Prompt Engineering basics using ChatGPT

1. Skillshare intro: Hey there, Are you looking for a course on prompt engineering that can get you started with Chat GPT right away? Well, you may stop looking now. This is the course you need. Studies show that prompt Engineering, the skill we're diving into in this course will be the most in demand in the coming years. Without any further ado, let's get started. The course is developed under the guidance of Mr. SNS. Roy with over 50 years of experience in India's topmost companies. My name is aka Sharma, I am a professional instructional designer. I have been using Chat GPT since its inception. Now we are excited to share all our learnings with you through this course. The course includes a complete section on basics of chair part navigating its interface, followed by the most in demand part, the prompt engineering. This part includes all the information you need to write effective prompts. The course also includes videos on custom instructions and how to reach out to chair jeopardy support in case you get stuck. And that's pretty much all that you need to use Chair Jeopardy like app. This is not just any other course with overwhelming prompt, guides to copy and paste. I'll teach you the basics of AI models and the art of prompting. Once you grasp these fundamentals, writing effective prompts becomes a piece of cake. So if you are ready to write the EI wave, if you are ready to be the part of the future, this is your chance, join us in the Chat Pty course now. 2. ChatGPT Let s talk fundamentals: Hello everyone, welcome to the first lecture. In this video, we are going to be talking about GPT and mostly it's fundamentals. Let's get started. In this video, you will learn about artificial intelligence, natural language processing architecture of EI models. Because right now there are a lot of EI models available. It's a chat GPT based course. I have not covered other EI models because they all work in the same manner. Let's get started first thing first, what is artificial intelligence? Because it might be the case that you haven't heard of Jet GPD, bar, mid journey, or any other AI models, but artificial intelligence has been there for a while now. We all are excited and a bit scared about the world that artificial intelligence will create. Some of it is because of the movies that we have seen so far and how it portrays artificial intelligence in general. But that's not really the case now that artificial intelligence is there and we are able to use it, we all know that how it's helping us, the possibilities are even bigger. We can't even think how it can help us make the world a better place. First thing first, what is artificial intelligence? It's a technology that makes a computer or software behave and think like humans. To some extent this is very important because we don't know any I that can totally imitate humans. Artificial intelligence are there. They're able to do certain things that humans can do, but they cannot behave completely like humans. Now you know what is artificial intelligence? Let us talk about how do we make it? He, we use an analogy to understand artificial intelligence. Suppose you have a friend. He's very intelligent. But how did he do it? He read a low top books, articles, research papers, etcetera, et cetera, et cetera. All this made him a very intelligent person. You can understand artificial intelligence in the same way, artificial intelligence is nothing but a system. A technology that can make your computer just like your friend. Now, where did the information come? It came from Internet. The technology that we use to feed this information from Internet to your computer is known as LP Natural Language Processing. You can think of NLP as a way that your computer uses to gather all this information and then make sense of it. In the same way, for your computer to be able to give you the response, it uses NLP. Nlp is very important here. You use the NLP to train your computer. It's a part of machine learning. Machine learning is nothing but you help your machine learn. As the name suggests, NLP is a specific tune language. You can feed text based data and your computer will read it, understand it, then we'll be able to help you in the same way there are other technologies like image processing, data processing, and video processing, et cetera. All these you can do to make your computer smarter. Now let's understand the architecture of AI models. If you think about your process of learning, you will very much understand how I works. You gather a lot of information. Basically you learn, then you process the information to make sense of things. When a new problem arises, you use the information that you already have. You process it, and then you create solutions. Architecture of artificial intelligence is very much similar to your brain. I also has a lot of neurons, just like your brains. What these neurals do is they take information, they interpret it, and then they generate results. This is neural network and it is known as transformer. When you feed information, it interacts with encoder. The function of encoder is to process this information and convert it into a language that your computer can understand. Then this information is processed in the neural network and it gives you a result. But before that, it is again coded. This is a general architecture of AI. There's a lot to it, but you don't need to understand all of these things. Basic understanding of how it works is sufficient. There's just one more thing that you need to understand is training. How do we train? There are two types of training. One is pre training. Pre training is a process of feeding your artificial intelligence model with a big set of data. It's not structured, it's just a big set of data that you feed into your system. And it just uses this information to learn things. It's known as unsupervised learning. You don't help your computer understand it, you just feed the data and then let the computer make sense of it. But then you do fine curing. Fine tuning is what made GPT bar and other big models possible. It is basically helping your computer with specific set of data to help make sense of entire data. Let's take a simple example. You are given all the ingredients available in the world. This is like pre trailing, everything is given to you. Then in the fine tuning, you are taught that if you mix ingredient 1.2 it'll create dish. This is how you are helping yourself understand how you make a fine tuning, helps your computer understand the context. It is very much needed because whenever you ask a question to your computer, it has to give you a specific response. It cannot just take out some data from this pre training data and give you a response. You want a very customized response and that is only possible because of fine tuning. Let us move on and understand the core of this course, which is Chat PT. Chat PT is nothing but a EI model. When we say it is fed with a lot of data, it has become a large language model. An AI model, which is trained on a big set of data is known as Large Language Model. How does it happen? It happens through NLP. We train a model using NL. Gpt is the name of the model. When you keep training it, it comes in Verss. Right now we are using GPT 3.5 or GPT four, which is paid in this course using GPT 3.5 Assuming that most of you are just starting out, wants to use the free version before you become confident and comfortable in paying for this technology. Gpt stands for Generative Pretrained Transformer. We have already talked about transformer. It is basically the architecture that is making all of it possible. This is at doing all the processing, everything. It is the do, it is doing all the things. Pretinasdel is trained on big language models. Generative is basically it can generate responses. If we go back to our example of your friend being very intelligent, what exactly is it doing? It is talking to you in your language. What have we created so far? We have created an AI model that can talk to you in your language and can help you in various tasks. Gp is nothing but a language model that can talk to you like a friend and can help you in various day to day tasks. It is the simplest definition that's there define a GPT precisely. It is a language model based on the GPT architecture that we have already discussed, which is nothing but a neural network that has been trained on a massive amount of text data from the Internet. Because of all this, we have a model like Chad GPT that we are learning in this course. Thank you. I hope this video will help you understand the fundamentals of J GPT. I'll see you in the next video. 3. Exploring the capabilities of ChatGPT: Hello everyone. Welcome to another video of this course. In this video, we're going to talk about the capabilities of chat GPT. Let's get started. If you are taking this course, that means you are already fascinated by the capabilities of Chat GPT and how it harnesses the power of AI to assist us in our daily lives. Gpt is like Swiss Army knife of I technologies. It can help you in almost every aspect of your life, Be it solving a problem, helping you with day to day tasks, or in case you need a conversational partner. Let's move ahead and talk about all the capabilities of Chat GPT and how it can help us in different tasks. