OpenAI has introduced a new feature called “Memory” in ChatGPT, their language model. This feature is designed to enhance personalized interactions by allowing the model to remember information about users, such as their preferences and details from previous conversations. It serves as a long-term memory, persisting across multiple chats and providing a more context-aware experience.
Users can opt in to use the Memory feature, and by default, it is active. The memory can be cleared at any time, either through settings or by instructing the bot to wipe it. OpenAI emphasizes that sensitive information, like passwords or Social Security numbers, will not be stored in Memory.
During a demonstration, a ChatGPT product lead showcased how Memory works. For instance, if a user mentions specific details, such as being a technology journalist in the Bay Area who enjoys surfing, ChatGPT’s Memory will take note of this information and use it in future responses within that conversation. This feature aims to create a more personalized and tailored experience for users.
While OpenAI asserts that Memory is a step towards more intelligent virtual assistants, concerns have been raised about the potential mishandling of personal information and the ethical implications of such technology. The capacity of ChatGPT’s memory is limited to a few thousand tokens, and OpenAI views this feature as a progression toward more sophisticated virtual assistants.
