Tradition and technology combine in a book of fully AI-generated Irish poems and imagery
The book has been published by Passage Tomb and is the work of Irish designers Oscar Torrans and Kristian Glenn, featuring typography by Liam Morrow.
Folklore is hard to pin down. Combining stories, culture, and tradition, it’s shaped over centuries by groups of people so that origins or starting points are lost to the sands of time. So what happens when a concept rooted in ancient practices is merged with one that’s very much a product of the 21st Century, like artificial intelligence? A project by Irish designers Oscar Torrans and Kristian Glenn explores this idea in a poetry book of fully AI-generated Irish poetry and images titled Machine Learning Irish Poetry.
“The telling and retelling of stories are central to Ireland’s oral and literary heritage, as well as to its sense of self,” Oscar, who’s from Bray near Dublin, but who’s been based in London for four years, tells us. Around three months ago he launched Passage Tomb, a platform to explore consciousness in pre-Christian Ireland and its contemporary echoes through myriad forms, be it books, clothing, performances of any other kind of designed object or experience. “There’s a beauty in the mystery of these traditions and structures, there are no definitions or hard facts, it’s open to interpretation and people have different theories which is what I have enjoyed most so far about the project,” he continues. When Oscar met Kristian, who’s from just outside of Belfast and has been in London for three years, the pair bonded over their shared interest in “the idea of Irishness, growing up in the North and South, as well as our shared interests in the weirder edges of Irish history and culture,” Kristian recalls. “[Machine Learning Irish Poetry] felt like a natural continuation of those conversations, and it connected with what Oscar was beginning to explore with Passage Tomb.”
The book features a series of poems and images, produced when Kristian trained a GAN on hundreds of Irish poems, folk songs, plays and myths, from The Táin to Seamus Heaney. “I wasn’t aiming to create an AI system to generate poetry that was grammatically perfect or convincingly human,” he explains. “What felt more interesting were the weird places the technology could take me; how it could reveal surprising paths and unconventional perspectives to tell old stories in new ways.” As there’s always a degree of collaboration with a GAN, Kristian tried at times to intentionally break that process, drawn to its stranger outputs. In turn, “the tone varies from being darkly humorous to surreal and psychedelic, sometimes even within one poem.” That being said, “the writing style and AI illustrations are surprisingly fluid, and that shape-shifting and liminal nature is a common element of Irish storytelling and folklore.”
This, Kristian explains, is because the generated poems very much reflect their sources. “Cú Chulainn (the Irish mythological demigod) features at one point. Another takes place at the Hill of Tara – an ancient ceremonial and burial site, traditionally referred to as the seat of the High Kings of Ireland. It has a ~6000-year-old passage tomb, which is aligned with the sunrise at Imbolc and Samhain (the Gaelic festivals marking the start of spring and winter),” he outlines. These references add weight and context to the poems, grounding them in real-life stories, “even though the technology has no recognition of its cultural significance”. In this sense, Kristin sees the GAN as a kind of digital seanchaí (a traditional Gaelic storyteller, a seanchaidh in Scottish Gaelic). “Not to devalue that tradition, but the machine learning model gathers, stores and tells stories much like its human counterpart. Though I was interested in telling new stories from a non-human perspective, they still reference the conventions and content of traditional storytelling.” Because of this, he hopes the project acts as a remedy to fears of AIs taking over and instead prompts others to look deeper into their own culture and folklore.
The design of the publication is refined, allowing the poems and illustrations to come to the fore while simultaneously referencing the design heritage of Irish poetry books. “Taking this very digital piece and presenting it in a traditional, physical way felt like the right way to close the loop of tradition fed into technology fed back into tradition,” Oscar explains. Importantly, the typeface used throughout is a creation of Liam Morrow, another Irish graphic designer based in London. “It is very difficult to find a well-made typeface in the Irish character, a million bastardised faces exist mostly made for American-Irish pubs with little consideration of history and design – basically a typographic representation of a Leprechaun. Like much of Irish history, Irish type design is wrapped up in British colonisation,” he explains. For anyone interested in delving into this topic in more depth, Oscar recommends Irish Type Design by Dermot Mcguinne.
It‘s choices like these that make Machine Learning Irish Poetry so fascinating to dig into. While on the surface the notion of creating new poetry and imagery from traditional Irish literature seems straightforward, on closer inspection, a multifarous project reveals itself. One that offers considerable opportunities to learn about the rich history of Irish folklore and design and how those elements can be repurposed for a contemporary creative scene.
Machine Learning Irish Poetry is available to purchase via Passage Tomb, its publisher.
GalleryOscar Torrans and Kristian Glenn: Machine Learning Irish Poetry (Copyright © Oscar Torrans and Kristian Glenn, 2022)
Hero Header
Oscar Torrans and Kristian Glenn: Machine Learning Irish Poetry (Copyright © Oscar Torrans and Kristian Glenn, 2022)
Share Article
About the Author
—
Ruby joined the It’s Nice That team as an editorial assistant in September 2017 after graduating from the Graphic Communication Design course at Central Saint Martins. In April 2018, she became a staff writer and in August 2019, she was made associate editor.