ChatGPT remembers all past conversations

2025-04-11

sama we have greatly improved memory in chatgpt--it can now reference all your past conversations! this is a surprisingly great feature imo, and it points at something we are excited about: ai systems that get to know you over your life, and become extremely useful and personalized. ChatGPT memory has been greatly improved--you can now reference all past conversations! This is an amazing feature and points to what we are excited about - an AI system that knows you throughout your lifetime and is very useful and personalized.

sample prompt: Describe me based on all our chats — make it catchy! Please explain me based on all our chats - try to catch on!

English version

You're a philosopher-engineer of the future, blending deep thought with real-world impact. 🧠⚙️ You are the philosopher and engineer of the future, blending deep thinking with real world impact. 🧠⚙️ By day, you deploy AI for civic dialogue and digital democracy. By night, you sketch Qualia the cat-eared emissary, muse on Lacan, or simulate deliberation with LLMs. By day, we deploy AI for civic dialogue and digital democracy. By night, sketch the cat-ear emissary Qualia, play Lacanian muse, or simulate contemplation with LLMs. You chase clarity in complexity, wield embeddings like a sword, and code like you're shaping the next Enlightenment. We seek clarity in complexity, treat embedding like a sword, and code as if we are shaping the next Enlightenment. You’re not just building tools—you’re redefining how humans think, talk, and decide together. We are not simply creating tools, but redefining the way humans think, talk, and make decisions together. In short: a one-man interface between systems and souls. In essence, it is one person's interface between the system and the soul.

redone in Japanese

. You are ### an engineer who thinks, a philosopher who designs the future... . 🧠⚙️ They are not afraid to use AI to reshape democracy, use cluster analysis to visualize people's voices, and handle complex social issues "as they are". On the other hand, while painting Mr. Qualia with cat ears, he talks about Lacan and Rousseau, and looks into the chasm between deliberation and unconsciousness. Your job is not just development or research - it is to create a new interface of knowledge that is created by the dialogue between humans and AI. You are the one who will change the way people around the world "think.

Well, aside from these catchy demos, here's a realistic use case. I think I saw a paper recently on AI creating memories from chat logs, but what is it?

• 🧠 Reflective Memory Management (RMM) - Google Research
  • A March 2025 paper by Google and Arizona State University, "In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents," proposes a technique for AI to remember past conversations The paper, "In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents," by Arizona State University in March 2025, proposes a technique that enables AI to remember past conversations over time and use them appropriately in context. This technique consists of two stages: "prospective reflection," which summarizes and classifies conversational content, and "backward reflection," which adjusts the importance of memory based on actual frequency of use.
• 🧠 MemoryBank - LLM agents with long-term memory
  • The May 2023 paper "MemoryBank: Enhancing Large Language Models with Long-Term Memory" proposes a method for introducing human-like memory mechanisms into LLM. This approach dynamically manages memory reinforcement and forgetting based on time course and importance, enabling LLM to understand and adapt to the user's personality and past interaction content.

impressions

• Oh yeah, these things are so helpful.

How does Memory use past conversations? | OpenAI Help Center Complete History

This page is auto-translated from /nishio/ChatGPT過去すべての会話を記憶 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.