OpenAI is testing a new feature for its generative artificial intelligence (AI)-powered assistant ChatGPT that can make conversations much more fluid. The AI firm is rolling out an update this week that will add long-term memory for the chatbot, giving it the ability to retain pieces of information beyond a specific conversation. The company says that users will have full control over what ChatGPT remembers and what it does not. There is also a manual setting for more granular control of the feature.
Making the announcement on Tuesday (February 13), OpenAI said the new memory upgrade will make future conversations more helpful and save users the trouble of repeating information in every new conversation. The feature is available for both free users and ChatGPT Plus subscribers. Currently, the feature is in the testing phase and has only been rolled out to a small portion of users. The company stated that it intends to share plans for a broader rollout soon.
With this feature, users can ask the AI assistant to remember their preferred length of essays, writing styles, where they work, if they own a business and more. However, asking it to remember is unnecessary as the chatbot will also pick information up on its own during conversations. Citing an example, the company said, “You mention that you have a toddler and that she loves jellyfish. When you ask ChatGPT to help create her birthday card, it suggests a jellyfish wearing a party hat.”
One of the biggest concerns in the AI space is data privacy, and some users might be apprehensive about letting the chatbot remember personal details about their lives. OpenAI highlighted that users will hold full control of the feature. To turn memory off, they can go to Settings > Personalisation > Memory, and toggle it off. While memory is off, ChatGPT will not create or use memory. Alternatively, users can use the temporary chat feature to have a conversation without memory or history.
Users can also command the chatbot to forget something, and it will. There is also an option to view and delete a specific memory. For this, going to Settings > Personalization > Manage Memory will show all the saved memories in an itemised list and give the option to delete them, too.
The AI firm also highlighted that it may use the content that users provide to ChatGPT, including memories, to improve the AI model. But users can turn this off too by going to Settings > Data Controls > Improve the model and tapping the toggle button. OpenAI does not train on content from ChatGPT Team and Enterprise customers.
Further, the company stated that as the memory feature brings more privacy and safety concerns, it is taking steps to assess and mitigate biases and pushing away from remembering information that can be sensitive, such as health details. These will only be remembered when a user explicitly commands ChatGPT to add it to the memory.
