OpenAI has announced the commencement of testing memory capabilities for ChatGPT. This memory feature allows the chatbot to recall information from previous conversations, eliminating the need for users to repeat themselves and making future interactions more beneficial.

Users will have full control over ChatGPT’s memory through the chatbot’s settings section. They can instruct the bot to remember specific details during the chat, verify what has been remembered, or request the bot to forget certain information.

ChatGPT’s memory will improve with usage, gradually enhancing the quality of interactions. For example:

• If a user explains that meeting notes should include summarized headings, bullet points, and an action list at the end, ChatGPT will remember this and format all future meeting notes accordingly.
• If a user mentions they have a young child who loves jellyfish and asks for help creating a birthday card, the chatbot might suggest drawing a jellyfish with a party hat.
• If a user works as a preschool teacher and typically designs 50-minute lessons for a group of 25 children, ChatGPT will remember this and take it into account when creating lesson plans.

The company emphasized that users can disable the memory feature at any time (Settings -> Personalization -> Memory). When memory is turned off, ChatGPT will neither save nor use any previous information.

More detailed information on managing memory can be found in the Help Center.

In the coming weeks, the memory testing will involve a small portion of free ChatGPT users and Plus subscribers. Further details about broader testing plans will be announced later.