Can we Build a Superintelligence Without Being Superintelligent?

Abstract

If we create an entity of greater intelligence to us, a superintelligence, it has the possibility to explode in intelligence, creating more and more intelligent entities. If the intelligence explosion argument holds, then the most important step to developing a powerful superintelligence is the development of the first superintelligence. This paper presents objections to the possibility of humans developing this first superintelligence. I argue that this is because we lack required knowledge about them, due to our epistemic position of not being superintelligent. I engage primarily with arguments from David Chalmers and Nick Bostrom about what superintelligences are and the nature of the intelligence explosion. I add my own detail to these areas and explore how to increase intelligence. I argue that my objections stem from flawed expectations of superintelligence such that we ought to change them. I then present my own alternative expectations for superintelligence.