GPT-4's New "Memory" Feature Is RELEASED! (ChatGPT Memory Update)

So the memory feature that I talked about in a video two weeks ago is actually finally here, and OpenAI have decided to slowly roll this out starting from today. If we take a look at their blog post, you can see that it is finally here. Memory and new controls for ChatGPT we're testing the ability for ChatGPT to remember things you discuss to make future chat more helpful. You're in control of ChatGPT's memory now. Why I find this so fascinating is that when I first uncovered this GPT memory from a few ChatGPT leakers, I was a little bit surprised that this could be a real feature, because of course, I was a little bit skeptical of any leaks. But now that this blog post is finally here, I can truly say that I am excited to announce that the leaks were real. So what exactly does this mean for us now that this feature is finally here? Who is going to be able to use this feature? Has it been rolled out yet and is there going to be some customization with this? There is a lot of information to go into, so let's take a look at the first piece of information, which is essentially, you're in control. So one of the first pieces of the blog is that they state you're in control. We're testing memory with ChatGPT. Remembering things you discuss across all chats saves you from having to repeat information and makes future conversations more helpful. You're in control of ChatGPT's memory. You can explicitly tell it to remember something, ask it what it remembers, and tell it to forget conversationally or through settings. You can also turn it off entirely. We're rolling out to a small portion of ChatGPT and Plus users this week to learn how useful it is, and we'll share plans for a broader rollout soon. So essentially with this, you can explicitly tell it to remember something, ask it what it remembers, and of course, turn it off entirely. But now, one of the things here that I do find really cool is how this essentially works, because as you do know, this might be a little bit confusing. So essentially, as you chat with ChatGPT, you can ask it to remember something specific or let it pick up the details itself. ChatGPT's memory will get better the more you use it, and you'll start to notice the improvements over time. And here is where they give some of the really, really cool examples. So, for example, you've explained that you prefer meeting notes to have deadlines, bullets, and action items summarized at the bottom. ChatGPT remembers this and recaps meetings this way. So if you are someone who does a lot of meetings, you say, look, I've got this meeting, can you give me some notes? It's going to format it in the way that you do. Like you've told chatgpt you own a neighborhood coffee shop. And then when brainstorming messaging for a social post celebrating a new location, ChatGPT knows where to start. And as someone who runs various businesses, this is definitely something that is very, very time saving. Because rather than stating a new chat, stating look, this is the business, this is what I'm working on. Having ChatGPT actually remember what's going on is something that is really useful. It then says here you mentioned that you have a toddler and that she loves jellyfish. When you ask ChatGPT to help create her birthday card, it suggests a jellyfish wearing a party hat. And then of course, the last example here, it says it remembers that you have 25 students and that you prefer 15 minute lessons with follow up activities. So I think this is going to be something that really takes GPT and ChatGPT to the next level because this, ladies and gentlemen, is what we were missing for quite some time. Now, if you want to turn this feature on or off, OpenAI have provided us with this small video here where it says turning memory on or off, all you literally need to do is go over to your account, click that, then click Settings, and then you'll see right here personalization. Then you can see the memory feature is going to be there. All you need to do is simply click that little button and it can be on or off. Now I'm not sure why you want to turn it on or off, but of course there might be some privacy issues. So what they've stated is that we may use content that you provide to ChatGPT and including memories to improve our models for everyone. If you'd like, you can turn this off through your data controls. And as always, we won't train on content from ChatGPT team and enterprise customers. So essentially what they're stating here is that if you're using this for personal use, they're probably going to use the memories and the content, including ChatGPT, to improve their models. Because what they do is when ChatGPT gives a response, they largely want to see if that response was good and and how they can make that better. But one thing that they won't do is they won't train on content from TPT teams and enterprise. That's because they do work with companies and These big companies proprietary data is something that does need to stay really, really private due to many different laws. So one thing that you do need to know is you can actually turn this off through your data controls. So as someone who is always actively engaged in what's going on with their privacy, that is something that you can manage. Now, for temporary chats, what you want to do here is this is bas like an incognito tab on Google. So if you want to have a chat which doesn't remember anything, you just want to ask something quickly, you can just put on the temporary chat and you can see that this chat won't appear in history, use or create memories. And for safety purposes, we may keep a copy for up to 30 days. And I think something like this is rather good because sometimes what you do want is you do want a rather quick question. And for those of you who know that using ChatGPT is obviously part of your daily routine, so something that happens is on the left hand side here. I have so many different chats that it can become a bit confusing to sift through them all. So having a temporary chat option is absolutely great, because if you have, you know, one off questions that you know aren't going to be part of a broader discussion, you're then going to be allowed to use this temporary chat feature to just have those one off questions that are going to be not that important over time. So I'm really glad that they actually finally added this feature because it was something that we do need. Now there was also this custom instructions allow ChatGPT to be more helpful. So custom instructions continue to allow you to provide ChatGPT with direct guidance on what you'd like it to know about you and how you'd like it to respond. And essentially they've added this in the blog post because I do think that the custom instructions bit has been forgotten. If you don't remember what custom instructions were, custom instructions were something that essentially when you're talking to ChatGPT, you can provide it a little bit of context. Now, this is quite similar to memory, and I think it's just a little bit different in the sense that ChatGPT over time is just going to pick up on certain things instead of you having to manually enter them yourself. So it will be interesting to see the differences between custom Instructions and how ChatGPT responds, because I think what ChatGPT and the OpenAI team want, which they previously said and even said in new interviews, is that they do want a more personalized AI system because that is the future that we are moving towards. So this feature doesn't surprise me. So I guess what you could do is, if you're waiting for the memory feature to be there, you could take advantage of the custom instructions and ensure that, you know, whilst you do have this feature, you do use it. And it's very easy to use custom instructions. Just