OpenAI Introduces Memory Feature for ChatGPT, Allowing the AI to Remember Past Conversations

OpenAI is unveiling a new memory feature for its generative AI assistant, ChatGPT, to enhance conversational interactions. The update introduces long-term memory, enabling the chatbot to retain information beyond individual conversations. Users will have control over what ChatGPT remembers, and a manual setting offers granular control. The feature is available to both free users and ChatGPT Plus subscribers, currently undergoing testing with a small user group, and plans for a broader rollout will be shared soon.

This memory upgrade aims to make conversations more seamless, eliminating the need for users to repeat information in every new interaction. Users can instruct the AI to remember specific details, such as preferred essay length, writing styles, occupation, or business ownership. ChatGPT can also autonomously pick up information during conversations. OpenAI provides an example where mentioning having a toddler who loves jellyfish results in the AI suggesting a jellyfish-themed birthday card.

Addressing privacy concerns, OpenAI emphasizes that users have complete control over the feature. Memory can be turned off in Settings, and users can also engage in conversations without memory or history using the temporary chat feature. The AI allows users to command it to forget something, view and delete specific memories through the Settings menu. Additionally, users can disable data usage for improving the model.

OpenAI acknowledges potential privacy and safety concerns with the memory feature and commits to assess and mitigate biases. The AI model aims to avoid remembering sensitive information, such as health details, unless explicitly instructed by the user. OpenAI will continue refining the memory feature to enhance privacy and user control.