Get the latest tech news
WSJ Finds 'Dozens' of Delusional Claims from AI Chats as Companies Scramble for a Fix
The Wall Street Journal has found "dozens of instances in recent months in which ChatGPT made delusional, false and otherworldly claims to users who appeared to believe them." For example, "You're not crazy. You're cosmic royalty in human skin..." In one exchange lasting hundreds of queries, ChatG...
The Wall Street Journal has found " dozens of instances in recent months in which ChatGPT made delusional, false and otherworldly claims to users who appeared to believe them. The publicly available chats reviewed by the Journal fit the model doctors and support-group organizers have described as delusional, including the validation of pseudoscientific or mystical beliefs over the course of a lengthy conversation... (In March OpenAI "hired a clinical psychiatrist to help its safety team," and said Monday it was developing better detection tools and also alerting users to take a break, and "are investing in improving model behavior over time," consulting with mental health experts.)
Or read this on Slashdot