Solar Powered Alba Generator
Proposal Type
Individual Talk
Location
Algorithms & Imaginaries
Start Date
July 2026
End Date
July 2026
Abstract
I present Solar Powered Alba Generator, a design for and implementation of a handheld solar-powered poetry generation device. The device consists of a low-power microcontroller (a Raspberry Pi RP2040), which draws energy exclusively from a small on-device solar panel. There are no rechargeable batteries on the device; it operates only when the panel draws enough energy to directly power the microcontroller. The device generates text using a radically small language model—a Markov chain—which is "trained" anew on the device each time it's activated. The project is meant to draw attention to some of the aspects of language that large language models erase: its transtextuality (all texts are in conversation with other texts) and materiality (a text can't be separated from its production). I also mean for the project to demonstrate that computation's role in poetry—in making language new—can be (and must be) accomplished without incurring the enormous environmental impact of conventional "generative AI."
In the talk, I discuss the motivation and theoretical context of the device, alongside a brief review of the device's software and hardware design. Additionally, I present a small annotated plain-text corpus of English language albas (and alba-likes) that are in the public domain in the United States. This corpus serves as the source data of the on-device language model.
Solar Powered Alba Generator
Algorithms & Imaginaries
I present Solar Powered Alba Generator, a design for and implementation of a handheld solar-powered poetry generation device. The device consists of a low-power microcontroller (a Raspberry Pi RP2040), which draws energy exclusively from a small on-device solar panel. There are no rechargeable batteries on the device; it operates only when the panel draws enough energy to directly power the microcontroller. The device generates text using a radically small language model—a Markov chain—which is "trained" anew on the device each time it's activated. The project is meant to draw attention to some of the aspects of language that large language models erase: its transtextuality (all texts are in conversation with other texts) and materiality (a text can't be separated from its production). I also mean for the project to demonstrate that computation's role in poetry—in making language new—can be (and must be) accomplished without incurring the enormous environmental impact of conventional "generative AI."
In the talk, I discuss the motivation and theoretical context of the device, alongside a brief review of the device's software and hardware design. Additionally, I present a small annotated plain-text corpus of English language albas (and alba-likes) that are in the public domain in the United States. This corpus serves as the source data of the on-device language model.
Bio
Allison Parrish is a computer programmer, poet, and game designer whose teaching and practice address the unusual phenomena that blossom when language and computers meet. She is an Assistant Arts Professor at NYU’s Interactive Telecommunications Program.
According to Ars Technica, Allison’s work “delight[s] everyone.” She was named “Best Maker of Poetry Bots” by the Village Voice in 2016, and her zine of computer-generated poems, “Compasses,” received an honorary mention in the 2021 Prix Ars Electronica. Allison is the co-creator of the board game Rewordable (Clarkson Potter, 2017), and her books, chapbooks and collaborations have been published by presses such as Counterpath, Instar, Aleator, and Nothing to Say. Her poetry has appeared in BOMB Magazine, Strange Horizons, Taper and Ninth Letter, among other publications. In 2024, she was the recipient of the Electronic Literature Organization’s “Maverick” award.
Allison is just out here every day doing her best. She is originally from West Bountiful, Utah and currently lives in Brooklyn.