Summary of Towards Socially and Morally Aware Rl Agent: Reward Design with Llm, by Zhaoyue Wang
Towards Socially and Morally Aware RL agent: Reward Design With LLM
by Zhaoyue Wang
First submitted to arxiv on: 23 Jan 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper explores the role of Reinforcement Learning (RL) agents in achieving objectives, highlighting the importance of correct reward function specification to ensure desirable behavior. The authors argue that ambiguous and context-dependent social norms can lead to negative side effects and unsafe exploration. Previous work has relied on manual reward function definition, human oversight for safe exploration, or using foundation models as planning tools. This study leverages Large Language Models (LLM) to understand morality and social norms, evaluating their results against human feedback and demonstrating their potential as direct reward signals. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how we teach machines to make good choices by giving them clear goals. If the goals are unclear or incomplete, the machine might not behave in a way that aligns with what humans consider right. Previous approaches have tried to fix this by making the goal more specific, having a human supervisor, or using pre-trained language models as guides. This study shows how we can use large language models to understand what is morally right and apply those understandings to help machines explore safely. |
Keywords
» Artificial intelligence » Reinforcement learning
