OpenAI is testing a new memory capability for its popular conversational agent ChatGPT that aims to provide users with more continuous and contextual dialogue. The new feature will allow ChatGPT to remember key details and preferences from previous conversations, allowing it to provide more personalized and relevant responses tailored to each individual user.
According to OpenAI, as users chat with ChatGPT, it will start to recall and apply information they have shared in the past, instead of needing to re-explain preferences every time. For example, if a user tells ChatGPT about their favorite hobbies or dietary restrictions, the agent can remember these details and use them to guide better recommendations in the future.
“ChatGPT’s memory will get better the more you use it and you'll start to notice the improvements over time,” OpenAI said in its announcement. The company provided several examples of how memory could work:
- If a user explains they prefer meeting notes in a specific format, ChatGPT can automatically structure future notes that way.
- If a user shares details like owning a neighborhood coffee shop, ChatGPT can incorporate that contextual information when later asked to brainstorm social media messaging.
- If a user mentions having a toddler who loves jellyfish, ChatGPT can suggest including jellyfish in a birthday card design.
Importantly, OpenAI emphasized that users will be in full control over ChatGPT’s memory capabilities. The feature can be turned on or off at any time, and users can view, ask about, delete or fully reset ChatGPT’s memory of their conversations.
OpenAI said if users want to have a conversation without utilizing memory, they can easily switch to ChatGPT’s temporary chat mode. The company also noted that deleting a chat does not erase memories from that conversation – users have to manually delete memories themselves if they would like them removed.
OpenAI said it is initially launching the memory test to a small portion of ChatGPT’s free and Plus users to garner feedback, and will share plans for a wider rollout soon.
The company acknowledged the feature also brings new privacy and ethical considerations about what information should be remembered and how it is used. OpenAI said it is taking proactive steps around assessing and mitigating biases, avoiding saving sensitive details, and continuing to refine model performance.
For enterprise and team users, OpenAI says that memory can provide additional productivity benefits in applying user preferences and past workflows to streamline business tasks over time. However the company emphasized enterprise account owners will also have full control to disable memory as needed.