How to Deal with ChatGPT GPT-4 Message Cap
00:01 All right, so somebody asked me how do you deal with the fact that GPT-4 has this cap of 25 messages every three hours?
00:10 If you're really getting into the groove and you're doing some good work and this has happened to me multiple times and then you hit that cap like, ah, it's so frustrating.
00:23 So what are some things that we can do to avoid that? Or better work within those. Within those parameters. So a couple of things.
00:36 So I came up with three, three ways to deal with it plus a bonus one. The first is to try to use more complex prompts to get better output the first time to cut down on the number of messages.
00:46 So we know that GPT-4, one of the great things about GPT-4 is that it can handle really complex prompts. Okay.
00:57 So whether that's a really well structured prompt that prompt template that you, you reuse over and over or whether that's just a, a blob of text that you, you recorded on or that you transcribed on your phone that, that you then pasted in and, and use it or you just literally did that
01:20 in the chat, GPT app on your phone. Whatever, whatever that is, GPT-4 can handle really complex prompts. So my advice to you is to get good in this, this takes practice and obviously everything that we covered in this, in this program, but get really good at writing those complex
01:45 prompts. And remember, prompt structure doesn't matter. The, the only thing that matters around prompt structure is when you're trying to reuse it as a human.
01:54 Okay. So if I'm going to give you a prompt template, it needs to be structured in a way that you can understand it.
02:00 And replace it. So you can basically, you know, variables where you need to. But in terms of the underlying language model, it doesn't really care.
02:06 Okay. So that's something that we, we can we, we can leverage that by, by making those, those prompts as complexes as, as we need them to be.
02:17 And remember, the only thing, what that means really is just make sure you're very explicit about what goes into it.
02:30 The output. And you're telling it, even, I would say, I'm well, if, if not more important, certainly of equal weight is what you don't want to see.
02:39 So make sure you tell it, hey, this is what I want the output to be. And then, make sure you don't do these things.
02:44 Okay. So if you put that together in, in a prompt, that's going to cut down on the, on the number of refinement prompts that you have to do.
02:52 So, which is going to cut down on the overall number of messages. I think there is a natural, kind of, cut off of, of what a, of what you can do in one prompt.
03:05 That's probably just literally going to be your own sort of ideation process. There's going to be things that you didn't think about until you see the output and you're going to have to go through that refinement process.
03:17 But I would also say this, and I, I also fall into this trap. Do you need to do a refinement prompt or can you just take the output and, and do it yourself?
03:26 I've literally found myself writing prompts to do something that is a, that like I could have done just as quickly on my own.
03:35 So make sure you're not using chat, GPT as you know, like for everything. Like there's, some things you just don't need it to do.
03:45 I don't need it to you know, change a few words here and there or something. And I, you know, that's just going to cost you messages.
03:53 If you're not going to be doing a lot of work, then fine. I mean, it doesn't hurt the model. It doesn't hurt anything to do that.
04:00 Each one of those messages where you write a prompt and then you do a refinement prompt, that's going to count against your, your allotment.
04:07 So that's one thing to think about. The other is using Bing chat directly outside of chat, GPT. What I mean by that is if you go to chat, GPT for the model and you see the browse with Bing.
04:22 That's GPT for. So that means anything you do in there is going to count against your messages. So you know, I've changed to that model and we're still at a cap of 25 messages.
04:36 So I, I want to use Bing chat for ideation for research for some brainstorming. It's not as, it's not as good when it comes to brainstorming, but it, it can do some of the heavy lifting for us.
04:53 And I want to bring that into, I want to take that from, from being and bring it into, to, to chat.
05:00 And then I can start to work from there. So there's no sense in sort of wasting your credits by doing stuff in chat, she PT that you could be doing in being you know, being has live search result.
05:16 I mean, you, you can do it here. Although I will tell you also just from a practicality standpoint, regardless of message cap, the search with being in chat, you PT.
