One of the biggest downsides to talking to an AI chatbot is that everything disappears after the conversation is over. The bot doesn’t remember who you are or what you asked about before. This is done for privacy reasons, but it also slightly hinders the technology from becoming a true digital assistant who knows you well enough to help with things. OpenAI is trying to fix this problem and is finally adding a memory feature to ChatGPT.
This will allow the bot to remember important personal data from previous conversations and apply that context to current queries so that its responses are more relevant.
Here’s how it works. You can ask ChatGPT to remember something specific, such as whether your child has a peanut allergy or how you sign emails. The system will automatically save this data at your request and apply it in future conversations and tasks. In addition, the system will accumulate information over time. It will naturally store data when you run queries. The goal is to make the chatbot smarter and, more importantly, tailored to your specific needs.
Also, each custom GPT will have its unique memory. OpenAI cites Books GPT as an example, as the bot will remember which books you’ve already read and which genres you like. The GPT Store is already full of unique chatbots that would benefit from this memory feature.
It doesn’t seem too different from how the internet already works. Applications store personal data, just like websites and social networks. This data is then used to create a unique user profile, which is used to develop personalized algorithms. However, there are significant privacy concerns, so OpenAI states that users will have control over ChatGPT’s memory and that the system has been trained not to automatically remember certain sensitive topics.
The company says you can just tell the bot to forget something and it will. There is also a “Memory Management” tab in the settings for finer settings. And if you don’t need this feature at all, you can turn it off completely. In addition, there is a temporary chat option for conversations that do not need to be memorized. These chats do not appear in history, are not stored in memory, and are not used to train models.
The service is currently in beta and will launch this week to a “small number” of ChatGPT free and Plus users, but the company will later share plans for a wider launch.
Read also:
- ChatGPT has been integrated into the robotic pet Loona
- Apple is testing an improved version of Siri using ChatGPT