ChatGPT’s evolutionary leap: Continuous memory across all chats

OpenAI has announced an improvement in ChatGPT's memory, which not only enhances its ability to remember and use information already mentioned by the user but also ensures that this context is not limited to the ongoing conversation and even updates it in real-time.

The company has modified ChatGPT's memory architecture, reports EP, so that this chatbot can have reference information with each new conversation and always keep user preferences in mind, for example, if they are vegetarians or do not want a certain person mentioned, to tailor responses accordingly.

Third update of the memory function

This is the third update that improves the memory function since the feature was introduced in 2024. At that time, users had to ask ChatGPT to remember specific information, although last year the possibility was introduced for it to remember previously mentioned information in the conversation to personalize responses.

The new architecture introduces a more powerful and effective memory system, which provides continuity to the context with synthesized memories, no longer limited to the current conversation but maintained for new conversations, without having to "introduce oneself from scratch," as OpenAI explains in a statement.

It also improves the understanding of the temporal context of information, for example, if users have mentioned they will take a trip in July, ChatGPT will be able to distinguish between present, future, and past times to update the information according to the moment, making it more relevant and useful.

Additionally, it introduces a summary of the information that ChatGPT "remembers," so users can review and modify it at any time and even specify the topics the chatbot can address and when it should do so.

This change in the memory function is available for users of the Plus and Pro subscriptions in the United States. It will also be extended to users of free accounts and Go users after reaching the necessary processing capacity to scale it and do so more efficiently, although its deployment will take place in the coming weeks.