The internet is a wondrous thing. When I’m writing an article and trying to steer clear of labelling something as ‘interesting’ – a term that teachers incessantly asked us to avoid – I may sift through the abundant world of the online thesaurus. I tend to struggle to find something appropriate for the sentiment I am aiming to express, because of all the choices available.
But on the other side of these digital choices and endless resources lies another issue: what are you really choosing from? In November, Google announced a ‘human-AI collaboration for writing poetry’ called Verse by Verse, which allows you to choose poet inspirations and make your own poem with a mixture of human input and artificial intelligence.
So I had a go at making my own Google poem. I was presented with twenty-two poets, three to choose as inspirations. Thirteen of these poets were white men (shock horror), eight of these were women and four of them were Black poets (one man and three women). I was asked to write my own title and first line, and given suggestions for further lines, style, and content. Putting aside the fact that it’s impossible to summarise a wealth of varied poets in twenty-two bite-size case studies, the race and gender breakdown is also an uncanny mirror to real-life hierarchies. A Creative Industries Federation report shows how much work the creative market needs to do to improve gendered and racialized discriminations.
However, another problem is precisely the inclusion of a tokenistic diversity where race and gender are made visible through AI technology. In this way, recognition becomes a tool for categorising groups according to identity. Researchers Clementine Collett and Sara Dillon put forward strategies for mediating AI discrimination with the University of Cambridge Leverhulme Centre for the Future of Intelligence. They articulate that the notion of classifying faces using AI technology reinforces socially constructed categories through repetitive processes, stating that “The concept of ‘classification’ and ‘recognition’ in general ought to be questioned as a legitimate and acceptable exercise.” In fact, a Harvard study emphasized that AI techniques employed by law enforcement for identifying potential ‘suspects’ echoed the techniques used historically to segregate and separate Black people from white people in the United States. Enforcing practices of recognition risks grouping identity with criminality.
I am a huge fan of digital landscapes as spaces for creativity and really dislike the rhetoric that implies technology is ‘killing creativity’. In fact, digital art can provide us with different modes of creativity, like graphic design or digital publication (or like me writing this article on Google Docs right now!). However, the specific insertion of AI technology into the production of poetry means that rather than the individual freely cutting up and piecing together digital space, the ‘answers’ are easily generated and ready. They don’t even really make any sense: I typed in ‘four corners’ and instantly was given some numerically themed lines based on my chosen poets – not exactly a liberal interpretation. Whilst the poem was fun to make and provided suggestions that sparked my imagination, the AI generator replaced a space for critical discussion and research with an existing world full of bias that you must mould yourself into.
Google is one of the giants that make up the term ‘Big Tech’, a phrase often repeated to the point of redundancy. Google owns and controls much of our digital creative and educational space, and it is really talented at distracting you from the bigger picture that it lives in. As journalist and theorist Rebecca Solnit writes in my well worn favourite essay of hers, ‘The Garden of Merging Paths’, Silicon Valley is a symbolic maze: “And the maze’s image is echoed in the circuit boards and silicon chips […] of centerless towns that melt into each other”, the landscape is “wholly interior” and works at “eliminating the world”. Verse by Verse stands as one path to this maze, where limited choice and outlined fun obstructs the inequalities of privatized and mass-produced education.
On the one hand, I think that there is much to be gained from understanding our identities as made up of digital and physical qualities. As esteemed theorist Donna Haraway argues in her ‘Cyborg Manifesto’, we should be wary of separating the notion of human from machine or animal for fear of implying anthropocentric authority or instating a kind of ‘human knows best’ rule. In fact, the mixture of human and AI technology in Verse by Verse could be a perfect instance of Haraway’s characterisation of human and machine intertwining.
Yet the Google generator is not an instance of merging, but a direct result of humans inserting preconceived notions of poetry and creativity into technology. Sure, Google can create some jovial lines and teach me about different poetic structures and styles, which can make up for the lack of creative access in lockdown. Yet, doesn’t it truly deny the essence of a poem, which is often based on the confusion of experience or imagination, not on the flow of superficially matching lines? It seems to encourage enforcing rules instead of breaking new ground.
In the end, it comes down to how we approach teaching and learning. As Feminist theorist Carolyn M Shewsbury states in her classic essay ‘What is feminist pedagogy?’, one of the foundations of learning horizontally and avoiding power imbalances is creating community, merging the listener and the talker. When I construct my Google poem, I am met with no response, no workshopping, no challenges. Creating genuine rather than tokenistic equality in creative or educational spaces is about creating communities that support and challenge each other. Otherwise, instead of collective creation, we are just one individual led around a garden maze.