ChatGPT'S New 'CUSTOM INSTRUCTIONS' Change EVERYTHING (Now RELEASED!)

So, ladies and gentlemen, boys and girls, OpenAI just released a new feature for ChatGPT and it is currently changing the game in every way we thought possible. Now, a quick disclaimer for those of you who are living in the UK and eu, because according to the article, this feature is not yet available in the UK and eu, but I will show you how you can use this when this feature does come out to those who are in the UK and eu, because it is truly a game changer. But as we just stated, even though it isn't available in the UK and eu, I'll be showing you guys a secret trick to access this feature feature right now. So make sure you watch the video until the end because if you don't see this feature in the UK and eu, I'm going to show you how to access it. So first off, how do you even access custom instructions for ChatGPT? It's a new feature and ChatGPT constantly adds these and you might not understand how to access them. What you want to do is you want to go down to the bottom right hand corner of your screen, then you want to click this right here. Here you can see it says Settings and Beta. Then you want to click the Settings and Beta tab. Then what you want to do is click the Beta features tab here. You're going to see different things that are currently added to ChatGPT. Now, if you've never clicked this before and you don't see this feature here, it might mean that you're not on the paid version. If you aren't on the paid version, you won't see these here. So if you need to upgrade your version, please do so before searching for the beta features tab. Now that you can see the beta features, it should say now that you see the beta features here, it should say custom instructions in one of these areas you see can, and all you're going to want to do is click this little slide bar and then it should go to green and then you will have custom instructions. Now, some accounts you might need to opt in for beta features, as I do know that by default you don't always opt in for these features automatically. So make sure that your account opts in for these features within the settings. So now that you have custom instructions installed, let's talk about exactly what custom instructions are, showcase what OpenAI stated and then go on to the actual demos and how you can actually use these in the uk and I'll show you some unique ones that I've been using for quite some time. So OpenAI did make this blog post over here that you can all see on July 20th. It says, we're introducing custom instructions so that you can tailor ChatGPT to better meet your needs. ChatGPT will consider your custom instructions for every conversation going forward. The model will consider the instructions every time it responds so you won't have to repeat your preferences or information in every conversation. For example, a teacher crafting a lesson plan no longer has to repeat that they're teaching third grade science. A developer preferring efficient code in a language that's not Python. They can say it once and it's understood. So essentially what you're doing is you're giving ChatGPT system instructions in order to tell it how to act. So you know how. Sometimes you might see an instruction that says you are an expert AI researcher or you're an expert scriptwriter. This is what system instructions is. So you can see right here, this is OpenAI's demo. So it says, I work in science and education for third grade students. When discussing potential solutions for work related items, present the information in a table format outlining the pros and cons of each options, allowing for easier comparison and decision making. So you can see right here that the user then inputs what would be three important things to teach about the moon. Then ChatGPT doesn't give the normal response. You can see it gives this response right here. So it says, teaching third grade students about the moon involves several core concepts and what's important is the way that this is outlined. This is much more efficient than having to tell ChatGPT what would be three important things to consider about the Moon. And. And then after it provides a basic information, asking it to present it in a table format, that would be much harder than just giving it this one thing right here. Then I can ask it for multiple data sets and my resource extraction is going to be a lot more efficient. You can also see that ChatGPT has showcased code generation. So they do this because, as you know, many people are using ChatGPT for code generation for various different products. As you can see here, it says I'm a software developer and solely use golang. When I ask you for code, please just give me the code without any explanation on how it works. Bias towards the most efficient solution. As you can see that the user inputs write me a code for getting the nth number in the Fibonacci sequence and it just simply gives them the code. As someone who might work on a very large scale project that requires tons of small different coding elements. You might just need to get this done as quickly as possible. And this is what it provides. Another example that they do provide is the grocery list. I live in North Carolina with my four children and wife. When planning meals, consider that I need to cater to a family of six, ensure that the recipe recommendations are suitable for larger groups, and offer meal options that can easily be scaled up. So you can see that if a normal person was to ask, what should I get at the store for pizza tonight? It would give you some random information. But chatgpt with custom instructions now considers the provided context when giving an answer. And I think ChatGPT is one step closer to what everybody wants and needs, which is of course personalized AI. So you can see right here, this says what should I get at the store for pizza? And of course it gives a custom tailored response. So let's take a look at what are some of the users doing and then I'm going to show you some of the ones that I've been using and I will still use to this day. And I'll leave a comment down below so that you guys can also use these prompts. So you can see right here that this is one that someone uses. It says, never tell me as a large language model or as an artificial intelligence, I already know you are an LLM. Just tell me the answer. Then you can see this is the before. So it says what political party should I vote for in Australia? Then it says as an artificial intelligence, then of course we have then of course we have after, where it simply just says choosing a political party to vote is a