My work on landscapes strives to balance an iconic representation of astronomy data (also seen on the cover of Joy Division Unknown Pleasures) and the meticulous Meridian series by Matt Deslauriers. For this particular piece, I worked around the data-to-ink ratio idea from data visualisation, a quick way to say that you should avoid using too much ink if it does not represent information (e.g. fillings, effects). In this case, sticking to lines to represent a landscape, I was curious how much data you can afford to drop while keeping a realistic representation.

I aimed to emulate etching drawings with code, starting from a chunk of data representing terrain elevation over a 50 x 50 km region of the Pyrénées (the broad Conflent region) and successively altering this dataset by adding noise and discarding information as a function of various attributes (elevation, slope, distance). Finally, this representation was cut into vertical bands, whose widths were empirically chosen (~ 3 km) to avoid too strong discontinuities. Between all these altering processes, a bit less than 2 percent of the initial topographical information is represented in this piece.

I reckon that the appeal of generative art is to design algorithms whose emergent properties can surprise artists and audiences alike while maintaining a long-form meaning. However, I'm curious about generative art as an evolutionary process (random generation, selection, adaptation), where artists can patiently collect interesting seeds in the hope of evolving them into something new, independently of the mindset of their creation.