ChatGPT becomes more personal: OpenAI tests memory function for more individual interaction

In a recent development, OpenAI has decided to implant a digital memory into its well-known chatbot, ChatGPT. This feature, currently still in the testing phase, enables the software to retain specific information about users. For instance, the AI could remember the name of a user's daughter or their fondness for jellyfish. In practical use, this means that ChatGPT could creatively embellish a birthday card with a drawing of a jellyfish in a party hat if asked for suggestions for the user's child. To be able to use these personalizing features of the chatbot, users must explicitly request it. Additionally, there is the option to directly ask the AI which data it has stored and to delete this information – either in part or in whole. This memory capability is designed to make the bot more user-friendly and to provide added value in personalized interaction. For users who wish to have an anonymous conversation, there are temporary chats available, the contents of which are not used for further training of the AI. Since its introduction, ChatGPT has caused sensational reactions – from euphoric visions of the future to dystopian concerns. AI chatbots like this are programmed through extensive training with massive amounts of data to generate texts at a level similar to humans. The process is based on a word-by-word estimation of the next part of the sentence. A challenge, however, is that despite correct informational foundations, the occasional incorrect answer can still be generated. The introduction of the memory function marks another step in the evolution of artificial intelligence and once again highlights the potential and challenges of this technology.