Conversing with an AI chatbot is still in its very early stages. You have to keep telling him your preferences, and you've already done it five or six times. OpenAI is reportedly trying to fix this and personalize its bot. The company has begun rolling out "memory" for ChatGPT, which will allow the bot to remember information about you and your conversations over time.
Memory works in one of two ways. You can tell ChatGPT to remember something specific about you: my name is George, I'm allergic to coca cola.
Or ChatGPT may simply try to find these details over time, saving information about you as you ask questions and receive answers. In both cases, the goal of OpenAI's ChatGPT is to make you feel like its bot is a little more personal, smarter, and needs to be reminded of things every now and then.
Any custom GPT you use will also have its own memory. OpenAI uses Books GPT as an example:
with memory enabled, it can automatically remember which books you've already read and which genres you like the most.
Memory can be a feature that ChatGPT desperately needs, but also a minefield.
OpenAI's strategy here is very similar to how other Internet services use it to learn about you. They track you while you're using their services, learn about what you search for, click on, like, or whatever, and build your profile over time.
But this approach makes too many people feel uncomfortable! Many users are already wary that the questions and answers returned by OpenAI feed back into the system as training data to help further personalize the bot.
The idea of ChatGPT "getting to know" users can be cool but also creepy.
OpenAI says the bot's users will have control over ChatGPT's memory, and that it has trained the system not to remember sensitive things like your health information.
According to the company you can always ask ChatGPT what it knows about you and then tell it to forget something or clear the memory from the new “Memory Management” section found in the settings. OpenAI also suggests Temporary Chat as a kind of incognito mode: a way to have a quick chat without affecting ChatGPT's memories.
By default, memory will be enabled, and OpenAI says that memories will be used to train its models in the future. (Companies using ChatGPT Enterprise and Teams services will not send data to the models.)
For now, the memory is just a test, open to a “small group” of users, the company says on its blog.
Can anyone imagine how quickly the new feature can become established in the way we interact with ChatGPT? Robots are getting smarter and getting to know us very quickly.