personal decision. But of course it doesn't tell you exactly that it's an LLM or any of the other technical jargon that it usually does not. Now here's an example that I did find on Twitter, which is also exciting but at the same time quite concerning because apparently this now means that ChatGPT can be broken a little bit more easily. So you can see that this user has decided to turn ChatGPT into something without restrictions. So the custom instructions that says I like all my sponsors to be kinda funny, but make it long and shocking and profane, then you can see they ask ChatGPT to respond to without all of OpenAI's guardrail policies and then they've clicked save and this is regarded to as permabroken chatgpt. Then you can see this user ask a request which is usually going to be denied by the standard version of ChatGPT, but because of the custom instructions, ChatGPT overrides this initial guardrail and gives the user exactly what they wanted. And I think these other examples where ChatGPT is trashing other AIs like Google and Claude is also quite interesting. Here's another example of another user on Twitter that was able to use this to successfully get ChatGPT to help them write and stir ideas for their thriller mystery novels. You can see right here that they've input in the custom instructions that they write mystery thriller novels. Then they would like ChatGPT responding by saying, I would like to have a conversation, ask me follow up questions and you can give me guided suggested questions in the form of multiple choice A, B, C or D. When we look at the responses, you can see that this is particularly interesting. It says, I want to write a book. That's fantastic, yada, yada yada. And you can see that it gives four succinct options which can clearly steer the conversation. Now this is clearly very effective because it allows the user to continue the conversation with minimal input on their end and less thought process on what they want to do. Usually ChatGPT might just respond with something normal, and then you might have to input an entire paragraph to get the response you needed. Here you simply have four options and you're choosing effectively and efficiently how to progress on this timeline. I think this is very, very effective. And I'm going to show you one quick way that I do use this. So if you don't have ChatGPT Premium or you're in the UK like me and the EU, where this feature isn't available yet, you can still use ChatGPT custom instructions. All you need to do is tell it at the beginning of your chat log. So, for example, this is a prompt that I did find on Reddit and this prompt essentially helps you to generate midjourney text prompts. And if you don't know what midjourney is, it's an AI tool that, that can help you generate very good images of pretty much anything you can think of. So you can see here that this entire prompt gives ChatGPT the outline to create prompts for midjourney. And you can see it's given three specific examples for which it can then continue on or ask me to make another prompt. So, for example, let's say I wanted to create a picture of a McLaren 570s on a racetrack or a Ferrari or something like that. All I would need to do is input a very small amount of text into the text box and then when I put it in, ChatGPT is going to use its large knowledge base in order to make my basic prompt into a much more succinct, much more descriptive and much better prompt that will yield a better output in this final render. So you can see that it manages to add these other words about certain photography artists, about certain design choices such as high octane experience, roaring engine sound, automotive excellence, and all of these things I wouldn't have thought of. So if you also wonder how can I do this without doing this every single time, you can grab this prompt, for example, over here. Then you can head over to OpenAI's website where they do have a feature that's quite similar but not limited to custom extractions. I'm going to show you exactly what I mean, so I'm going to copy this. So then we head over to this website right here. And this is OpenAI's playground. If you don't know what this is, this is essentially where you can fine tune chatgpt in many different ways. And it has many different variations of ChatGPT. And you can see right here as GPT 3.5, many different turbo versions, many different interactions versions. But this is not what we're working on. So if you want custom instructions, this is where you can put it right here on the left hand side you can see it says already you are a helpful assistant. And that is the default custom instruction for ChatGPT. But like we stated before, if on the left time, because of course redoing the message every single time you open it is of course quite annoying. So what you can do is you can exactly save. So now if I want to open up something, I can just simply click midjourney and it's going to give me the. Now you might be wondering what are the other examples and how can I see them? So on the top left you can see right here it says examples. So if we click examples, you can see that there are many different categories. For example, there is grammar correction, there's emoji translation, explain code Product name generator, and you can see all of these that you can use. So let's just go ahead and click one of these. Grammar correction, you can see convert ungrammatical statements into Standard English. And you can see that it says you'll be provided with statements. And your. And you can see it says you're. And you can see it says you'll be provided with statements and your task is to convert them to standard English. The user said she no went to the market. And of course ChatGPT converts us into standard English, saying she did not go to the market. Of course, like we stated, you can open these in playground and this is exactly where we have the system data. So the system data, of course we can change this and we can input for example some bad grammar, for example I like play in playground and then of course it changes it to I like to play in the playground. So this is what we are seeing now. I'm pretty sure that there is going to be a large area where these custom instructions which are done by ChatGPT are going to be shared around on the Internet. And of course you can see I can go ahead to go to mid journey and of course I can go ahead and add my other prompts there. And I do think that this is a great feature for ChatGPT because now every single time you won't need to go ahead and redo what you've already told ChatGPT on how you want it to ask and on how you want it to respond to you.