ChatGPT will soon be able to remember things
OpenAI is planning to roll out a memory feature for ChatGPT. Users can let the chatbot pick up details, or they can ask ChatGPT to remember something in the conversation. Memory will roll out to a select number of users, with plans to eventually make the feature broadly available.
Memory will enable ChatGPT to remember things discussed across all chats, making it more helpful and saving users from repeating things in conversations. Users will control ChatGPT's memory by asking it to remember or forget things within their conversations with the chatbot or shutting it off altogether from settings. From February 13, memory in ChatGPT will be rolled out to a small number of free and paying users for testing purposes. Plans to make the feature generally available are yet to be determined.
Although users always have the option to let ChatGPT know that they would like the chatbot to remember a piece of information, memory is also meant to enable ChatGPT to pick up details by itself. For instance, if users specify that they would like ChatGPT to take meeting notes following a specific format, then the chatbot will remember this and present all meeting notes in the preferred format. Moreover, since memory is not linked to conversations, deleting a conversation will not affect ChatGPT's memory. Memories should be deleted by asking ChatGPT to forget or by deleting all memories via settings.
With the introduction of memory, OpenAI will also introduce temporary chats, which will not be stored in the conversation history and will not use memory. Thus, users should start a temporary chat with ChatGPT if they do not want it to generate memories. Independently of whether ChatGPT is using memory or not, Custom Instructions are still available to make the chatbot more helpful. They can easily be combined: Custom Instructions can hold any explicit information or instructions that ChatGPT should consider when replying to users' prompts, while memory can pick up complimentary information in conversations.
Unless individual users opt out, their conversations and memories will be used to improve OpenAI's models. This does not apply to Team and Enterprise users since their data is never withheld for training. Moreover, administrators in Enterprise accounts will have the option to turn memory off for the entire organization. Enterprise and Team access to memory will be part of the general rollout. Likewise, GPTs will get memory once the feature is more broadly available.
