ChatGPT forgets. When using GPT-4, ChatGPT has a context length (memory limit) of 32,768 tokens, which is roughly 24,000 words or 50 pages of text. This allows it to remember details across average-length conversations or documents. As conversations get longer or more complex, it becomes clear that the model either starts to lose context or repeat itself. If you notice this early in a task, you can somewhat mitigate it by summarizing context, breaking down conversations into smaller pieces, and explicitly telling ChatGPT to remember key details. If you don't notice "memory loss" early in your task, the results can (and do) continue to degrade until you do notice. So far, this has been an unavoidable occupational hazard.
Now, ChatGPT will remember. OpenAI announced an awesome new feature that allows ChatGPT to remember personal details, preferences, and context from previous chats with a user. For example, if a user mentions their daughter Alexis who likes pink butterflies, ChatGPT can later generate a birthday card for Alexis with those details.
You can store information in Memory two different ways: either directly tell ChatGPT to remember certain facts (like coding language preferences) or have ChatGPT automatically pick up and remember relevant details from natural conversations. What's nice is that the memorized information is persistent across conversations, and users have control over what is remembered and can view, edit, delete, and even turn off Memory. ChatGPT is designed to not store sensitive information without explicit consent.
Memory allows ChatGPT to become more personalized and act like a digital assistant that evolves with the user. Potential uses include remembering formatting preferences, professional/personal context for better answers, and improved recommendations (like what books you’ve previously read). For enterprise customers, shared team memories could store preferences like company document styles.
Memory is currently being tested with a small number of ChatGPT users and will later be rolled out more broadly. When fully released, it will be available in both the free and paid tiers of ChatGPT.
ChatGPT will still have memory limitations, but this new persistent memory feature allows it to better learn and retain user details over time. This new feature gives users some new ways to manage memory and prevent excessive resource consumption, but core model limitations persist, especially as conversations scale up. The best way to understand how this may help you is to try it for yourself.
As always your thoughts and comments are both welcome and encouraged. Just reply to this email. -s
ABOUT SHELLY PALMER
Shelly Palmer is the Professor of Advanced Media in Residence at Syracuse University’s S.I. Newhouse School of Public Communications and CEO of The Palmer Group, a consulting practice that helps Fortune 500 companies with technology, media and marketing. Named LinkedIn’s “Top Voice in Technology,” he covers tech and business for Good Day New York, is a regular commentator on CNN and writes a popular daily business blog. He's a bestselling author, and the creator of the popular, free online course, Generative AI for Execs. Follow @shellypalmer or visit shellypalmer.com.
