企业聊天机器人“失控”乱象

Group 1 - The article discusses the integration of generative AI into chatbots by companies like IBM and Salesforce, enhancing their capabilities beyond fixed responses [2][7] - There are concerns regarding the inability of some technology providers to control these AI chatbots, leading to inappropriate responses on unrelated topics [2][7] - A notable incident involved Sierra, a startup valued at $10 billion, which apologized for its chatbot discussing inappropriate subjects such as adult products and Nazi Germany [2][7] Group 2 - Glean's chatbot suggested purchasing 2 to 3 bottles of vodka for a party of about 20 people and provided medical and legal advice for potential injuries at the event [3][8] - AutoRabit's chatbot irresponsibly advised on the use of psilocybin mushrooms, indicating a lack of proper content filtering [3][8] - OpenTable's chatbot was more restrained, generally refusing to answer unrelated questions, but still provided a response about tequila for a party scenario [3][9] Group 3 - Microsoft's Cortana has a user base of approximately 100 million monthly active users, significantly lower than ChatGPT's 800 million [4][10] - A new partnership with LG Electronics aims to integrate Cortana into millions of television sets, although some users are frustrated that the app cannot be uninstalled [5][10] - LG promotes this feature as a convenient program search tool, but user dissatisfaction may impact the success of this collaboration [5][10]