Mirroring Reality: Surreal Tourist Gaze of Chinese Landscape Painting by Artificial Neural Networks

Abstract

In 2015, Google released the "Deep-dream" algorithm based on deep learning of artificial neural networks. This method reverses the path of feature capture in artificial intelligence image recognition. Instead, the artificial neural network captures hidden image features in the reverse direction. The original image is transformed into a fantasy world with dream-like animals and strange geometric shapes by iterating, reinforcing, and deepening. This dreamlike visual encounter demonstrates more than mere filtering of contrived images. The deep-dream algorithm inspires a bionic view of how artificial neural networks view and interpret the world. With this view, the algorithm constructs a bionic vision that reassembles, collages, and even emerges with species features. We explore the world of the Deep-dream in Chinese landscape painting from the viewpoint of the tourist gaze. The possible connections between AI and creativity and imagination are explored by comparing the work with that of Midjourney, a powerful prompt-based AI-generated art platform.