Openai has updated Chatgpt behavior-now the chat bout will no longer give direct tips in response to personal and sensitive questions, like “throw a girl or not?”
The model was taught to help the user to reason, and not make categorical solutions. Instead of an answer, she will ask clarifying questions, offer to weigh the pros and cons, so that the person himself comes to the conclusion.
The company explained this by the desire to make ChatGPT truly useful, especially in difficult situations. They also admitted that in the previous update the model became “too consonant” – now this behavior has rolled out.
In long sessions, ChatGPT will begin to gently remind the user of the need to rest. How exactly these tips will work, it is still adjusted so that they are perceived naturally.
In addition, Openai creates an expert group with the participation of specialists in mental health, teenage psychology and interaction with AI. More than 90 doctors from 30 countries were also attracted to work.
