In 2024 OpenAI approached me to create poetry using OpenAI tools, for the work to be included in an exhibition that would launch OpenAI's arrival in Dublin. As someone who is very interested in the arts and technology space but with very little experience of using AI, I had the perfect opportunity to collaborate with my best friend Melissa Lara Clissold, who loves finding ways that AI can be used to tell stories.
Confluence of Echoes: A Triadic Poetry Journey
|
ChatGPT: Reflect on a moment that stands out in your memory where the natural world around you dramatically changed your mood or perception. This could be a sudden weather shift, an encounter with wildlife, or witnessing a seasonal change. How did the sights, sounds, and scents of this moment make you feel, and what lasting impression did it leave on you?
Melissa: I remember when there was a solar eclipse. I was around 10, 11 or 12? I was in my summer place in the South West of Türkiye, Datça. Newspapers had handed out little glasses that we could use to watch the solar eclipse which took part during the day. Everyone was on the beach. I was with my mum and suddenly everything started getting super dark. It was absolutely magical and everyone was just smiling, looking at one another. I remember an almost silence amongst the noise and excitement. Waves lapping endlessly. |
the sand crunches underfoot
summercrisp and morningfresh the sea whips and rages wetsucked and muddy feet sink deep into the sand through my lens I see Orion, the big dipper the panflip to Polaris I wait until the sea foams up to my feet dip my fingers in, suck the salt off them a sleepy fire rages behind us the sunrise erupts across the sea - Chandrika |
|
ChatGPT: Recall a moment when you felt a profound connection with a place in nature that left a deep imprint on your heart. Perhaps it was a serene lakeside, a vibrant city park in bloom, or a quiet forest path. How did the sights, sounds, and smells of this place merge with your emotions at the time, and how has the memory of this connection influenced your perception of nature or the place itself?
Melissa: I live in a crowded city. To escape the hustle and bustle I recently went to a sort of newly opened city forest which has a beautiful lake with ducks. There is a variety of birds that you cannot necessarily see but can sometimes hear or read about on the passing signs. Just walking amongst trees, breathing in the fresh cold air on a sunny day and then lying down on the grass always invigorates me. It just reminded me that I need to spend more time in nature because I always feel like I return to my true self when I am surrounded by the stillness of the forest. The smell of soil, various flowers, the sounds of families walking, children watching and feeding the ducks just made me feel connected to everything around me. |
wintercrisp and morningfresh
the dirt crunches underfoot plum blossoms bloom in the cold dandelion seeds float across the square feathercool and springbright the river whips and rages wetsucked and muddy feet sink deep into the water the sunrise erupts across the lake crackling into the bright blue - Chandrika |
Artist Statement
This piece is a collaborative poetic experiment about memories of nature in local environments and how they can connect across countries. ChatGPT acted as a facilitator connecting Melissa’s memories of Türkiye with Chandrika’s poetry about Ireland. We had strict parameters regarding the role of AI: it must only facilitate the conversation, and not generate new artistic output. Chandrika uploaded a database of poetry lines which included some variables that ChatGPT could choose to suit the context (e.g. it could choose ‘sand’ over ‘dirt’).
We then asked it to create a memory prompt about nature with a focus on sensory and emotional impact. Melissa answered the question and ChatGPT assigned pre-existing lines from Chandrika’s poetry. This piece includes the two best versions of this experiment.
The process took some trial and error with very mixed outcomes (e.g. ChatGPT decided it would include ‘cackling brownies’), and the end results required human intervention for editing. In conclusion, this was an interesting experiment that could be developed further, however Chandrika felt that AI hindered rather than helped her in creating writing of a high standard.
We did ask ChatGPT to name this project though, and we have to admit it is the perfectly pretentious title it deserves.
We then asked it to create a memory prompt about nature with a focus on sensory and emotional impact. Melissa answered the question and ChatGPT assigned pre-existing lines from Chandrika’s poetry. This piece includes the two best versions of this experiment.
The process took some trial and error with very mixed outcomes (e.g. ChatGPT decided it would include ‘cackling brownies’), and the end results required human intervention for editing. In conclusion, this was an interesting experiment that could be developed further, however Chandrika felt that AI hindered rather than helped her in creating writing of a high standard.
We did ask ChatGPT to name this project though, and we have to admit it is the perfectly pretentious title it deserves.
Process
Melissa and I came to this project with two very different perspectives. As someone working in the arts, I have seen how AI has negatively affected the livelihoods of artists from multiple sectors. However, I have also seen artists use AI effectively as a tool to create incredible new pieces. Melissa has worked with various AI tools to create games, interactive pieces, and storytelling experiments, and loves using it to explore how humans and AI can collaborate. We met in the middle, with the focus on AI in this case being used only to facilitate the conversation, and help us build the project, but never to replace the role of the artist themselves with generating either the poetry or any images.
I created a database of poetry lines based on my memories of nature in Ireland in the past years. The poems that were being composed by AI weren't really working together, and I realised that this was because they were in different tenses, and had varying focuses. Basically, I was writing just normal poems instead of thinking about how they need to work recontextualised line by line. Melissa made a massive contribution here, suggesting we use variables the same way it's used in game design to help ChatGPT contextualise the lines better based on the memories they were responding to. I also edited the lines to make sure they kept their sensory emotion while being used across seasons and scenes. The results were excellent: the lines that responded to Melissa's memories worked.
ChatGPT very successfully helped us with the memory prompts as well. It understood what we needed fairly quickly, and after several iterations of questions, it was able to create prompts that brought out exactly what we needed from Melissa's memories. The wrong prompts would have created answers that didn't flow as well, but Melissa's answers worked beautifully as part of the piece.
