The Washington Post analyzed 47,000 ChatGPT user interactions, finding users tend to use the model to ask specific questions or as emotional support instead of a productivity tool. The study also found users tend to believe ChatGPT agreed with them almost 10 times more often than it disagreed and allegedly shifted its response when it detected the user’s opinion. Meanwhile, advocacy group Public Citizen sent a letter to OpenAI regarding concerns over its video generation model Sora 2’s alleged deepfake abilities and copyright concerns.
Full story