Jonathan McCabe - Nervous States

Canberra artist Jonathan McCabe is currently showing some digital prints at the Front gallery in Lyneham - the show is called Nervous States, ostensibly referring to the neural net behind the generative process... but it seems to have much wider implications just at the moment, too. I wrote about McCabe's Butterfly Origami Method on generator.x a while ago, and was impressed by the elegance of the generative mechanism and the visual richness of the results. Nervous States is just as elegant, and visually psychedelic, but uses a completely different generative approach.


Like the Butterfly Origami images, there's a sense of materiality here... which is paradoxical, considering the abstraction of the generative techniques. Each image is essentially a visualisation of the output state of a small neural network. The X and Y coordinates correspond to two variables in the connections of the network; the colour of the pixel at that point is a representation of the network's behaviour for those parameters. So the image is a map of system states; coherent colours show areas of relative stability or gradual change; edges show sharp jumps in the output; marbled swirls show complex oscillations.

Technically, this work is pushing the edges in several ways. To select images from the vast range that the system can produce, McCabe first uses an automated analysis based on variation in the image at three levels of scale: the software varies the weighting of the inter-neuron connections, and selects images (maps) with the most variation. However this automated process still generated 6000 candidate images, which McCabe then whittled down to nine for this exhibition.

Generating these images at very high resolutions is a hefty computational task. The solution for McCabe was to make use of the parallel-processing grunt available on the video card. Using the Brook language from the Stanford Graphics Lab, the images are rendered using the parallel pixel processors on an nVidia graphics card.


This work also makes me wonder about communication, meaning and generative art. As McCabe explains them, and in the context of the "nervous" metaphor, the generative system is poetic in itself; the images can be read in that context, as mysterious maps of complex dynamics - or they can function on a more "retinal" level, as sheer visual stimulus - or perhaps both. But how comprehensible is the generative system for a wide audience? Does it matter? Understanding the images as state maps, rather than physical (or even simulated physical) traces and gestures, is a considerable leap of abstraction. And at a time when open-source tools are drawing more and more artists and designers to generative techniques, McCabe's work issues a similar challenge: underneath the initial challenge of learning to code is the conceptual process of understanding, designing and visualising generative systems, and it's those systems that (I'd say) are at the core of the work.