, because there is a core human quality which cannot be captured by any rule set, and that is the ability to rebel—to step outside the assumptions and boundaries of any given set of conceptual rules.

Following closely upon that, I think we would agree that some of our most human endeavors, like art and poetry, express that quality and insofar as they do, they also can’t be captured by any set of rules. This is something Chuck Rybak touched on last week.  The computers that win at Jeopardy or beat human chess players are able to dig through thousands upon thousands of games and possible moves to come up with the best solution. This process merely represents the operation of a finite set of programmed rules operating on a large data set, but progress in poetry can no more rely on cutting and pasting ideas and phrases that already exist as parts of other poems any more than it can on recycling a given set of rules.

To illustrate this, let us reflect on the fact that Watson can be unbeatable at Jeopardy, but it cannot ever step back and decide that it has wasted its life by spending so much time playing Jeopardy and wants, instead, to spend more of its time watching sunsets, nor can it disagree with the wishes of its current programmers by deciding to use its Jeopardy winnings to fire them and hire new programmers in their place. Nor, in the specific sphere of poetry, could a computer programmed with volumes of poetic works come to realize that it detests some of the poets that have been given to it as exemplars. Byron, as a great example, showed meaningful dislike of towering figures like Coleridge and Wordsworth, and that was the cornerstone of his particular greatness. The great poet has an experience which pushes her out of the normal rules for thinking about some part of the world around her. She then twists and stretches language past its normal rules for saying something so that the reader is, in turn, pushed out of his normal set of rules for thinking about the world around him. In a move that would make great logicians like Gödel and Cantor proud, we can see that given any set of rules we might create for poetry or for robots 100 or 1000 years hence, humans will have the capacity to step outside of that rule set to contradict them, to create new rules, etc.


The computer advocate can interject here:  If poetry can be taught, it can be programmed. Is that not exactly what happens with textbooks and poetry classes? Are we not creating a set of instructions that will allow someone to write good poetry? Consider, for instance, that however poetry has changed in our imagined future, the new textbooks, or e-texts, or whatever, will give students the new rule sets they need to be good at that evolved version of poetry. Couldn’t a good programmer translate that textbook to computer code and feed it into the robots?


Certainly that’s a logical train of thought, but as anyone who has ever tried to teach poetry knows, we can’t even ‘program’ every bright student how to be a great poet. Certainly by interacting with other poets over time, some of the students can become great, but that interaction is significantly different from something that could be captured by a set of rules—even a whole stack of poetry textbooks. The fact that we can only do this with some of the students proves that there are necessary ingredients beyond what is contained in the textbooks and the professor’s scribbled menagerie of class notes and marginalia.


What is more, in the process of coming to understand minds and computers enough to have built such a robot, we—the race of robot-makers—will have changed in meaningful ways from where we are now, and that creation threshold cannot be breached. It is questionable that we will ever be able to create machines faster than we can change our thinking to dream them up.

Erik Richardson lives in Milwaukee, Wisconsin, with his family and assorted pets. The whole group is a tangle of dandelions in the middle of the suburban lawnscape (and takes pride in that). In addition to teaching math and computers, he mentors a high school robotics team and runs a small business with his wife fueling sci-fi/fantasy fandom.

1 comment:

John Peter said...

Excellent post. I want to thank you for this informative read, I really appreciate sharing this great post. Keep up your work.
online shopping for computers