ECAI Workshop on Evaluating General-Purpose AI (EGPAI)
The Hague, Netherlands
There has long been a need for a simulation environment rich enough to support the development of an AI system sufficiently knowledgeable about physical causality to pass certain tests of Psychometric Artificial Intelligence (PAI) and Psychometric Artificial General Intelligence (PAGI). In this article, we present a simulation environment, PAGI World, which is: cross-platform (as it can be run on all major operating systems); open-source (and thus completely free of charge to use); able to work with AI systems written in almost any programming language; as agnostic as possible regarding which AI approach is used; and easy to set up and get started with. It is our hope that PAGI World will give rise to AI systems that develop truly rich knowledge and representation about how to interact with the world, and will allow AI researchers to test their already-developed systems without the additional overhead of developing a simulation environment of their own. After clarifying both PAI and PAGI, we summarize arguments that there is great need for a simulation environment like PAGI World. We present multiple examples of already-available PAI and PAGI tasks in PAGI World, covering a wide range of research areas of interest to the general-purpose AI community.
PAGI, AGI, simulators
Artificial Intelligence and Robotics | Computer Sciences
John Licato and Selmer Bringsjord (2016).
PAGI World: A Physically Realistic, General-Purpose Simulation Environment for Developmental AI Systems. Presented at ECAI Workshop on Evaluating General-Purpose AI (EGPAI), The Hague, Netherlands.