Dreaming: Enhanced Memory Technology for a More Useful ChatGPT

OpenAI is launching a new system that improves ChatGPT's memory synthesis capabilities to provide users with a more useful and personalized experience. This update focuses on addressing issues of memory Freshness, Continuity, and Relevance.

The existing Saved Memories approach had the limitation of requiring users to explicitly ask it to remember something, and information could become inaccurate or useless over time. To address this, the newly introduced Dreaming technology references conversation history to automatically curate and synthesize memory in the background.

The newly launched memory architecture has the following features:

• Dreaming-based architecture: Through a background process, it synthesizes learnings from multiple conversations to provide the most up-to-date and relevant context.
• Memory summary feature: Users can review a summarized form of what ChatGPT knows about them, and edit or update the information.
• Personalization optimization: It learns the user's preferences, projects, and constraints through natural conversation, maintaining context without the need for separate requests.

This update is currently being rolled out first to Plus and Pro users in the US, and is expected to expand to other countries and to Free and Go users within the coming weeks.