
Recent discussions have emerged regarding the capabilities of ChatGPT, particularly its memory function. Users have noted that if memory is not activated, ChatGPT does not retain past interactions, leading to misconceptions about its ability to recall personal information. Some users suggested a playful activity where individuals ask ChatGPT to create a portrait based on their interactions. However, experts clarified that this only works if the memory feature is enabled and contains accurate data. Without memory, ChatGPT generates responses without any personal context, effectively making up information. This has raised concerns about the understanding of how large language models (LLMs) operate in relation to user data.
This. This prompt only works if you have memories toggled on and the memories are good quality. Else ChatGPT is just making things up. https://t.co/dRzmsg6kkT
I think this only adds to confusion about what LLMs know about you. If you do not have "memory" turned on for ChatGPT, and you do this into a new chat, then the AI literally knows nothing about you & is making everything up. If there are saved memories, it will riff off of those https://t.co/nkgmJNoIlB
I'm seeing lots of people ask ChatGPT to tell them something about themselves. Reminder: ChatGPT doesn't have access to your old conversations other than (i) memory notes and (ii) custom instructions. Just type "show me your memory notes". Correct them if they're wrong. https://t.co/uOVlWvutMt





