Low Impact Artificial Intelligences

Abstract

There are many goals for an AI that could become dangerous if the AI becomes superintelligent or otherwise powerful. Much work on the AI control problem has been focused on constructing AI goals that are safe even for such AIs. This paper looks at an alternative approach: defining a general concept of `low impact'. The aim is to ensure that a powerful AI which implements low impact will not modify the world extensively, even if it is given a simple or dangerous goal. The paper proposes various ways of defining and grounding low impact, and discusses methods for ensuring that the AI can still be allowed to have a (desired) impact despite the restriction. The end of the paper addresses known issues with this approach and avenues for future research.

References in corpus (1)

• Game theory, maximum entropy, minimum discrepancy and robust Bayesian decision theory

Cited by in corpus (9)

• Scalable agent alignment via reward modeling: a research direction
• AI Safety Gridworlds
• Penalizing side effects using stepwise relative reachability
• Avoiding Side Effects By Considering Future Tasks
• Evolutionary Computation and AI Safety: Research Problems Impeding Routine and Safe Real-world Application of Evolution
• Challenges for Using Impact Regularizers to Avoid Negative Side Effects
• Towards Empathic Deep Q-Learning
• Learning the Arrow of Time
• Combining Reward Information from Multiple Sources