Poetry is scary.
It’s that bizarre art of words that everyone talks about, yet some of us have difficulty understanding. I find poetry closely related to most of contemporary art. I don’t know if it’s good or bad. I like some of it; I’m confused by a lot of it, and then there’s the kind where I can’t even have an opinion because I don’t know what I feel when I come across it.
We learn the classics since childhood. Shakespeare’s sonnets are the Mona Lisa of rhyme and verse. You’ll find them being read during a wedding, a funeral, a high school graduation, and any place where you could also see someone using Gandhi’s “Be the change you wish to see in the world” quote. If art is what makes us human, then a few stanzas are the easiest proof of our true emotions.
But, in this world where even the most delicate handmade artifacts are exchanged for their more accessible, i.e. cheaper, knockoffs, could beautiful language be outsourced as well?
In 2011 an undergrad at Duke University had a poem accepted for publication at the school’s literary journal The Archive. There was only one small detail: he hadn’t written the poem. Before you think this is a tale on the perils of plagiarizing, it’s not. Zachary Scholl had created an algorithm that could create poetry. His main interest was not to be seen as the next Ezra Pound but to find out if his program could fool people who read a lot of poetry.
Which of the two following poems do you take as the human versus the A.I.?
Poem 1
Oh, but it is dirty!
—this little filling station,
oil-soaked, oil-permeated
to a disturbing, over-all
black translucency.
Be careful with that match!
Father wears a dirty,
oil-soaked monkey suit
that cuts him under the arms,
and several quick and saucy
and greasy sons assist him
(it’s a family filling station),
all quite thoroughly dirty.
Poem 2
A home transformed by the lightning
the balanced alcoves smother
this insatiable earth of a planet, Earth.
They attacked it with mechanical horns
because they love you, love, in fire and wind.
You say, what is the time waiting for in its spring?
I tell you it is waiting for your branch that flows,
because you are a sweet-smelling diamond architecture
that does not know why it grows.
So which one was written by a flesh and blood author versus one programmed to emulate a real person? The first was an excerpt from Elizabeth Bishop’s “The Filling Station”. The second one was Scholl’s published poem that had been provided by his poetry generator. What’s interesting is how you could totally buy that it was written by college student. It’s not formal, with a slight cheesy sentimentality, and even a whiff of naïveté.
Oscar Schwartz and Benjamin Laird are two PhD students at Monash University who’ve come up with a curious research subject: Can a computer write poetry? In 2013 they created the website Bot or Not, where people can go online and test whether they think a poem was written by a person or a machine. The results have been surprising, since at times the algorithms seem more believable than those of actual poets. It’s become a sort of mimicking of Alan Turing’s Test for Artificial Intelligence. Since the mathematician claimed that if, when conversing with a computer, at least 30% of study participants believed they were talking to another person, then that machine was considered as able to pass for human.
Then things got interesting. They noticed how certain poems written by actual human beings, famous poets even, were not passing the Turing Test. This led to a question he posed in his 2015 Ted Talk: “Do we take that William Blake is somehow more of a human than Gertrude Stein? Or that Gertrude Stein is more of a computer than William Blake?”
For that answer, we need to back to Alan Turing himself. During a BBC broadcast from May 15, 1951, titled “Can Digital Computers Think?” this is a part of what he had to say about artificial intelligence: “It seems that the wisest ground on which to criticize the description of digital computers as ‘mechanical brains’ or ‘electronic brains’ is that, although they might be programmed to behave like brains, we do not at present know how this should be done.” That’s also a conclusion reached by Schwartz and Laird, a computer can reproduce whatever you tell it to do. You give it a few sonnets, and you’ll have your iShakespeare soon enough. The difference is that humans are not constants.
People fluctuate. They get inspired by random things they start adding into what they’re used to doing. They change their minds and, sometimes, change them back to their initial set-up. Human spontaneity cannot be predicted or copied, not because there are no brilliant minds trying to do so, but because we ourselves don’t know how or why it happens. We can’t predict triggers. We might foresee some obvious symbols, but in truth we won’t know until faced with one.
I find two optimistic sides to this debacle. The first being that these generators can help freshman college students, even high school ones, play with format and words. They can put in X number of poems by one of the famous poets, and they might be able to locate patterns based on their results. They’ll see how imagery can be used in a different way while staying close to the author’s voice.
On another level, the idea of evolution and change is positive in the way that people can shift their perspectives based on what they see, hear, or do. They can come back from mistakes, specially if they learn from them. It takes a machine to show us just how much we’re not like them. We can be fed negative information and still be able to rise above it all.
Is it curious that it takes a computer scientist and mathematician like Alan Turing to present us some of the most philosophical ideas out there on human nature? Or is the ultimate use of technology to demonstrate what sets our minds apart from other organisms?
You can try your hand at telling the human from the digital at the Bot or Not website or listen to the rest of Oscar Schwartz’s Ted Talk.
Have you heard of the poem that will get you arrested if you post it on your social media? What work of verse is painfully and constantly misread?