Thomas Depas

Princess of Parallelograms
What will happen when our imagination itself is externalized in machines? Artificial intelligence constructs its own world-truth that is beyond our sensory perception. Generative Adversarial Networks (GANs) use algorithms to synthesize and generate images in a completely new way. These images have almost uncanny aesthetic characteristics, seeming to emerge from an ocean of data, a kind of pixel soup. Rather as if we were observing the emergence of artificial thought.” The machine learns to understand the “essence” of a thing, be it an animal, the face of a celebrity or a body of text. It is then able to generate new images of this thing, including faces of celebrities who do not exist, mutant animals, or new texts. Eventually, AI will be capable of instantaneously and dynamically emulating all representations. The era of the optical machine and the capture of reality will then be at an end, supplanted by the era of machines that generate their own reality.

Chris Cheung

No Longer Write – Mochiji
Powered by artificial intelligence’s Generative Adversarial Networks (GANs), the collected works from ancient Chinese Calligraphers, including Wang Xizhi, Dong Qichang, Rao Jie, Su Shi, Huang Tingjian, Wang Yangming, as input data for deep learning. Strokes, scripts and style of the masters are blended and visualized in “Mochiji”, a Chinese literature work paying tribute to Wang Xizhi. Wang is famous for his hard work in the pursuit of Chinese calligraphy. He kept practicing calligraphy near the pond and eventually turned the pond for brush washing into an ink pond (Mochi). The artwork provides a platform for participants to write and record their handwriting. After a participant finished writing the randomly assigned script from “Mochiji”, the input process is completed and the deep learning process will begin. The newly collected scripts will be displayed on the screen like floating ink on the pond, and slowly merge with other collected data to present a newly learnt script. The ink pond imitates process of machine learning, which observes, compares and filters inputs through layers of image and text, to form a modern edition of “Mochiji”.