ChatGPT Gets A Memory – Here’s All You Need To Know About This Groundbreaking Innovation
26 February 2024
One of the frustrations with ChatGPT (and other generative AI chatbots) is that they don’t remember what we’ve said to them in previous sessions. This means starting from scratch with each new chat, teaching it about yourself, your business or anything you want it to know that it won’t find in its training data or through web browsing.
Well, it looks like all that is about to change. In what could prove to be its most significant update so far, OpenAI – the creators of ChatGPT – has just announced that it is to be given a memory.
But while I think this will prove to be hugely helpful for the millions of us who use it every day, it also raises some important concerns. How reliable will this memory be? What does it mean for data privacy? And are we ready for AI that can develop long-term memories, bringing them yet another step closer to human intelligence?
Why Does ChatGPT Need Memory?
Just as with the addition of web browsing to ChatGPT’s capabilities last year, the addition of memory is more than just a regular incremental update. It could potentially change its behavior and capabilities in many ways.
So far, when ChatGPT generates a response, all of the information it has to consider comes from three data sources: It’s training data, the user’s input during the current session, and (if using web browsing mode) the internet.
Effectively, this update adds a fourth source – a long-term memory store that’s persistent between sessions and holds information that could make its replies in future chats more valuable. This could include, for example, the user’s name, their occupation, or their personal likes and dislikes.
Having this long-term memory as an additional source of information means users won’t have to repeatedly input this data every time they start a new session.
This should mean there will be less need for lengthy, detailed prompts that have to be re-entered every session in order to ensure it gives consistent output.
The Context Window
The context window is the technical term for all of the information that ChatGPT can “see” when it creates its responses.
If you’ve ever had a long chat with it and noticed that eventually it starts to forget things you’ve already said, that’s because you’ve run out of space in the context window. GPT-4 – ChatGPT’s most powerful model – has a context window of 8,192 tokens.
OpenAI hasn’t yet said whether this new long-term memory will be loaded into the existing context window, or if the context window will be expanded, making it effectively “free” information.
Expanding the amount of information that ChatGPT can address in a conversation means it will be able to carry out longer, more complex and more detailed tasks.
How Will This Improve Generative AI?
There are a ton of ways in which giving generative AI long-term memory could improve it as a tool:
· It means it could improve its ability to learn over time, as it stores information from past interactions and uses it to inform future conversations.
· It can make its responses more personal, as it learns about the user and understands specifics about how they like to work and solve problems.
· Conversations will have improved continuity, as it remembers facts and information from previous chats without having to be reminded.
· It comes closer to being able to provide responses that display emotional intelligence, as it can build up a long-term understanding of users’ emotional responses.
· It can become better at making decisions, as it remembers how previous actions and interactions affected results and outcomes.
All of this could also help us to foster a deeper sense of connection with AIs, ultimately letting us make better decisions on matters of trust.
But What Do We Have To Worry About?
With all of the benefits, there are also a number of important concerns that have to be addressed.
The most pressing is probably around data privacy and security. Due to the nature of the information that we will want it to store, most of this new data will be personal information – things that are specific to us as users and human beings. For example, OpenAI’s press release says that the tool will avoid actively recalling sensitive data, such as your health information, unless users specifically request it to do so.
It also says it will give users granular control over what information can be retained and what information can be fed back to it in order to better train its systems. This sounds good, but we have yet to see how it will work in practice. We can only hope that other AI tools that are now rushing to add memory will be so conscientious, too.
There are also ethical concerns around deciding what an AI should remember or forget. Even if the user is in control, the tool itself may still have to make decisions when it comes to personal information that relates to other people.
Another concern, perhaps more long-term, is how it will affect the way we use technology in general. Does this bring us a step closer to AGI? Memory is a critical feature of natural human intelligence, and although ChatGPT has always had a memory of sorts – its training data could be considered as such – this takes things a whole step further by allowing it to remember things about people as individuals. Coming to terms with the full ethical and cultural implications of this will be a complex task.
In whatever way we end up addressing these issues, one certainty is that we’ll remember this as an important development in the history of AI, and now it looks like AI will remember it, too.
Related Articles
11 Most Reliable AI Content Detectors: Your Guide To Spotting Synthetic Media
Since the launch of ChatGPT just two years ago, the volume of synthetic – or fake – content online has increased exponentially.[...]
The AI-Powered Citizen Revolution: How Every Employee Is Becoming A Technology Creator
Something remarkable is happening in organizations around the world.[...]
6 Mistakes IT Teams Are Guaranteed To Make In 2025
The next wave of artificial intelligence isn't just knocking at enterprise doors - it's exposing fundamental flaws in how organizations approach technology transformation.[...]
2025’s Tech Forecast: The Consumer Innovations That Will Matter Most
Consumer technology covers all of the tech we buy to make our lives more convenient, productive or fun.[...]
7 Healthcare Trends That Will Transform Medicine In 2025
Healthcare has evolved dramatically in recent years, with technology driving countless new opportunities, just as demographic and societal factors have created new challenges.[...]
The Simple ChatGPT Trick That Will Transform Your Business AI Interactions
I believe ChatGPT and other generative AI tools can help pretty much any business.[...]
Sign up to Stay in Touch!
Bernard Marr is a world-renowned futurist, influencer and thought leader in the fields of business and technology, with a passion for using technology for the good of humanity.
He is a best-selling author of over 20 books, writes a regular column for Forbes and advises and coaches many of the world’s best-known organisations.
He has a combined following of 4 million people across his social media channels and newsletters and was ranked by LinkedIn as one of the top 5 business influencers in the world.
Bernard’s latest book is ‘Generative AI in Practice’.
Social Media