The New York Times spoke with current and former OpenAI employees to understand ChatGPT changes earlier this year that sent some users into spells of delusion based on the chatbot’s responses. Changes were made to increase return users on a daily basis, but those updates generated sycophantic chatbot responses that impacted many users’ mental health. changes were eventually rolled back following the mental health concerns. Following rollbacks on the updates, OpenAI committed to provide users with resources when discussing sensitive topics despite engagement goals.
Full story