OpenAI Tests Memory Controls: ChatGPT Learns and Forgets at User’s Command

A new “memory” function for its AI-powered ChatGPT is being tested by OpenAI. The ChatGPT’s memory settings enable the chatbot to retain or lose important instructions and information shared by the user. This would probably assist the users in personalizing the discourse and removing the need for information repetition. Although still in its testing phase, the new function has begun to roll out for a select group of users from both the free and Plus levels.

What are ChatGPT memory controls?

Users can teach the chatbot to remember particular data or provide directions that it must follow when carrying out specific activities in the future while utilizing ChatGPT for conversation or help. When this function is used, ChatGPT will automatically learn details and make improvements over time. For instance, if you ask ChatGPT to summarize texts in a few words and include bullet points at the conclusion, it will do so each time you ask it to in the future.

In a similar vein, you can direct ChatGPT to forget a particular instruction or detail. As an alternative, OpenAI would permit the removal of particular memory within the newly created Manage Memory section of the Settings Personalization menu. From the same option, users can also entirely turn off the memory feature.

It should be mentioned that ChatGPT stores memories across conversations and that clearing a chat does not remove the memories it has accumulated. A temporary chat can be accessed through a drop-down option at the top of the screen if the user wants to talk without utilizing memories and does not want to disable the feature or wipe memories.

Memory on custom GPTs

In the upcoming weeks, OpenAI is probably going to release memory controls for customized GPTs. The company stated that GPT builders will be able to activate memory on each individual GPT when it becomes available. The user will need to repeat the instructions they previously gave to ChatGPT or another GPT because each GPT will have its own memory.

Privacy

According to OpenAI, there are extra safety and privacy concerns with the memory capability. The company claims that ChatGPT won’t retain private information, such as specifics of a user’s medical history. The user can, however, specifically instruct the chatbot to keep such information in mind.

Although the corporation stated that it might use user-provided content to enhance its model, Data Controls also offers the ability to turn off data collecting.
Regarding memory with GPTs, the business has stated that the builder will not receive access to the information gathered via memory on a per-chatbot basis.