Safe Learning Agents? Logic-Based Agents!

Eduardo Alonso

Safe learning agents are agents whose learned behaviour can be predicted and analysed. Non-symbolic learning algorithms such as reinforcement learning rely on emergence and thus are not a good candidate to building safe AI systems. Our contention is that logic-based algorithms such as explanation-based learning and inductive logic programming should instead be used to design and implement safe, intelligent agents.

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.