In public discussions of Artificial Intelligence people have, for the most part, focused on how it will affect employment and the reliability of news. These are certainly serious matters that will relatively soon affect the livelihood of many individuals and the shape of society. They also prompt us to ask, as AI becomes more varied and sophisticated, how it might change our understanding of ourselves as human beings. To grapple with this large question it may be helpful to ask more apparently playful and peripheral questions. One such question asks whether AI could write good poetry.
A cynic might argue persuasively that the question is idle. The task would demand that the Lords of the Technological universe commit great resources to such a project, and the Musks, Bezoses and Zuckenbergs would never see a buck to be made from taking over poetry. Nevertheless, assuming that the enterprise was financed, I believe that Artificial Intelligence could produce competent poems that might win the praise of knowledgeable critics.
As an aside, the reference to critics in this context will twitch the ears of many Australian readers. It will evoke the Ern O’Malley affair. This was initiated by two Australian poets who valued form and clarity in poetry over the contemporary fashion that favoured spontaneity and looseness of structure. They carved up lines from the Parliamentary proceedings in Hansard and submitted the resultant poem to an avantgarde literary magazine under the fictitious name Ern O’Malley. It duly published it. The schadenfreude and scepticism about critical judgment that followed the unmasking of the affair were monumental and their traces endure.
That said, the resources that AI could be provided with if it turned its fleshless hands to poetry are considerable. It would have at its working command the resources of the Oxford English Dictionary with its attention to the historical usage of words, the thesauruses of synonyms and antonyms, the multifarious books on rhetoric and modern manuals on poetry with their reflections on sound, rhythm and imagery, the works of all the recognised English poets of the past and present and the judgments of them, just for a beginning.
With these resources and after painstaking editing of the instructions it would build on in using them, I do not doubt that AI could produce on any topic verse that would competently imitate in its language, rhythms, and style the work of writers of a particular period. Assuming, too, that the capacities we have so far seen in AI bear the same relationship to those that will quickly be developed as did the capacities of the Gutenberg Press to those of modern printing presses, I am confident that I, at least, would eventually be led to describe many poems it produces as competent and as the work of fine poets.
Are competent poems, however, the same as good poetry? I think not. The writing of poetry involves more than management of technique. Certainly AI has the potential capacity to meet all the how-to challenges that face a poet, amateur or professional, in their writing. It perhaps also stir a corresponding response in readers. It cannot, however, replicate the inner relationships and processes within a poet when writing a poem, which define it as poetry.
'Poetry reminds us of the earthy glory and limitation of human beings at their most social and creative.'
These qualities include the urgency to catch an experience, thought or relationship that defies simple description. This sparks the writing of poetry. Poetry includes also the relationships to an imagined audience and to other poets that enter the writing even if it is never to be published. Within it, too, is the struggle to find the right words, the right sounds, the right form that matches the impulse behind the poem to its expression in words. It includes also both the satisfaction of writing something that half-way works mingled with the disappointment, the sense of failure, that the words have failed quite to capture the elusive quality of the original impulse to write. A finished poem inevitably has the same relationship to its origins as does a pressed flower in a book to the flower in the garden. The pressed flower inevitably lacks the life of the original. It remains, however, a flower quite different from a plastic or paper image of it even though this may be faultless.
Good poetry is good because the poem embodies these processes and relationships in its writing. They mark it out as the fruit of a human activity that bears the traces of a large network of intersecting relationships within poets and with the world outside them.
Underlying this analysis lies the crucial relationship between human beings, the tools they use, and their artefacts. The poem is an artefact that human poets produce. The process of production involves self-reflection, bodily experience, valuing and agency, and as a human artefact cannot be fully understood without reference to these qualities. As originator of a poem the poet is quite different than Artificial Intelligence which is a human instrument or tool. It could produce fine imitations in different styles, but a copy remains different from an original.
The risk of Artificial Intelligence does not lie in the quality of what it produces but in the risk that we shall be seduced by it to see human beings as instruments for the production of artefacts and not as agents. That view is already common in economic theories that regard human beings as instruments for the production of profits, and so with no second thoughts see them replaceable by machines. That is one step from seeing human bodies as artefacts open to the re-creation by privileged human minds and wills of gender, health, emotional state and intelligence through the available instruments of genetic engineering, surgery, pharmaceutical intervention and artificial intelligence.
Poetry reminds us of the earthy glory and limitation of human beings at their most social and creative. It differs from verse that considers only form, technique and use. Many poems written by Poets Laureate for civic events, of course, are little more than verse. In that respect in future these poems might be entrusted to Artificial Intelligence. But not the poetry that is written by the same poets with serious intent.
Andrew Hamilton is consulting editor of Eureka Street, and writer at Jesuit Social Services.
Main image: Westminster Abbey stonemasons prepare permanent memorial to poet Philip Larkin in Poet's Corner. (Dan Kitwood / Getty Images)