ChatGPT can now ‘remember’ what you like and how you want it to talk

The feature is undergoing testing at the moment and OpenAI hasn't specified when it'll be rolled out widely as of yet.

ChatGPT's new memory feature. (Image: OpenAI)

ChatGPT is receiving an upgrade – the ability to remember your previous conversations to provide more customised and helpful responses. It’s currently being tested with a small number of free and paid Plus users.

The addition of memory represents an important step in making conversations with ChatGPT feel less repetitive. Up until now, users have often had to re-explain preferences around things like email formatting every time they ask ChatGPT to draft a message. But with memory enabled, the bot can simply recall those choices for future interactions.

Some examples highlighted by OpenAI include:

– Recalling that a user prefers meeting notes formatted with headlines, bullet points and action items. Subsequent notetaking performed by the bot will automatically apply this structure.

Story continues below this ad

– Remembering key details when a coffee shop owner asks for messaging ideas about a new location opening. ChatGPT can seamlessly incorporate that context.

– Suggesting a jellyfish wearing a party hat when helping a parent create a birthday card for their toddler who loves jellyfish.

– Tailoring lesson plans to a teacher’s preferred 50-minute structure and follow-up activities based on previous conversations.

Users have full control over what ChatGPT remembers. The bot can be explicitly asked to commit specific details to memory. Alternatively, it will passively pick up relevant information from regular chat. Memories can also be viewed and deleted individually or collectively at any time.

Story continues below this ad

For those wanting private conversations, the feature also provides temporary chats that won’t be remembered or used to train its algorithms.

If you are one of the few users who’ve been selected for testing out this feature then there are a couple of things you need to keep in mind. The feature is turned on by default and OpenAI also says that these memories will be used to train its models going forward. If that bothers you, you can turn memory off via Settings > Personalisation > Memory.

There are exceptions though. OpenAI says it’s trained its system not to remember sensitive things like information about your health. Also, companies using ChatGPT Enterprise will not have their data used for chatbot training.