OpenAI Is Trying To Give ChatGPT A Memory
On February 13, OpenAI announced that it's trying to add a feature in ChatGPT that will allow it to remember things about the user from their chats so that future conversations are simpler and the user doesn't have to repeat the same things again and again.
The same announcement was also made over X (formerly Twitter) with a video walkthrough of where to find the feature on the app and how to use it.
We're testing ChatGPT's ability to remember things you discuss to make future chats more helpful. This feature is being rolled out to a small portion of Free and Plus users, and it's easy to turn on or off.OpenAIOnce they are sure that the feature is actually useful and ready for use, they'll announce their plans for a wider roll-out.
This new feature won't be limited to ChatGPT alone. All the other GPTs such as Books GPT and Artful Greeting Card GPT can also work with memory. To avail of this feature, you'll have to manually enable the memory option.
Here it's important to note that deleting a chat history doesn't erase its memories of the chat. You have to explicitly go to Manage Memory to erase it.It's important to note that each GPT will have a distinct memory. So for instance, if you shared your book preferences with ChatGPT, you might have to do it again with Book GPT.
The best part is that the users will have complete control over the bot's memory. They can directly ask it to remember something with a prompt along the lines of Remember I'm allergic to shrimps" and can also make it forget information through the app's settings.
To view and delete memories, go to Settings > Personalization > Manage Memory. Here you'll get a list of the memories it has from your chats and an option to do whatever you want with them
If you don't want any hassle, you can use their Temporary Chat tool. These chats won't show up in chat history or memory. What's more, they won't even use it to train their AI models.
Is It Safe?The biggest concern with an AI being able to remember things about you is how safe it is. Addressing this issue, OpenAI has confirmed that they are working on evolving their security and privacy policies to help maintain privacy despite this new update.
Adding this memory feature certainly complicates drafting privacy rules. Mostly because what type of information should be stored and what shouldn't be is quite debatable.
In response to this concern, OpenAI said that they are trying to improve their AI model so that it doesn't automatically retain sensitive information unless the user explicitly asks it to.
For now, users can choose what type of information they want the chatbot to remember through the Custom Instructions setting.
OpenAI has also shared that they might use things you share with the AI and prompt you to ask it to remember things to train the model. But if you're not comfortable with that, you can easily turn it off in the settings.
The post OpenAI Is Trying To Give ChatGPT A Memory appeared first on The Tech Report.