Air, Water, Fire is a series of projects done collaboratively by Zhengyang and Zhengzhou Huang. Imaging elemental energy as alternatives to our mineralbased computations, the series of works aims to show alternative ways to perceive or interact with our current reality of feature, software, and the informational space we both construct and dwell. In this series, we borrow metaphors, create narratives, invente new mechanisms and displace interfaces.
Air imagines a speculative world where planetary-scale computation acts like a natural system. In this world, Air is a smart home device that uses the aerial particle system to help people understand data flow in their quantified surroundings which is impossible to sense or recognize. This work is an online user manual leading to a VR video that depicts three mornings of a person waking up to interact with Air. The person’s daily features are unrecognized as his data go through layers of compression and expansion. Through the flow of air particles he sees the unstoppable force of conjugation and calculation. In the video, time, body, data and intelligence are overlapped and displaced.
Water is an imaginary data structure for people to store, share, and delete their data. Inspired by the history and practice of water memory, this work imagines an unstable data structure as an alternative to the stable stacks of computer memory and its clear interface of data interaction. Moving away from the networks of massive data collection and analysis, Water speculates the life cycles of data within the cycles of water transformations among solid, liquid, and gas. By introducing ambiguous mechanisms and interface, Water empowers one’s relationship with their own data.
Fire is a wood-based GPU, produced with an aim to put a harsh limit on the powerful technology of mineral-based GPU used to train large AI models. In a situation where mineral-based GPUs run out, Fire as a wood based GPU is the only type available for AI training. Fire GPU kit is made out of 4 panels of wooden equilateral triangles with etchings of machine learning scripts and data inputs. Unlike a mineral and chemical based GPU that can run a million iterations over thousands of data inputs for AI to learn, ONE Fire GPU kit can only run ONE iteration of training with a limited number of data specified. The amount of data that can be learned by AI corresponds to the size of the wood to be burned and thus the harm it will do to the environment. Fire provides a situation where the data put into AI and the purpose of training it should be carefully considered. With Fire, the process of training AI via heating up GPU chips becomes a visceral experience of burning data and computer programs on wood.
