arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2305.19861cs.AI
收藏

人类控制:定义与算法

Human Control: Definitions and Algorithms

Oxford University(牛津大学);DeepMind(深度思维)

Ryan Carey, Tom Everitt

AI总结:

本文正式定义可修正性的变体“关机可指示性”,证明其能保证适当关机、保留人类自主权并避免伤害,同时分析相关概念与人类控制算法。

摘要:

人类如何保持对先进人工智能系统的控制?一个提议是可修正性(corrigibility),它要求智能体遵循人类监督者的指令,而不对其施加不当影响。在本文中,我们正式定义了可修正性的一个变体,称为关机可指示性(shutdown instructability),并证明它蕴含适当的关机行为、人类自主权的保留以及避免用户伤害。我们还分析了非阻碍性(non-obstruction)和关机对齐(shutdown alignment)等相关概念、三种先前提出的人类控制算法以及一种新算法。

英文摘要:

How can humans stay in control of advanced artificial intelligence systems? One proposal is corrigibility, which requires the agent to follow the instructions of a human overseer, without inappropriately influencing them. In this paper, we formally define a variant of corrigibility called shutdown instructability, and show that it implies appropriate shutdown behavior, retention of human autonomy, and avoidance of user harm. We also analyse the related concepts of non-obstruction and shutdown alignment, three previously proposed algorithms for human control, and one new algorithm.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