Memory Prompts in Current Conversations

How can we ensure that the GPT is able to recall infomration processed in the current conversation. For instance, after a back and forth on a topic I get the GPT to understand and see my side to the argument, with supportive proof, not just false information. Then, a few minutes later I go on a tangent that involves the new information, however the GPT responds with it’s originally known information rather than taking the new side into considersation on the conversation.