IMMATERIAL DISRUPTION & REPAIR
[the below video is embedded - please allow a few moments to load.]An experimental audio track was organically created using a KORG Minilogue XD synthesiser and a Roland TR-8S drum machine, recorded in a single track through Ableton Live, and exported with higher levels of compression and lower quality, intentionally removing fragments of the creation. This audio was then used to generate visuals through mathematical structures that render 3D shapes and pointclouds: particle-based residues that respond dynamically to audio.
In implementing the audio into TouchDesigner, I allowed the machine to attempt to fix its own errors, introduce new patterns to its interpretation of the mistakes, and create a visual response to where the audio had been compressed or altered in its export process. A bit like passing the book; one area of the machine made the mistakes, but another area was expected to recover from it.
In researching the concepts of machine vs. human learning, one of the key questions I pose is ‘what do humans possess in our frameworks of learning that machines lack, that differentiates how we process, adapt from, and teach information?’. One part of this is the notion of consequence, that the machine/the cyborg would lack the emotional regulation that becomes so integral to humans when we learn through experience - learning theory in humans suggests that learning occurs when an outcome deviates from expectation, and is a form of prediction error. The threshold between prediction and learning lies within the plasticity of the human brain, and the mathematical and/or neurological responses to external stimuli, and whether there is a mismatch between preconceived ideas of what is about to happen, and what actually happens. Essentially, when we go through something that challenges or questions our previous knowledge, this actively informs our future responses to similar stimuli.
Perhaps the distinction lies not in whether a machine can recognise an error, but whether it can understand that an error has occurred. The machine can identify deviation, calculate difference, and produce a new response, but the consequence of that deviation remains external to it - in this process, the machine is given an imperfect piece of information and asked to adapt to it. It learns, in a way, from the prediction error, but without experiencing any of the conditions that made the error meaningful in the first place. What results is another interpretation entirely, as opposed to a correction; a residue of the original mistake, that has been transformed through another system of learning. The resulting forms in ‘Immaterial Disruption & Repair’ have become traces of an exchange between human intention and machine interpretation, which queries who can take accountability for errors and for trying to fix them.
Maybe learning doesn’t just require the capacity to change in response to information. Maybe we also need the capacity to understand why that change matters. Maybe that is also something missing from the machine.