The Verge websiteThe report stated that OpenAI's artificial intelligence service ChatGPT can already remember the identities of users who interact with it, and can even further respond to questions based on user preferences.
This feature, called "custom instructions," is still in beta testing and can only be manually enabled by paid ChatGPT Plus subscribers. During use, users can explain their occupations, preferences, and needs to the ChatGPT service, which will then generate more appropriate responses.
Joanne Jang, who is in charge of behavioral modeling and product projects at OpenAI, explained that this function allows users to indicate that they are school teachers and automatically generate course content suitable for students, or to indicate the number of family members to generate appropriate recipe cooking quantities, and even to respond with more appropriate answers based on different professional needs.
However, such a feature would inadvertently collect more user privacy, so OpenAI will obviously be testing it with greater caution, while also avoiding violating local privacy regulations. On the other hand, many believe that once artificial intelligence knows more about users' private information, it could also pose more potential risks, so OpenAI may also face more opposition to this feature.
