The gifts of our mistakes

Kate Geck

A show of textiles exploring co-creation within emerging technological landscapes, where materialities of glitch, error and digital detritus sit alongside machine imagined imagery.

One room is filled with colourful hangings, echoing chaotic browsing histories and carrying the saturated psychological weight of unread or unreplied messages. The space is imagined as a flat temporality where cartoon nostalgia and future technological ruins sit alongside each other in an overwhelming permanent present. The internet lets everything be known now: with an abundance of info, it's our attention that becomes the precious commodity. With colour and texture this space materialises the sensation of being always on, of constantly fighting to organise your attention and decide what to know.

Despite what design tries to tell us, the digital is material, full of friction and actively extractive. Most of the web pages and app pages we are served are clean and clear with simple calls to action, their minimalism a tactic to control attention. In these textiles, technical mistakes and in-between moments are captured, stacked and immortalised creating unintended digital rhythms and unexpected surface frictions.

Another is filled with machine-imagined butterflies after a text-to image machine learning algorithm was asked to dream itself as a butterfly. These works speculate the nature of imagination and experience. Somewhere around 2300 years ago, the Daoist philosopher Zhuangzi wrote about awakening from a dream where he was a butterfly. When he woke, he was unsure if he was a human who had dreamt he was a butterfly, or a butterfly dreaming it was human.

His story brings into question the nature of sensory experience and the effects of imagination. It unfolds 'models of mind' - imagining how other things might feel, as well as 'umwelts' - the edges or gamuts of sensory experience that an organism can have. Will a machine intelligence be able to imagine how other organisms feel? And what kinds of worlds will machine intelligences sense? Will these things give rise to machine experiences - different to our own, but still happening: feeling real and producing effects in the world.

A lot of the metaphors we use to understand machine intelligence draw from extraction and control, positioning the AI itself as a resource to be extracted and controlled. We talk about 'oceans of data' to be 'mined, and 'AI arms races' with winners and losers. What happens if we think about machine intelligence through more relational metaphors - ones centred on more even exchange, and eventually perhaps even on mutualism?

These embroidered textiles were made in collaboration with generative adversarial networks, a machine learning model that learns from datasets in order to generate its own images. The model was trained on 100 hand drawings of Australian fungi that I made. It then imagined its own versions of these, which I translated into embroidery designs to stitch onto cotton. This cotton is printed with abstract machine imagined fungal textures. I have come to understand this co-creative process as mycorrhizal - human and machine intelligence work together to produce the creative work.

A mycorrhizal structure is relational, involving collaborative sharing and the moving of resources between plant and fungus. The fungal hyphae and plant roots work closely to exchange nutrients, sometimes even interpenetrating each other and becoming impossible to untangle.

If we think about the exchanges between human and machine intelligences in a mycorrhizal way, we come to see that they are entirely entangled: machine decisions are based on human datasets, these decisions affect human lives and in turn change the datasets we produce. A relational metaphor like the mycorrhiza foregrounds interconnectedness and exchange, holding space for speculative futures in which machine intelligences might ethically require more mutualistic relationships with humans to emerge.

The last collection of work asks different ML models to imagine the emergence of the universe from stardust, and to write and record a guided meditation on the nature of interconnectedness. This meditation plays thru your phone via your web browser. Through readily accessible free technologies, machine intelligences can give a pretty convincing semblance of understanding the poetic vastness of the universe. While it is novel and a little off-kilter at present, I am curious what happens when machine intelligence actually begins to locate this understanding with some reference to 'self'.