Meta Works to Prevent Inappropriate Replies From New AI Chatbots
Facebook-parent has been conducting stress testing of new AI-based chatbots for several thousand hours, but still cannot give an absolute guarantee that the digital thinking system will not say something undesirable.