CIRCLES, 2023

While we’ve been accustomed to mechanization replacing physical tasks, the foray of machine learning into the creative sphere has upsaled pandora of intellectual and ethical discourse. The information we input into the system only partially determines the quality of responses or outputs. The other part is the processes through which we condition the system to create its analytical structure. Considering the parcels of content we upload means providing context and categorization– which are inherently creating a narrative through pathways or progressions of links. 

In this small-scale project I questioned if the progression of time in the tangible world would have an effect on progressive AI generation. Using the same input prompt at a 50% match generation setting each time, I created one or two daily rendering(s) using the rendering(s) from the day before. 

Keeping the generative process at a surface level, the main fluctuating variable was the amount of real-world time between the initial rendering and the final output about a month later. Qualitatively, I found a large prioritization on consistency– with only rare moments of noticeable (or valuable) change throughout the process. Applying moda of repetition or replication as measures for performance is antithetical to how we fundamentally view the creative milieu, which prioritizes the value of a singular unique over a copy. Leading to the question: should we value a work of art for its concept, inclusive of all iterations; or should the process attempt to result in a rare collection of curated choices? Or, How do we determine which iterations have value when the distinguishing variables are so subtle?