ChatGPT will soon have memory. The conversational robot will thus remember all the elements that you have specified to it during your various discussions to be even more efficient and more relevant.
One of the best things about ChatGPT is that it is able to later use things you told it earlier in the conversation, as if you were naturally chatting with a person. It’s very simple, he takes into account what you say to him and takes these elements into account to adapt his words. The problem is that all this information is automatically erased from the chatbot’s memory when you start a new conversation. It’s not very annoying if you use it occasionally, but it’s much more annoying when you work or have to regularly perform the same tasks with it.
Also, to counter this problem, OpenAI is testing a new function that allows ChatGPT to remember all kinds of information about you, and this constantly, as the company announces in a statement. To put it simply, with this “memory”, the AI will draw from conversation history and remember important information so that you don’t have to constantly remind it that you are allergic to peanuts or that you prefer peanuts. Meeting summaries are presented in three separate parts. Quite practical!
ChatGPT Memory: Control the AI’s “memories”
ChatGPT’s memory will improve as you work with the bot. You can also ask him directly to remember this or that instruction. For example, you can tell them that you’re a teacher and you want to create 50-minute lesson plans followed by activities. No need to remind the AI of this with each new conversation. Other examples: You told ChatGPT that you own a neighborhood coffee shop, which will help them, when brainstorming a social post celebrating a new establishment, know where to start. Likewise, if you have a toddler who loves jellyfish. When you ask ChatGPT to help you create their birthday card, they will suggest a jellyfish wearing a party hat if you ask them to create their birthday card. AI can even remember your tone, voice, and format preferences and automatically apply them to blog post projects. When you code, you can tell it your programming language and frameworks. It is able to remember these preferences for subsequent tasks, simplifying the process.
Note that this function will not be limited to ChatGPT, since GPTs, these personalized versions of ChatGPT, will also have the right to this memory. OpenAI takes the example of Books GPT, which advises a user on their next book. This function will allow it to preserve “your preferences, like your favorite genres or your best books, and adapts its recommendations accordingly.” Obviously, this information will not be shared with the GPT developer.
But this incredible memory feature raises the question of data privacy – which is why OpenAI is cautious in its deployment. If the firm allows itself to use this information to train its language model, it still offers a management panel in the personalization settings to erase all or part of it. Note that deleting a discussion will not delete the memories linked to it. In addition, certain sensitive data, such as health data, may never be stored, unless you specifically request it from ChatGPT. You can also maintain complete control by telling the AI to forget this or that information. Finally, to chat with the chatbot while putting aside its memory, a “Temporary Chat” option will appear in the drop-down menu – the one which allows you to select the desired language model. The memory feature will be offered to a reduced number of free users and paid ChatGPT Plus subscribers, which will allow OpenAI to discover possible areas for improvement.