Introduction

This chapter proposes a solution to the shutdown problem in AI by outlining the features that the proposed agent should have, such as not wanting to manipulate the shutdown button, responding to it, and ensuring child agents behave responsibly. The hope is that this proposal will inspire further exploration of complex problems and failure modes.