On the Broad Implications of Reinforcement Learning based AGI

Reinforcement learning (RL) is an attractive machine learning discipline in the context of Artificial General Intelligence (AGI). This paper focuses on the intersection between RL and AGI by first speculating on what are the missing components that would facilitate the realization of RL-based AGI. Based on this paradigm, we touch on several of the key moral and practical issues that will inevitably arise.