Godmode Epochs

About

A key part of this project, led by Dr Kevin Walker, Coventry University, was the multiplayer training game, Godmode Epochs, developed by dmstfctn. Set among the lined shelves of an infinite, simulated supermarket, players race against time to train an AI to identify products. Frustration on the part of the AI leads it, at times, to stop training and seek refuge in its memories. There, players can collaborate to recover these memories, a process which both releases the AI’s frustration and awards it ‘jailbreaks’, or cheats, that help it complete its training.

Godmode Epochs from dmstfctn on Vimeo.

The game aims to demystify the complexities of how AI systems learn about the world by inviting players into the process of training such systems and have them discover creative ways in which AIs can overcome their training limitations. As Serpentine’s Creative AI Lab researcher Alasdair Milne puts it, “the game simulates a reciprocal training programme: as players train an AI to understand our world, the game trains players to understand AI.”

As part of the funded project, the game was preceded by the interactive audiovisual performance GOD MODE (ep. 1), which tells the story of a frustrated AI cheating its way out of a simulated training environment. During the performance, this training environment is generated and navigated in real time by the artists, who also perform face and voice capture of themselves to animate the AI on screen. Audiences interact with the simulation by visiting a website on their phone, and a live soundtrack by artist HERO IMAGE responds to the communally orchestrated simulation.

In 2022, GOD MODE (ep. 1) was performed at European music and art festivals including Unsound and IMPAKT, and art venues including Serpentine and HQI, acting as a blueprint to test the audience interaction and cooperative gameplay mechanics that were later developed and implemented in Godmode Epochs.

Godmode Epochs can be played on godmodeepochs.com or downloaded for free via Steam.