Abstract

A long-standing goal of Artificial Intelligence is to program computers with common sense. Virtual worlds could be especially valuable in this endeavor because they represent people and objects in 3-D space interacting with a world similar in many ways to our own. I propose the Human Experience Project, a new type of annotated corpus for collecting common sense, based on scenes in a virtual world. The proposal calls for volunteer contributors over the internet to create 3-D scenes about everyday life. The contributors' annotations will describe the objects in the scenes, the actions, and the hidden mental states of the actors. The goal is to amass a collection of scenes and annotations that researchers can use to model common sense. I present a sample scene and outline the broad requirements of this ambitious project.

Share

COinS