Impossible Evolutions

Full description

Background
‘Impossible Evolutions’ is a creative research project using generative machine learning (ML) models in the design of woven tapestries. ML is used to unfold the stories of vulnerable organisms, while textiles provide novel perspectives for locating emerging relations between human and machine intelligences. The project offers ways of positioning artistic practice with ML models through ‘textillic thinking’. Textile practices foreground creativity, collaboration and care, and textile language provides a useful material vernacular for discussing these seemingly intangible technologies. The lens of textillic thinking (drawing together Tim Ingold, Elaine Igoe and Rosi Braidotti) positions ML technologies as unfolding, ongoing processes, and importantly emphasizes the agency, responsibility and level of care of the humans involved in their design.

Contribution
The work comprises 3 computer woven tapestries each 200cm X 150cm. Generative Adversarial Networks (GANs) and diffusion models are trained on datasets and drawings of endangered Australian butterflies and wildflowers to imagine new iterations of these creatures. These ‘impossible’ machine-imagined evolutions gesture to the futures we face as biodiversity decreases. I developed processes for collecting, sorting and generating these images using GANs, and working them into digital compositions able to transpose into physical textiles.

Significance
Australia Council Grants:
2022: (in part) $10,900
2021: (in part) $14,000
Competitive EOI:
2022: exhibition space and support through City of Melbourne at Assembly Point.
2023: Wild Hope exhibition for Now or Never festival at RMIT Design Hub Gallery.
International Symposium of Electronic Art 2023, Paris France