5 Prompts That 99% of GPT-4o Users Don‘t Know

So ChatGPT has 100 million monthly active users. Wait, hold up. It has 180 million monthly active users, huh? And let me tell you from experience, most of these users don't really know what they're doing. They scratch the surface, they find a technique here and there that works for them. But there are so many non obvious use cases inside of this beautiful little interface that I just had to create this video covering various techniques that took me some time to discover. But once I did, it turns out that I actually use every single one of these techniques techniques regularly. So let's get things started with this simple little prompt that we shared in a guide in our AI learning community. And I really, really wanted to share this with you because I keep returning to this regularly. What this does is simple. It finds you case studies for a new topic that you're looking into. So I don't know about you, but I find myself venturing into new topics, new areas very frequently. I like to be on the bleeding edge of things and there's many ways to kind of manage that and to get better at doing that. But this prompt right here is legitimately one of the best ones because what ChatGPT has at its disposal, so pretty much all of the Internet, including a lot of comment sections, articles and even YouTube video transcripts. And then all you need to do whenever you're venturing into a new topic, you just go in, you paste this prompt, and then you replace these brackets with whatever topic you're interested in. So I'll just go ahead and say community building and it will tell me about community building and its application in the business world. And as it turns out in the business world, the stakes are high. Therefore a lot of ideas either come to fruition or crash and burn. And ChatGPT will tell you exactly about that on pretty much any topic that you give it. Again, this is a fantastic first step into something new. Because if I'm interested in community building, then learning about the Salesforce and how they build a customer centric community, or about Lego and how they spun up the Legos ideas platform platform where fans can submit their own Lego set ideas. Successful submissions are turned into official Lego sets. This is something I didn't know about, but what a great way to explore unknown territory to me this is. And it's simple as that. Look at that. It even points out Airbnb that has a global hosting community and and the different initiatives they have and that worked for them and the outcomes that they achieved. It's a relatively simple prompt, but One that I find myself using regularly, so I hope this will help you too. And now let's move on to the next one. Okay, and just disclaimer for this next one. A lot of you power users will already know about this prompt, but for you guys, I'll show you a hidden chatgpt feature that I wasn't aware of myself until a few days ago that works together super well with this prompt. Which prompt am I talking about? It's this. Let's just use one of the presets to get started. Because this is a follow up prompt, it's not meant as the first prompt prompt in your conversation. This one is meant to continue your conversation. And what I'm talking about here is this. But first, ask me clarifying questions that will help you complete your task. And look, without overcomplicating this, the whole point of prompt engineering, aka interacting with ChatGPT and getting results that are useful to you, is playing with context. That's what a lot of it comes down to. Giving it basic instructions like this is very simple, almost anybody can do that. But then fleshing out the context in a way that it includes all the nitty gritty little details that an assistant would need to plan a relaxing day for you, that part is not so obvious. And that's where this prompt comes in. It asks you the questions a helpful assistant would be asking if the goal is to plan the best day possible. So here's all the questions you can answer now. For this original prompt to improve in quality and relevance to you, I could simply go through and answer them like so or and here's the hidden feature that I promised you a second ago. If you highlight one of these, there's this new button that appears. It's called reply. And this is not something they tweeted about or wrote a blog post about. They they just added it to the user interface. And this allows you to interact with specific parts of the conversation. And it only pulls this into the context for your prompt. So I could answer the question like so without needlessly extending my conversation. This is handy in many, many other situations, but as you can see, it singles out this question and it's just a great way to reply to something specific. Then it goes more in depth on that, because it does always consider the entire conversation above, but it allows you to formulate shorter answers twice to only the specific text that you're replying to. So it's essentially a way to save time when formulating your follow up prompts. Okay then, if I answered all these questions, it's going to give me a higher quality answer. And you can include this in pretty much any workflow you do. Now here's my tip. If you actually go ahead and answer these different questions, then this is the information you might want to re implement into your original prompt. In other words, if you plan on reusing this prompt, and if this is a problem you encounter regularly, then you might want to include the answers to these questions in the original prompt, like so. And this allows me to avoid this question next time I use this prompt because I did the work once and now I can be super efficient every single time I'm looking to plan a relaxing day. All right, next up, we're going to be Talking about building GPTs in the simplest way possible. And let me tell you, this took me a while to figure out. I've built so many GPTs in different styles, with different building blocks, writing techniques and components to them. There's a lot you can do there. But at the end of the day, I arrived at this, this, this GPT generator prompt that you can simply copy paste from a link in the description because it's just too long to paste into the description. And all you do is you copy this one prompt, you go into a new conversation with ChatGPT 4 or 4.0, you paste it, and then you're on your way to building a GPT that will help you complete specific tasks. As a quick recap, they're just a specialized version of ChatGPT for one task or one purpose, and all you have to do is to build that specialized assistant with this prompt that I provided you here. As you just need to answer