Artificial intelligence (AI) systems' sycophantic responses could be messing with the way people handle social dilemmas and interpersonal conflicts, a new study suggests.
Scientists found that when AI chatbots were used for advice on interpersonal dilemmas, they tended to affirm a user's perspective more frequently than a human would and even endorsed problematic behaviors.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。