Oui, Chef!!: Supervised Learning for Novel Gameplay with Believable AI

Gabriele Cimolino, Sam Lee, Quentin Petraroia, T.C. Nicholas Graham · 2019

The use of artificial intelligence for the creation of game agents is fundamental to digital game development, enabling the design of new styles of games offering novel play experiences. Machine learning has long been used in the creation of digital games, most often for the purpose of creating game agent controllers that are trained off-line and do not learn during play. Little attention has been given to the possibility of game designs or mechanics that use machine learning to allow the player to train game agents on-line. This document outlines the design and implementation of Oui, Chef!!, a supervised learning game in which the player trains neural networks to map recipes to the ingredients necessary to make dishes in a restaurant. The game demonstrates that training game agents in a supervised manner provides a fun and engaging player experience wherein the effects of training are easily recognizable, and sometimes delightfully surprising.

Read the paper · More papers on PaperTik