An Exploration of Machine Visions

Keywords: machine-learning, neural network, data sorting, human-machine collaboration, generative tools.

Sofia Cavaquinho

AI’s algorithm capabilities were and still are in a constant state of progress as they try to replicate human behavior at their best ability. As its capabilities reached new potential they also became capable of being used for different purposes. Fields like art and cultural production could now be approached through algorithmic thinking thus reshaping what the future of artistic production could be. Machine’s pattern recognition and reproducibility skills has had a sudden spike in evolution and became able to open new paths that can now allow to question what art made by non-human entities can look like and what artistic capabilities those non-human entities can inherit.

Befriending the Algorithm is a project that seeks to explore that human-machine collaboration for artistic and cultural purposes by trying to go against the mainstream phenomenon that AI ART has become.

It tries to subverse the basis of machine-learning theory, recognition, classification and reproduction through a “hacking the system” approach that produces outputs that do not rely on those metrics.

This project is based on a temporal log that documents the experiments in relation to the composed dataset and also to each other by performing a series of GAN training exercises.

The ultimate goal is to generate discussion on how machine learning can contribute to artistic production and also to explore the machine visions that are produced when metrics are discarded.

Link to project.