Algorithmic bubble ecology of Deep Dream Generator page on Facebook

Abstract

In this work we seek to demonstrate the relationship between the algorithmic image tracking mechanism (neural networks) and the deep dream aesthetic, which produces images through the intervention of this tracking mechanism. For this, we start from the description of how neural networks work, demonstrating how they are part of the Google and Facebook organization algorithm, until we arrive at the creation of the deep dream image production algorithm. The focus of this investigation is the Deep Dream Generator page on Facebook, from which we were able to observe how the deep dream images themselves are tracked and organized on this site. The peculiarity found on the analyzed page is that it has a social media application that allows image editing for deep dream plastic. Given the structural complexity of the analysis focus, we set up a method consisting of documentary digital ethnography and network analysis using the Netvizz application and the Gephi software. As a result, we have found that the ecosystem that produces the algorithmic bubbles is beyond the limits of each social networking site, being a network of connections in constant transformation.

Keywords: Algorithmic bubble. Neural networks. Deep dream. Ecology. Cyberculture.