Why do chatbots hallucinate, and how to stop yours
中文摘要
聊天机器人因训练机制倾向于预测而产生幻觉，导致虚构政策等问题。本文探讨其成因、代价及预防方法。
English Summary
Chatbots hallucinate because they are trained to guess, causing errors like invented policies. This article explores why this happens and how to prevent such mistakes.
Original Excerpt
Chatbots guess because they were trained to. Why support bots invent policies, what it has cost two companies, and the rules that make one… Continue reading on Medium »