Zero likes is a generative, machine learning project exploring the aesthetics of the neglected, negative spaces of the internet. An AI was trained to respond to more than 100, 000 Instagram posts that received zero likes, investigating the potential for machines to respond to abstract, human questions. Another AI was trained to respond to respond to the images generated by the first. The result is a glitchy, unsettling artwork accompanied by disconcertingly straightforward, automated captions.
Likes’ are the primary metrics used by social network algorithms to quantify what is good, interesting content vs bad, uninteresting content. A post with zero likes will be pushed to the edges of the internet where it might be unlikely to be seen by a human again.
By reducing the the possibilities of internet space into a user interface, the platform begins to encourage our behaviour. In the case of Instagram, the structure of the platform informs our aesthetic decisions. It teaches us to create content that maximizes user engagement and share-ability. Predictably, most of the content for the zero likes project came from older Instagram users with limited follower accounts or people who are new to the system. These people have yet to learn from the platform, and are feeling their way through an unfamiliar system. They haven’t learned the tropes yet.
The project exists in its current form as an automated, organic creature operating in a popular social network, or more simply, it is an art bot. The displayed works are selections from a growing collection of over 20,000 images. Images from this collection are being shared daily at @zero_likes on twitter.
PRELIMINARY STUDY TOOL
The tool below was created to study the results of the first experiment with the zero likes data set with DCGAN - a deep learning algorithm.
(It might need some time to load.)