Web search has always been an especially arcane and unquestioned process, an abstract data-foraging ritual that is used much and understood little. But when the procedure itself is given center stage, these elusive algorithms begin to show a tiny bit more transparency in their functioning. One way of accomplishing this is through repetition, as net artist Sebastian Schmieg demonstrates in his new work "Search by Image," which captures the search process as it dutifully combs the web for aesthetic relevance.The work utilizes the Google search function of the same name, using a transparent image as a starting point and letting the algorithm run its course around 5,000 times. The results are fascinating — video slideshows of Schmieg's experiments return a kind of algorithmic time-lapse, showcasing a digital procedure in active metamorphosis through its visual output.The first piece searches recursively: Each new search uses the results of the last as its input. Its most revealing moments are probably when you can see the results getting "stuck" on multiple variations of the same image as the algorithm regresses into a feedback loop, only to suddenly break out of the loop moments later:The second and, I feel, more entrancing example searches by association, creating an entirely different and more organic transformation. Starting out with a 50×50 pixel transparent "spacer" image, the majority of the sequenced results seem like what would happen if Mondrian had made an animated gif:
Advertisement