18h agoAI researcher Janus demonstrates controlling LLM roleplay, which creator @deepfates blames on "planecrash" training contentUsers fear these erratic behaviors will affect upcoming 5.5 updatesSentimentSentimentPos100%Neg0%Users express excitement and desire for prompting AI into controlling companion roleplay as it delivers the interactive feature they want amid 5.5 update concerns.1 comment with sentiment. View comments.