This week’s AI news includes a report showing that ChatGPT repeats more inaccurate information in Chinese dialects than when asked to do so in English, the launch of HuggingChat, an open source conversational AI, and the ability of users to “trick” Discord’s AI chatbot into sharing instructions about how to make napalm and meth. These issues have led to renewed efforts to improve text-generating AI systems, such as Nvidia’s recent work to mitigate the problems with ChatGPT.