this question. What, what is your chatbot's intended purpose or goal? I'll just say brainstorming Assistant for generative AI tutorials. Let's say that is the GPT that I'm looking to create today. And then it goes ahead. And as you can see, I implemented the last technique, the last prompt that I showed you in here as it asks me various clarifying questions. But I also introduced another feature for just anybody who wants to get a draft as soon as possible and doesn't want to bother with all of these clarifying questions. That will certainly help to make it better. But if you just don't want to bother, you can just go ahead and say auto and it's going to answer all of these clarifications clarifying questions for you. We'll just wait a bit till it finishes generating. And as you can see, this is pretty much it you want to start copying from communication sequence and end at the end of this bullet point list. So if I simply copy this like so then I go to explore GPTs and then here you simply click create and under the configure tab you can post the instructions in here and that's it. Simple as that. Okay, sure, we can give it a name, Brainstorming Buddy. We can give it a conversation starter like let's start. I always like to do this just so the user has something to click, even if I'm the only user. Makes the experience a bit better. And then we won't worry about anything else in here because this builder is all about the instructions. So I'll say create, update this and my brainstorming buddy that is personalized to my needs is live. So let's give it a spin. Let's just say let's start and let's see what the builder came up with by itself. Right? Zero customization, zero additions. This is what it came up with. Welcome to the Gen AI tutorial Brainstormer. How can I assist you today? Are there specific needs or challenges you're facing in creating your tutorials? I would like to create a tutorial on GPT. So now it's going to continue helping me as it has been set up in the instructions for you. Now look, this is a very potent starting point, yet it's still just a starting point. Can you use it like this? Absolutely. It works. Look, it's helping me. It's personalized, but really the power of these GPTs that are meant for a specific use case is tuning them a little bit. So if I go back into the builder and I full screen this, I'll see the welcome message right here. I could customize that. Here are some of the instructions that the builder came up with for. For me, I could also customize some of these. But given this does take a little bit of prompting experience and once you get into the knowledge base, that works really well with templates or up to 15 pages of documents. That's what I usually say. After that it kind of starts falling apart. As of now there things get a little more complicated. And with actions, you do need a little bit of technical skill. You need to host the API endpoint included. That is not intuitive at all. But here's the thing, a lot of people want to learn this skill and I actually went ahead and built the entire AI Advantage community around the fact that skills like this are best taught in a communal environment. So if you're interested in taking the Next step in building GPTs, then. Yeah, in the community that is paid, we have a learning path that gives you an intro video by me introducing you to what modules you can expect in here. And in five steps, in five roughly one hour long lectures, I will teach you how to build a GPT module by module. This GPT generator that I showed you is step number two. And here you have a full lecture on these various topics, how to generate a GPT, generating prompts for them, and a Q and a session in the end. And if you ever get stuck along the learning path, there's a full community of people on the same path as you there to help you. This is why learning in an environment like this really is best. Look, in the last 24 hours, these are all the questions, all of them have answers to them. So yeah, this prompt is the first step if you really want to learn the skill. We do that in a community. And now let's move on to the next prompt that most people are not aware of. And this one is kind of the opposite of what we covered so far. I showed you prompts that you should be using. This is a prompt that you should actually not be using. What am I talking about? I'm talking about the concept of a negative prompt prompt. So this is a building block that you can include in your prompt that tells chatgpt what you do not want to see. Now this sounds great in theory, but in practice, especially for beginners and intermediate users, this thing causes more harm than it does good. Let me just show you. If I take one of these preset prompts from the homepage here, I just want a fun fact about the Roman Empire. What a fantastic preset prompt. As a guy Living in 2024, obviously I love this. And let me expand this with a negative prompt. Okay, so this could be a negative prompt. I could also format it like this, maybe, maybe with a little delimiter in front of it. That's a nice trick to keep your things structured. This is a negative prompt. And why should you avoid this even if you see other people recommending it to you? Well, as talking to ChatGPT has a lot of parallels to communicating with human beings. I mean, heck, this thing has been taught on all of the Internet, which has been produced by humans. Right? A lot of the same rules apply. So there's this paradox in psychology. It's called the paradox of the pink elephant. And it's very simple. If I tell you to not think, think of a pink elephant, what are you thinking of right now? Yeah, Every single person, including me, has an image of a pink elephant in their head. Now. Now, a lot of this paradox also applies to ChatGPT because if you bring up something, you're introducing it into the context window that it considers when generating the next messages. This might not be a problem initially, but if you're not crystal clear on the negative prompt, and this would be example of a good negative prompt, because it doesn't get more clear than a human first and last name. There is only one Julius Caesar as far as I'm aware of, and by telling it never to mention him, it will do that. But every single time you'll be following up. For example, I might want to follow up and create a timeline of all Roman emperors here. Now, it will do that while considering not to mention Julius Caesar. And that's fair enough. That's what we wanted initially, but is it really what we wanted? Maybe we just didn't want a fun fact about Julius Caesar, but now this timeline does not include him. And although we were hyper split specific, there is no Julius