ChatGPT only ever got us about 90% of the way. In terms of editing, I had to made a good few edits to make the poems work as well as they could, sometimes deleting a line, editing punctuation, deciding how the form would be formatted in their stanzas etc. And sometimes editing a word that just didn't work. However, we tried to keep this to an absolute minimum. Also, in the actual process, ChatGPT often started getting notions of itself as a poet, forgetting that we had instructed it not to make any major changes to the text. Yet it created entire lines of, frankly terrible poetry, which was comedically grandiose in nature. It also randomly decided to change the words that weren't our chosen variables - but it all worked out in the end!
Below is the final list of poetry lines that were used to create the piece:
the [frost/sand/dirt] crunches underfoot
[winter/summer]crisp and morningfresh
the heat melts the sunscreen into my eyes
my eyes sting with the sweat and heat
plum blossoms bloom in the cold
plump fruit crushed on the pavement
[red/green/yellow/dead] leaves on the wet pavement stuck underfoot
trailing mess into a freshly cleaned home
[cherry blossoms/dandelion seeds] float across the square
feathercool and springbright
freshly baked brownies steam from a kitchen window
we lick ice cream from sticky wrists
cackling into the bright blue
unfurling into the morning
the [sea/ocean/river/water] whips and rages
wetsucked and muddy
[feet/shoes] sink deep into the [bog/sand/water]
the freshgreen leaf of the pothos plant
from the window I see the big dipper, the panflip to Polaris
silvery purple leaves were once only a cutting
tradescantia loops and tumbles from the shelf
wait until the [sea/ocean/water] foams up to my feet
dip my fingers in, suck the [salt/sugar/juice] off them
an aloe plant, cut from one belonging to a mother, a mother's mother
the basil wilts and crumbles, spinach-bruised
small spindly flowers so elusive and so loud
the tree down the street erupts into a violent blue
a sleepy fire raging behind us
the sunrise erupts across the [lake/water/sea/ocean/mountain/hill/field/city]
I created a database of poetry lines based on my memories of nature in Ireland in the past years. The poems that were being composed by AI weren't really working together, and I realised that this was because they were in different tenses, and had varying focuses. Basically, I was writing just normal poems instead of thinking about how they need to work recontextualised line by line. Melissa made a massive contribution here, suggesting we use variables the same way it's used in game design to help ChatGPT contextualise the lines better based on the memories they were responding to. I also edited the lines to make sure they kept their sensory emotion while being used across seasons and scenes. The results were excellent: the lines that responded to Melissa's memories worked.
ChatGPT very successfully helped us with the memory prompts as well. It understood what we needed fairly quickly, and after several iterations of questions, it was able to create prompts that brought out exactly what we needed from Melissa's memories. The wrong prompts would have created answers that didn't flow as well, but Melissa's answers worked beautifully as part of the piece.
ChatGPT only ever got us about 90% of the way. In terms of editing, I had to made a good few edits to make the poems work as well as they could, sometimes deleting a line, editing punctuation, deciding how the form would be formatted in their stanzas etc. And sometimes editing a word that just didn't work. However, we tried to keep this to an absolute minimum. Also, in the actual process, ChatGPT often started getting notions of itself as a poet, forgetting that we had instructed it not to make any major changes to the text. Yet it created entire lines of, frankly terrible poetry, which was comedically grandiose in nature. It also randomly decided to change the words that weren't our chosen variables - but it all worked out in the end!
Below is the final list of poetry lines that were used to create the piece:
the [frost/sand/dirt] crunches underfoot
[winter/summer]crisp and morningfresh
the heat melts the sunscreen into my eyes
my eyes sting with the sweat and heat
plum blossoms bloom in the cold
plump fruit crushed on the pavement
[red/green/yellow/dead] leaves on the wet pavement stuck underfoot
trailing mess into a freshly cleaned home
[cherry blossoms/dandelion seeds] float across the square
feathercool and springbright
freshly baked brownies steam from a kitchen window
we lick ice cream from sticky wrists
cackling into the bright blue
unfurling into the morning
the [sea/ocean/river/water] whips and rages
wetsucked and muddy
[feet/shoes] sink deep into the [bog/sand/water]
the freshgreen leaf of the pothos plant
from the window I see the big dipper, the panflip to Polaris
silvery purple leaves were once only a cutting
tradescantia loops and tumbles from the shelf
wait until the [sea/ocean/water] foams up to my feet
dip my fingers in, suck the [salt/sugar/juice] off them
an aloe plant, cut from one belonging to a mother, a mother's mother
the basil wilts and crumbles, spinach-bruised
small spindly flowers so elusive and so loud
the tree down the street erupts into a violent blue
a sleepy fire raging behind us
the sunrise erupts across the [lake/water/sea/ocean/mountain/hill/field/city]
Bios
Chandrika Narayanan-Mohan is writer, performer, and cultural consultant from India living in Dublin. Her work has been published by Dedalus Press, UCD Press, Lifeboat Press, Banshee, Stinging Fly, Poetry Ireland, and others. She was Writer in Residence for the Institute of Physics for 2023.
|
Melissa Lara Clissold is an Istanbul-based communicator and artist. With a foundation in English Literature, Philosophy, and Social Anthropology, she integrates impact storytelling in her multidisciplinary projects ranging from game design to social justice initiatives to music production. As Communications Lead at ATÖLYE, she champions justice, equity, diversity, and climate action.
Check out her website here, and her AI project here. |
Chandrika and Melissa met in High School in Ankara, Türkiye and have been friends ever since. Both coming from multicultural backgrounds with an interest in art, technology, and nature, this is their first collaborative commissioned piece.
|
