36   Simon Colton’s Poetic Fool

If you unwrap a computer poem, don’t look for humanity in it because it was written by software.

—Simon Colton163

Simon Colton is ubiquitous in the computational creativity field. Besides The Painting Fool, he has also developed, with colleagues who include Jacob Goodwin and Tony Veale, an algorithm that writes “unpredictable yet meaningful poetic artefacts.”164 This offers a “stronger definition of computational creative poetry,” as opposed to texts that emerge out of stochastic—that is, random—algorithms.165

Some poets intentionally write poetry that cannot be understood without textual analysis. Nevertheless, random words on a piece of paper are just that, Colton contends. An example is Robopoem, an algorithm that creates nonsense poetry from random words in a cut-and-paste fashion, much like the Dadaists did. With the Robopoem phone app, people can generate endless poems while waiting at a bus stop, then try to find meaning in them. This, as Colton writes, only highlights the fact “that people have an amazing capacity to find meaning in texts generated with no communicative purpose.”166

Colton’s poetry generator is loaded with a vast amount of material. He avoids existing poems to stay clear of accusations of plagiarism and pastiche. Instead, he throws in material from the Guardian newspaper, together with similes checked with the Google Ngram Viewer, pronunciation dictionaries, and a sentiment dictionary to see how many times they’ve been used. As with The Painting Fool, the software determines the mood of the poem depending on the newspaper articles it’s read that day. It has about twelve thousand articles to choose from and adds to its collection each day. It chooses one article, from which it extracts key phrases, then uses these phrases to put together a template that determines the number of stanzas, the number of lines in each stanza, and the rhyme and meter. Then it assembles words and phrases using this structure to create a poem.

The aesthetic content is governed by the mood of the newspaper article that the software has chosen. The mood is expressed through the frequency with which certain words appear in the poem to evoke lyricism, sentiment, or flamboyance. Colton assesses all of this by making a mathematical analysis of the poem generated.

Finally, the software produces commentary on why it chose to write that particular poem in that particular mood, choosing from a number of templates written by Colton, then going back through the stages that produced the poem. The software also comes up with a title for the poem from a frequently recurring phrase in it.

In each session, the software creates over one thousand poems. It assesses the poems according to the aesthetic criteria of relevance to the newspaper article chosen, lyricism, sentiment, and flamboyance, and it keeps the highest-ranking poem.167

Here is an example. The software speaks in the first person:

It was a generally bad news day. I read an article in the Guardian entitled: “Police investigate alleged hate crime in Rochdale.” Apparently “Stringer-Prince 17 has undergone surgery following the attack on Saturday in which his skull, eye sockets and cheekbone were fractured” and “This was a completely unprovoked and relentless attack that left both victims shocked by their ordeal.” I decided to focus on mood and lyricism, with an emphasis on syllables and matching line lengths, with very occasional rhyming. I like how words like “attack” and “snake” sound together. I wrote this poem.

Relentless attack

a glacier-relentless attack

the wild unprovoked attack of a snake

the wild relentless attack of a snake

a relentless attack, like a glacier

the high level function of eye sockets

a relentless attack, like a machine

the low-level role of eye sockets

a relentless attack, like the tick of a machine

the high-level role of eye sockets

a relentless attack, like a bloodhound

You can see the relevance to the rather violent theme it chose. Who knows what a contemporary poetry critic would make of it?

Colton remarks that people’s opinions of a poem are colored by who they think the author is, particularly if they discover that the author is a computer. To prove his point, he gave a performance in which he told his audience he was going to read a poem by Maureen K. Smith, describing her experiences of childbirth. After a few minutes, he stopped and said, “Sorry, it’s not by Maureen K. Smith. It’s by Maurice K. Smith, who witnessed the birth.”168 The audience was incredulous: a man writing a poem about childbirth? A little later Colton revealed that Maurice was a pedophile and had written the poem in prison. By now, the audience’s mood was distinctly dark. The joyous words in the poem had begun to sound sinister. Then he said, “Good news. The poem was actually written by a computer program.” The audience’s dark mood changed to disbelief at the realization that a machine had been able to play with their feelings.

“The text lost a lot of its value,” Colton notes, when he told the audience it was written by a computer. “We’ve reached the stage where such poems would be award winning if they were produced by people, but if they are written by a computer they are not considered to be real poems.”169

Colton refers to “The Death of the Author,” an essay written by French philosopher Roland Barthes in 1967. Barthes argued that literature should be taken at face value, without taking any account of the identity of the author. The message was that anyone—be they street sweepers or philosophers—could write poetry. “This very liberal movement has not survived into the computer age,” Colton laments.170 “Indeed, if you have to project humanity onto authorship of poems, why not just read poems written by people?” He suggests calling computer-generated poetry “c-poems, just like we say e-books.”

As to the argument that machines can’t feel the necessary emotions to inspire poetry because they’re not out in the world, they can at least read the web and experience, however vicariously, hunger, thirst, love, and passion. Painting machines, including Colton’s The Painting Fool, acquire experiences of the world through their portraiture. Colton is currently researching ways of storing this information in the system’s memory.

At present, Colton sees his poetry system as “fairly rudimentary [though it] can function on the majority of levels to be taken seriously as a poet, albeit in a simplistic manner.”171

Colton is dismissive of Project Magenta’s end-to-end philosophy. He is critical of using neural networks in creative projects because the programmer puts in so much information about the world that it “often means less imaginative thought.” He prefers his own highly structured, rule-based approach. Colton’s poetry system, along with The Painting Fool, certainly seems to show some of the behavior we associate with creativity, such as imagination and intent. No doubt the optimal approach is somewhere in the middle. We have yet to discover what it is.

Notes