Artificial Open World for Evaluating AGI: A Conceptual Design

By Bowen Xu et al

Published on June 2, 2022

Read the original document by opening this link in a new tab.

Table of Contents

Abstract
1 Introduction
2 What Intelligence is
3 Evaluation
3.1 The Trap of Developer's Experience
3.2 Overall Principles
3.3 Conceptual Design of Artificial Open World Generation

Summary

This paper discusses the concept of Artificial Open World for Evaluating AGI. It addresses the challenge of evaluating Artificial General Intelligence (AGI) and proposes a new evaluation method named Artificial Open World. The paper emphasizes the need for an open world where developers cannot rely on their experience to solve problems, thus avoiding the trap of developers' experience. The design includes principles such as Independence, Similarity, Openness, and Asymmetry. The world generation process involves differentiation, generating causations, and importing the mind to make the world vibrant. The paper aims to provide a new framework for evaluating AGI that is independent of developers' specific knowledge.