Frosty Morning Prototype
2018
Auido-Visual Installation
large-scale screen, loud speakers
04:33 loop"

Frosty Morning Prototype is a single channel Audio-Visual work based on artist’s research on Machine Learning and Neural Networks. The original architecture of the code is designed to train a machine which can predict future images from consecutive video frames. After all the research, the artist decided to use the front page of the newspaper as a narrative view of our world instead of the video sequence, then downloaded all the front page of The New York Times during the last 6 years and used them as a data set. So after training a machine to see daily newspapers from last six years, it becomes a sort of perpetual newspaper generator or probably a machine trying to predict “future.”

However, it is actually not a correct approach to predict by simply using front page of newspaper images. In the process of training, this machine only reads every pixel which is the result of deconstruction. Nevertheless, when we review the training process and generated results, the blurred image seems to reflect the reasonable and overwhelming world in the present.