Cloud Theory is a relaxing videogame where you lay on the ground and look at clouds. Every once in a while you or your friend might see a shape in the clouds, which sparks conversation about everything from the origin of poodles to the deepest mysteries of life. Launching sometime in 2023.
Cloud Theory is part I of the Watching Simulators series, a triptych of games about looking. In videogames, a ‘walking simulator’ is a term used to describe games where moving through the world is a core principle. Taking this ponderous form one step further, the Watching Simulator series focuses on the simple act of looking, providing an ambient form of entertainment that is based on observation and the wanderings of the mind.
Cloud Theory is also part of a research project called Machine Unlearning, an exploration of the ways in which “intelligence” is within the mind of the beholder, not the algorithm. Cloud Theory explores apophenia, the psychological phenomenon of seeing patterns where none exist. We see shapes in clouds, but we also apply meaning to the GPT-based dialogue. If language based machine learning models are “autocorrect on steroids”, our minds are ‘apophenia engines’ that perceive meaning, intentionality, and life in digital worlds. How different is this process in simulated worlds from the meaning we perceive in the real world? Is the mind a sophisticated probability engine? As Anil Seth describes, is perception a form of hallucination? Machine Unlearning is an attempt to ask these and other questions about the nature of intelligence and perception in an entertaining and playful way.
For more on this project, see this presentation at WordHack from February 16th, 2023.
Cloud Theory has been supported by funding from the New School Research Assistantship Grant and the New York City Media Lab.
Support from: Jason Li (programming, conversation engine), Anna Garbier (dialogue design and GPT tuning), Emi Sato (computer vision), Apurv Rayate (programming), Lavannya Suresh (machine learning), and Terry Kahn (visual effects, programming).