arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.11441cs.AIcs.HCcs.MAecon.GNq-fin.EC
收藏

民主AI的人性化机制设计

Human-centered mechanism design with Democratic AI

DeepMind;University of Exeter(埃克塞特大学);University College London(伦敦大学学院);University of Oxford(牛津大学)
Raphael Koster, Jan Balaguer, Andrea Tacchetti, Ari Weinstein, Tina Zhu, Oliver Hauser, Duncan Williams, Lucy Campbell-Gillingham, Phoebe Thacker, Matthew Botvi… 展开作者

Raphael Koster, Jan Balaguer, Andrea Tacchetti, Ari Weinstein, Tina Zhu, Oliver Hauser, Duncan Williams, Lucy Campbell-Gillingham, Phoebe Thacker, Matthew Botvinick, Christopher Summerfield

摘要:

构建符合人类价值观的人工智能(AI)是一个尚未解决的问题。在这里,我们开发了一个名为Democratic AI的人在回路研究管道,其中使用强化学习来设计一种人类大多数人喜欢的社会机制。一大群人玩了一个在线投资游戏,其中涉及决定是否保留货币捐赠或与他人分享以获得集体利益。共享收入通过两种不同的再分配机制返还给玩家,一种由人工智能设计,另一种由人类设计。人工智能发现了一种机制,纠正了最初的财富不平衡,批准了搭便车,并成功地赢得了多数票。通过优化人类偏好,民主人工智能可能是一种有前途的价值取向政策创新方法。

英文摘要:

Building artificial intelligence (AI) that aligns with human values is an unsolved problem. Here, we developed a human-in-the-loop research pipeline called Democratic AI, in which reinforcement learning is used to design a social mechanism that humans prefer by majority. A large group of humans played an online investment game that involved deciding whether to keep a monetary endowment or to share it with others for collective benefit. Shared revenue was returned to players under two different redistribution mechanisms, one designed by the AI and the other by humans. The AI discovered a mechanism that redressed initial wealth imbalance, sanctioned free riders, and successfully won the majority vote. By optimizing for human preferences, Democratic AI may be a promising method for value-aligned policy innovation.

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。

↑