Google Gemini Gets a Memory Upgrade

Gemini is introducing a “reference past chats” feature to remember context across conversations. Initially, the memory upgrade will be available only to English-language Advanced subscribers via the Google One AI Premium Plan, with plans for expanded access in the coming weeks.

Rival ChatGPT has had memory capability since 2023, but Gemini’s new feature appears to be more comprehensive. It offers users greater control, allowing them to see memorized conversations, review and delete entries, manage the duration of stored conversations, and toggle specific memories on or off.

“Whether you’re asking a question about something you’ve already discussed or asking Gemini to summarize a previous conversation, Gemini now uses information from relevant chats to craft a response,” explains Google’s Dave Citron.

In late 2024, Gemini introduced a narrower memory feature called “Saved Info.” This feature allows users to create a personalized knowledge base within Gemini by saving specific information like dietary restrictions, favorite book genres, or key details about their projects. However, the new “reference past chats” feature eliminates the need for manual saving by automatically learning from conversations to generate more relevant, personalized responses.

ChatGPT’s memory works more passively, remembering details across conversations without letting users directly view or edit this information. Users can only request ChatGPT to update, forget, or not retain certain details while chatting with it. It’s also worth noting that unlike Gemini’s, ChatGPT’s memory feature is available for free.

While Google has upgraded Gemini with a faster and more efficient model and continuously improves its performance with new features, its chatbot still struggles to beat ChatGPT’s first-mover advantage.

Following lawsuits targeting its search business and the emergence of chatbots capable of performing real-time web searches, Google’s parent, Alphabet, is doubling down on AI. Its investments span everything from foundational research (Google AI, DeepMind), and advanced model development (LaMDA, PaLM, Imagen), to product integration (Search, Cloud, Android, Workspace), and AI infrastructure (TPUs, data centers).