Published Apr 26, 2024

EP 259: ChatGPT's New Memory Feature - the good and the bad

Explore the transformative potential and privacy concerns of ChatGPT's new memory feature with Jordan Wilson, as he delves into its impact on tailored user interactions, the process of enabling the feature, and how users can manage their data for optimal control and privacy.
Episode Highlights
Everyday AI Podcast – An AI and ChatGPT Podcast logo

Popular Clips

Episode Highlights

  • Enabling Memory

    Jordan Wilson explains the steps to enable ChatGPT's new memory feature. Users need to access the settings and look for a personalization tab where they can toggle the memory feature on or off. This feature is currently available primarily for paid users, and a pop-up notification will guide users through the process 1.

    You can always come back and ask me and I'll answer. All right, so here's what you need to do. Well, number one, you need to enable it in the settings.

    ---

    He also notes that the feature was initially previewed by OpenAI two months ago and is now rolling out gradually 2.

       

    Functionality

    The new memory feature aims to improve ChatGPT's ability to remember information across different chats. Previously, each chat was isolated, but now ChatGPT can carry over details and preferences, making interactions more seamless 3. This change addresses a common complaint about large language models forgetting previous interactions.

    ChatGPT will carry what it learns between chats, allowing it to provide more relevant responses.

    ---

    Jordan emphasizes that while the feature is promising, it may still have some limitations and bugs to iron out 4.

Related Episodes