Calvin Wankhede / Android Authority
TL;DR
- OpenAI has announced a new memory feature for ChatGPT.
- ChatGPT can now remember context and details across conversations. It can pick these up on its own or be told by the user to remember something specific.
- Users can turn off the memory feature or use temporary chats to have conversations without memory.
ChatGPT is a great tool that serves various needs, whether it be saving you from writer’s block or planning a vacation on a budget. But one of the downsides with the AI virtual assistant is that you need to specify everything you need on every new instance of your conversation. While ChatGPT remembers conversational contexts within that conversation, it does not carry it forward to other conversations in the future. OpenAI is now gifting ChatGPT the power of memory, letting it remember things if you want it to.
OpenAI has announced that it is testing memory with ChatGPT. Memory lets ChatGPT remember things you discuss across all chats and saves you from repeating information. Memory is rolling out to a small portion of ChatGPT free and ChatGPT Plus users, and the company will announce a broader rollout soon.
Some examples of its memory functions include telling ChatGPT you own a neighborhood coffee shop. The next time you brainstorm messaging for a social media post, ChatGPT will remember this fact. In another instance, you can mention your preference in seeing your meeting notes summarized in a certain way, and ChatGPT will remember it across meetings.
Pre-emptively addressing concerns, OpenAI says that users can control ChatGPT’s memory. Users can tell it to remember something (while it will also just remember things conversationally), but you can also turn off the feature entirely. If you prefer a middle ground, you can ask ChatGPT what it remembers and tell it to forget that fact conversationally or through settings.
Further, you can also converse in temporary chats to have conversations without memory (with the added benefit of them not appearing in history and not being used to train models).
OpenAI does reiterate that the content you provide ChatGPT, including memories, could be used to improve its models for everyone. Users can turn off this behavior through Data Controls. Content from ChatGPT Team and Enterprise customers will not be used for training.
The company also announced that custom GPTs will also have their own distinct memory, though enabling memory will be an option exercised by the builder. Memories from users are not shared with builders.
What do you think of the memory feature on ChatGPT? Let us know in the comments below!