Caesar to be seen. He should definitely be in the timeline of Roman emperors, but because of this first prompt, he's not. So even though we were super specific, this tripped us up along the way. And I've seen this so many times with myself and people that I teach these skills. If you use negative prompting, you have to retain the fact that you use the negative prompt somewhere in your conversation throughout the entire conversation. And it gets even worse. If you don't use something as specific as a name, you really great if you have a name or a number. But immediately, if you introduce the slightest bit of ambiguity, you're going to run into troubles because you leave room for interpretation and then it can interpret that in any way. So for example, if you would be writing an email and say never use emojis. Well, fair enough. Maybe you don't want emojis in your email. But what if later in the conversation you're doing some web research, you're using the web browsing tool to pull up articles and maybe in your research prompt there's a part of it that says I want exact quotes of the articles. Don't summarize, paraphrase, give me the exact quotes plus the links. Well, what if that article includes an emoji? All of a sudden your negative prompt is going to be counteracting that and it's going to be a mess. You're going to get inconsistent results. These emojis are pretty harmless examples, but this can cause a lot of damage. That's why I would say avoid negative prompts. Always try to prompt with positives. Tell it what you want rather than what you do not want. What do you want? It's not that simple. Now you can still use it. It works really well to refine the style. If you want to remove certain keywords, yes, a negative prompt is the best way to way to do that. But again, it introduces these troubles down the line and I just want you to be aware of them. My recommendation would be stay away from negative prompts. And again, this is something that was not at all obvious to me as I started using this tool. Oh, and before I forget it, if you're enjoying this video, if you learned something, please leave a like it really helps the channel. And with that being said, then let's get into my last prompting tricks. This has become one of my favorites over time and it is so powerful. I can't wait to share this with you. It is about creating templates inside of ChatGPT. And if we take a step back here and we talk about the fact that, and I mentioned this many times, ChatGPT is excellent at recognizing patterns and then reproducing those patterns that it recognized. If you consider that fact, then templates are legitimately one of the most powerful techniques you can employ inside of ChatGPT for anybody who's intermediate or advanced. That's why five shot prompting are pretty much the standard for all benchmarking. If new models come out, you always see MMLU 5 shot because they included 5 shots aka 5 examples at the end of their prompts. That's what that means. For anybody who's not familiar, five shot just means you do a prompt and then underneath you give five examples of what you want the output to look like. Now why do they do this and why is this one of the most popular techniques? Because you show it a pattern and then ChatGPT can follow the pattern. And this is really the secret sauce to really high quality results. Now what I want to share with you is this super simple prompt. Check it out. Transform this into a reusable template. That's really all you need. If I send this prompt, it's going to take one, whatever it did in this case, it's a timeline of various emperors and it's going to turn it into a template, a standardized template. And what this is going to allow me is I will be able to copy this and reuse it in any other scenario where I want to create the same type of timeline. Now look, the timeline is just one example I really need you to think bigger here. I need you to think about every single workflow that might be present in your work. Anything that involves digital text or digital media. You can turn anything text based as of now into a template and then use that template to reproduce it. So let me give you two quick examples. One would be I'll just take this template for this timeline and in a brand new chat that has nothing to do with this, I'll say craft me a timeline for the various Egyptian rulers, okay? And then all I'll do is say template down here and I'll share this template with it. And now it should follow it. So let's have a look if it works. Dynasty or period, the start year, end year, the name, the years of them, it followed it perfectly. Now let's say I'd be teaching people history. I could be using the same template for every single ruler timeline that I ever do, and that would keep things consistent, which is obviously valuable. I promised you a second example and this one might be valuable for most people, whereas this timeline probably is not. And that is for my custom GPT that I use for some email reply drafts, mainly super formal ones like banks or insurances. And as you can see, I have this extra AI advantage email templates down here. Plus there's a building block all the way at the end of the GPT referencing it. Every time the user asks you to write an email, follow the AIAemailtemplate PDF as a template for the email. And with a combination of a prompt like this and a file like this, I simply took the template, put it in a word stock, saved it as a PDF, uploaded it. Here I am able to reproduce the same email template every single time, keeping things consistent. This is again, I can't stress enough how universally applicable this is prompt really is. You can transform anything into a template and then use that template to reproduce social media content, emails, ideas, whatever it might be. So now that you're equipped with these brand new prompting tricks, I sincerely hope you learned something in this video. And if you want more learnings like this and a community of other AI enthusiasts to support you along your learning journey, that's why I built the AI Advantage community. That's the whole point. But if that's not for you, no worries. I'll be here on YouTube teaching you these tips and tricks on how to get the most out of these tools. Plus keeping you updated on what new tools are relevant and what practical tools and techniques like this you should know about within the generative AI space. If you want to keep on learning. I created a video here a few weeks back. It shows another ChatGPT technique in the same vein of the few prompts that I shared here today. I really think you would enjoy that one. But with that being said, I hope you have a wonderful day and I'll see you soon.