OpenAI introduces memory feature for ChatGPT: How it works
OpenAI is introducing a "memory" feature for ChatGPT, designed to make the chatbot more personal and intelligent. This addition allows the bot to recall user-specific details and conversations, reducing the need for constant repetition. Memory functions in two ways: users can directly ask ChatGPT to remember something, or the bot can learn information through ongoing interactions. Each customized GPT has its own memory, improving user experience across various applications like Books GPT, Tutor Me, Kayak, and GymStreak.
Custom memory for different GPTs
However, the new memory feature raises questions about user privacy and data management. OpenAI assures users that they remain in control of ChatGPT's memory. The system is designed not to store sensitive information, such as health data. Users can ask ChatGPT what it knows about them and either instruct it to forget specific details or remove them in the 'Manage Memory' section of settings.
User control and privacy measures
Users can also use the Temporary Chat option. It acts as an incognito mode for brief conversations without affecting memory. Users can also disable memory entirely across their accounts. However, memory will be turned on by default and shall be used by OpenAI to train its models going forward. This excludes data from firms using ChatGPT Team and Enterprise.
What about its rollout?
Currently, the memory feature is being tested by OpenAI with a select group of users. A broader rollout is expected soon. Once it is done, this might become an important part of the way users interact with ChatGPT.