Tag: creative practice

Un/telling y/our story, part two

Southerly Guest BloggerBlogLeave a Comment on Un/telling y/our story, part two

Amelia Walker Acknowledgement: I live and write on the lands of the Kaurna people. I pay respect to Kaurna Elders, and to all Aboriginal and Torres Strait Islander peoples. Sovereignty was never ceded. This always was, always will be Aboriginal land. I’m still on the bus. That’s a lie. Or is it? As I commit these words, it is later, many laters. I’m curled on a couch, notebook on knees, its pages filling with blue scrawlings I cross out, rewrite, cross out, rewrite, letters smudging as my frantic hand drags across not-yet-dry ink. Now I’m at a desk, squinting as…

… read more

Archival bots: My mother, my model for language

Southerly Guest BloggerBlogLeave a Comment on Archival bots: My mother, my model for language

Jenny Hedley   During two stages in my life—at age seven and again in my twenties—I acted as surrogate speaker when head-and-neck cancer muted my mother’s voice. In the years between, we collaborated on so many creative projects that our voices often became enmeshed. I have perpetuated this mother–daughter entanglement eighteen years after her death by training a small language model (SLM) on a dataset of personal writing which includes her voice. Specifically, my JenAI language model’s ‘memoir’-tagged voice trained upon my unpublished manuscript ‘Hermit Crab Diary’ which weaves together my mother’s and my 2005–2007 diary entries with my contemporaneous…

… read more

Generative AI model training: Why do you sound like me? Because I am you

Southerly Guest BloggerBlogLeave a Comment on Generative AI model training: Why do you sound like me? Because I am you

Jenny Hedley   My first fine-tuned small language model (SLM) was a boring failure who sounded didactic, encyclopaedic. JenAI v1 had trained on approximately 280,000 words of my creative writing and yet still had the (un)distinctive (non)voice of a bot trained on textbook-quality data.[1] I had tagged each of my training datasets with <author>, <genre>, and <subgenre> so that the model could learn that Jenny Hedley’s nonfiction writing—in memoir, academic, or poetic form—should read a particular way. My Python training scripts needed adjusting to make my self-miming language model more faithful to my voice, tone, and the themes of my…

… read more

To each author, a mimetic AI model

Southerly Guest BloggerBlogLeave a Comment on To each author, a mimetic AI model

Jenny Hedley  Imagine a future where writers control their own small language models (SLMs) trained on select high-quality data including their creative works. Unlike resource-guzzling, copyright-be-damned large language models (LLMs), compact models can be run locally on minimal hardware[1], keeping privacy intact while conserving resources. That future is already here for the tech-curious who are willing to accept coding assistance from LLMs. While suggesting such purposeful use of LLMs, I feel reproached by my dear colleague Beau Windon who wrote ‘Is it Time to Bully Generative AI Users?’. Am I the LLM user/‘slopsucker’/‘botlicker’ who, according to Windon, treats ‘the water-guzzling,…

… read more