Selective learning of information for the generation of personalized responses by a generative response engine
Topics: LLMO / GEO, OpenAI / ChatGPT, Personalization, User Signals
This OpenAI patent describes a technology that enables a generative response engine (like a chatbot powered by a large language model) to automatically learn and remember personal information about a user across multiple conversations. Rather than requiring users to explicitly instruct the system to remember details, the engine selectively identifies facts, preferences, and context from natural conversations and stores them in a “personalization notepad.” The stored information is then used to generate personalized responses in future interactions, and users can also request the system to forget specific learned information.
