Small data and model crafting
The ethics of Big Data and large foundational generative models are often questionable. Besides the obvious vampirization of online data and artworks, they promote a generic, yet biased, aesthetic.
In response to large prompt-based models, the Metacreation Lab created Autolume, a no-code system enabling artists to train their own AI models with their selected works. Autolume allows non-coders to craft, mix, and finetune their own generator, and manipulate in real-time their many parameters to produce both still and animated outputs.
The result is a selection of four artworks from collaborations practicing model crafting as a way of moving forward without falling prey to the generic AI aesthetic of pre-trained, large commercial models.
-
Autolume Mzton
Jonas Kraasch (DE), Philippe Pasquier (CA/FR)
Autolume Mzton explores the notion of birth using Autolume audio-reactive features. Driven by the piece Mzton, from the analog modular rhizome of the French band Robonom, the neural aesthetic of generative visuals unexpectedly evokes early experimental analog cinema.
-
Dreamscape
Erica Lapadat-Janzen (CA), Philippe Pasquier (CA/FR)
In response to AI-generated art using Big Data, the Metacreation Lab developed Autolume, a no-code system for artists to use their own works in training AI models. This tool allows non-experts to create both still and animated outputs.
-
Ensemble
Arshia Sobhan (IR), Philippe Pasquier (CA/FR)
This collection melds the traditional art of “siyah-mashq” in Persian calligraphy with AI model crafting. Each piece features generatively evolving and fluid calligraphic forms, accompanied by a background sonic texture.
-
Longing + Forgetting
Matt Gingold (AU), Thecla Schiphort (CA), Philippe Pasquier (CA)
Longing + Forgetting explores pathfinding algorithms as a metaphor for our personal and collective searches for solutions. Combining physical and algorithmic choreography, the work consists of a collection of artificial agents finding their way through the projection surface.
Simon Fraser University – Vancouver, School of Interactive Arts and Technology, Metacreation Lab for Creative AI (CA)
The Metacreation Lab for Creative AI runs a research-creation program at Simon Fraser University in Vancouver, Canada. Focused on the design, evaluation and application of generative systems since 2008, the Metacreation Lab is at the forefront of creative AI developments. Its scientific and artistic work has been shown on six continents, and the Lab’s projects are reaching global creative communities through open-source releases and collaborations with the software industry.
Credits
Longing + Forgetting
Concept: Matt Gingold, Philippe Pasquier and Thecla Schiphorst
Video Design and Code: Matt Gingold
Sound Design: Philippe Pasquier
Choreography: Thecla Schiphorst and Matt Gingold
Set Design: Greg Snider
Associate Producer: Kristin Carlson
Lighting Design: Ben Rogalsky
Editing and Animation: Josh Burns and Matt Gingold
Code Development: Matt Gingold
Performers: Shannon Cuykendall, Matt Duncan, Sarah Fdili Alaouim, Meghan Goodman, Marcus Marshall, Joshua Ongcol, Priya Rajaratnam, Bladimir Santos Laffita, Nathalie Sanz, Cara Siu, Yawen Wang, and Martin Wong
Dreamscape
Concept and Visuals: Philippe Pasquier and Erica Lapadat-Janzen
Generative AI Assistant: Arshia Sobhan
Autolume Mzton
Video Generation programming: Jonas Kraasch
Sound Design: Philippe Pasquier
Ensemble
Generative Calligraphy: Arshia Sobhan
Sound Design: Philippe Pasquier
We acknowledge support from CCA, SSHRC, NSERC.