05:31 Just sometimes it doesn't work. Like it tends to throw a lot of errors. I really, I really have not been super happy with it.
05:43 So I still, so I've just, I go to being chat like directly. That also has the side effect of not counting against your credits.
05:54 Next one is drop to GPT 35 for less creative work. So when you hit that that threshold and it'll throw an error, I didn't take a screenshot of that, but it'll say you've run out of credits come back one year.
06:08 No, come back in like three hours or whatever. It'll tell you the time. And for me, I usually hit that, you know, about 1 30 a.m.
06:18 And it's like, come back at four a.m. And I'm like, man, probably not. But that's a whole nother, a whole nother discussion.
06:29 So in that error that it throws it will, it will tell you you have the option to just drop back to three five.
06:37 There's nothing wrong with GPT 35. It cannot handle the more complex prompts. So you have to break things up. It's not as creative.
06:52 So if you're doing less creative work it's fine. You know, if, if, if you have, if you have some refinement prompts that you could, you could run.
07:03 I think it's going to be just fine there. So don't be afraid. I guess my point is, and, and I, I, again, I will, I will say I kind of didn't want to.
07:14 You know, I kind of didn't want to drop back to three five because four is so powerful. But when I did, it's like, oh, okay.
07:23 That's, that's not bad. It's, it's actually pretty good. And it's super fast. So it's, it's, it's, it's fine. So, I want to work in GPT four as much as I can.
07:34 But if I'm trying to save credits, if I have some things that aren't super heavy lifting, just some, some more repetitive tasks, I can definitely drop to three five.
07:43 The other thing is, and so those are my, those are my three that, that I do. This is how I handle working around the message cap.
07:53 And honestly, I didn't even think about that being something that I was doing until somebody asked me how to do that.
08:01 And I was like, okay, interesting. Like that, that's the thing that we probably need to talk about. The last thing is I've heard that since you can share chats, that some people will buy a second or third.
08:15 I've never heard more than third, but, you know, subscription to GPT plus. And do it in a different browser. I suppose you could do it in Cognito mode, but a lot of times I'm not sure how what they're doing, but, you know, there's, there's certain fingerprinting that they can do to, to keep
08:33 you from doing this in the same browser. So you might have you know, Chrome and and edge set up and that would be probably good enough.
08:43 Yeah, you're to your two different subscriptions, but because you can come in here. So let's go to this webinar title suggestions chat and I can say share this and I'm not sure why it's.
09:01 Now there we go. Okay. So I can go to share and I can copy the link and, and you can actually share this chat.
09:10 Now it says messages you send after creating your link won't be shared. So I'm still, I'm, this is like still an interesting model.
09:20 It's, I don't know that it's. As collaborative as we would like, but if I'm just doing this myself, I could just take the chat that I've been working on.
09:31 Share it to my other account. Go to my other account in my other browser and continue. And just kind of keep doing that and now I have a workflow set up that, that would work.
09:43 This is a total hack, a workaround, but that's definitely a possibility. So again, that's sort of a bonus one. I haven't done that because all of this works just fine.
09:58 Getting better at prompt engineering, making sure that your initial prompt does what it needs to do to get you. So, the output that, you know, it's close to the output that you're looking for initially will cut down on the refinement process.
10:11 That's going to cut down on the number of messages. And just being smart about, you know, using Bing chat directly and being okay with dropping to GPT-35.
10:22 Because it's really good. It's just that GPT-4 is like so much better, but GPT-35 is, is really good. And then, you know, if you want to go this route I will say, I, you know, again, I don't know, but if they caught you doing this, do, do they, do you get in trouble?
10:41 I, I wouldn't think so because you're paying them, but you never know. It might be a violation in terms of service.
10:47 I don't know. They're probably violating our copyrights. It's, we're all just in violation of everything. All right. There you go.
10:57 Three ways plus a bonus on how to deal with GPT-4 message camp. Okay.