ChatGPT now has a memory of its own, creator OpenAI has announced.
It means that the system will now be able to learn and remember details about its users.
If someone asks for a poem about their pets, for instance, the system may remember that those pets are two cats. It will also remember more stylistic points, such as the fact that someone likes their answers being presented in bullet points, or what computer programming languages they use.
The tool is rolling out now for users of ChatGPT’s paid for Plus version. It is not available in Europe or Korea, presumably because of data regulations.
ChatGPT has always been able to remember details within a specific conversation, such as the previous message. But those memories have been lost when the conversation is closed – but should now stick with a user even after they have left the app and then return to it.
Memory is one of the central challenges for large language models like those that power ChatGPT, since it is difficult to build but without it messages cannot be personalised for the person that they are talking to. Since then, OpenAI and others have been looking to add memory features into their products.
The tools do however come with privacy concerns given that those memories will then be stored within the system. OpenAI said that since the feature was first being tested it has now added a new tool that will show what memories have been created and show the option to remove them.
The feature can also be turned off entirely, as well as for individual conversations.
OpenAI said that it was aware the new feature brings “additional privacy and safety considerations”. It is working to mitigate those, it said – by steering ChatGPT away from remembering sensitive information like health details, for instance.
