1. Introduction

Many have argued that in the next twenty to one hundred years we will create artificial general intelligences (AGIs). Unlike current 'narrow' AI systems, AGIs would perform at or above the human level not merely in particular domains (e.g., chess or arithmetic), but in a wide variety of domains, including novel ones. They would have a robust understanding of natural language and be capable of general problem solving.

The creation of AGI could pose challenges and risks of varied severity for society, such as the possibility of AGIs out competing humans in the job market. This article, however, focuses on the suggestion that AGIs may come to act in ways not intended by their creators, and in this way pose a catastrophic or even an existential risk to humanity. We will organize and summarize the proposals that have been made so far for responding to catastrophic AGI risk, so as to provide a map of the field to newcomers and veterans alike.

to read the rest of the article, click HERE