go into your settings and just make sure it's enabled. Now, something that they're actively doing is they're evolving the privacy and safety standards. So they said that memory brings additional privacy and, and safety considerations such as what type of information should be remembered and how it's used. We're taking steps to assess and mitigate biases and steer ChatGPT away from proactively remembering sensitive information like your health details, unless you explicitly ask it to. So it will be also interesting to see how ChatGPT chooses what it's going to remember and what it chooses to forget. There are many different things that it could choose to remember and what it chooses to forget, but they've said that that one of the things that it doesn't want to do is sensitive information like your health details. Now, one thing I do want to urge you guys not to do is don't enter sensitive information into ChatGPT because they've repeatedly stated that this is a system in which they are training on in terms of using that to improve the model. So people from time to time will be looking at certain conversations. So don't be putting sensitive information in there because there is always the risk that a human reviewer, someone reviewing the certain conversations, could potentially see it. And that was something that we did see with Gemini, because of course, with the AI competition and the AI race, what people do want to do is they want to improve the models and they can do that by looking at the conversation. So don't input anything like your bank details or anything absolutely crazy like that, or any, you know, something that you wouldn't want ChatGPT to remember. Now, they also actually do talk about teams and essentially teams and enterprise is going to be a little bit different. They said for, for team users and enterprise users, memory can be useful. When using ChatGPT for work. It can learn your style preferences and build upon past interactions. This saves you time and leads to more relevant and insightful responses. For example, ChatGPT can remember your tone of voice, format risk preferences and automatically apply them to blog posts without needing repetition. And this is one of the things that I think most people did miss about this announcement. It says here that when coding, you tell ChatGPT your programming language and frameworks and, and it can remember these preferences for subsequent tasks, streamlining the processes. So I would want to see if this is something that does work because I've worked a little bit with code in ChatGPT and sometimes you do have to keep telling it the same thing and it is genuinely one of the most frustrating things because you're trying to save time, but you have to keep telling the AI system what to do. So I will be trying to run some tests on this to see if the memory feature is actually good. And then of course they stated for monthly business reviews, you can securely upload your data to ChatGPT and it creates your preferred charts with three takeaways each. And it also does say memories and any information on your workspace are excluded from training. On our models, users have control on how their memories are stored and used in chats and of course enterprise account owners can turn off their memory or their organization at any time. So this is going to be pretty good for those of you who are working in teams and you want to be able to work with your team very effectively and you don't want to have to tell each team member to format it in this way. So this is going to be something that saves more time. Now there was one thing that I did actually want to take a look at and this is of course the GPTs. So it does state here that GPTs will also have memories. GPTs will also have their own distinct memory and builders will have the option to enable memory for their GPTs. Like your chats, memories are not shared with builders. To interact with a memory enabled GPT, you will also need to have memory on. So for example, the Books GPT helps you find your next read and with memory enabled, it remembers your preferences such as your favorite genres or top books and Taylor's recommendations accordingly without needing repeated inputs. So I think the very best thing here is that of course that the memories are not shared with the builders because that would be a huge privacy issue. And of course I'm really glad that we do have memories on the GPTs, because this is going to mean that each time we start a chat with GPT, it's going to remember our previous interactions. And this makes a lot of sense because if you're interacting with the GPT for whatever reason, it's it means that you're likely trying to save time and get a shortcut to an issue. And with this memory feature, this is going to be really good. Now what's Also cool is that each GPT does have its own memory, so you might need to repeat details you previously shared with ChatGPT versus a GPT. So it says if you're using the artful greeting card GPT to create a birthday card for your daughter, it won't know her age or that she loves jellyfish. You'll need to tell it the relevant details you now one question I do have, and this is because there was a recent feature which we know is going to be coming, and I know it was rolled out to just a few people, is that if we do get memory for different GPTs and those are independent of each other, what happens when we combine GPTs in a chat with ChatGPT? Will it remember certain details? Either way, it's something that I will have to test because things are slowly evolving. But I do think that this does show us the slow iteration of GPT4 into a more personalized AI model shows us that potentially GPT5 is going to be even more so personalized. Now one thing that is the big question is that if you don't have access to this now, when are you going to get access? So it says memories for GPTs will be available when we roll it out more broadly. That is of course very specifically vague and they says we're rolling it out to a small portion of ChatGPT 5 free and plus users this week to learn how useful it is and we'll share plans for a broader rollout soon. And of course if you're wondering what kind of time frame you can expect, I'm kind of predicting that since they've been working on this for quite some time, I wouldn't expect it to be too much longer, especially since they've done a blog post. So I do expect it to be likely within the next two to four weeks. And in addition, if we are talking about any kind of rollout, I would expect it to largely be USA based first and then of course to the rest of Europe now something that most people did miss again because because with a lot of these updates the problem is that people just look at the update and don't actually tie them into a lot of other stuff happening in the AI community. And that's what many people did miss with Copilot. So if you don't know, Copilot is something that is really, really cool. But Copilot actually today got this feature where personalization is there. So you can see personalization is currently on. You can of course then turn it off or undo and keep it on, and Copilot is a really, really good tool. I'm going to be making a video on the entirety of that because it is really, really outstanding. But that goes to show that some of the leaks are actually real, and some of them do manage to come to fruition based on certain things that you can see within the code. So GPT4's new memory feature is, of course, finally here. Now, my question to you all Is this something that you guys do want? Are you guys just itching for GPT5? And of course, are you going to be turning this on or off? If I do have this feature, I definitely will be using this because I do want to save as much time as But I'm excited for the future because this is a step in the right direction.