Infinite Adversaries uses ChatGPT to generate perpetual, randomized encounters, pitting you against a never-ending army of imaginary enemies. Inspired by the classic “Choose Your Own Adventure” books from the 1980's, as well as Dungeons and Dragons, this project is a new take on an old theme: stories written and narrated by Artificial Intelligence.
You begin by discovering a table, and are presented with a choice of weapon. Behind the scenes, ChatGPT is prompted to create a random list of weapons, which changes with each visit. This weapon is then equipped for every encounter.
For each encounter, ChatGPT is asked to pick an adversary at random, and to imagine a location where a physical encounter is imminent. ChatGPT also imagines a set of four possible actions to take, as well as the outcomes of each action. Some actions result in damage. If you survive an encounter, you move on to another.
The actions and outcomes are influenced by the weapon you selected. To visualize the encounter, DALL-E is used to generate a random image based on the weapon, adversary, and location. In fact, as part of the prompt, ChatGPT is asked to come up with a one-sentence summary of the encounter, appropriate for a DALL-E image prompt. Because hey - why have a human tell an AI what to do, when you can have an AI tell another AI what to do? What could possibly go wrong there?
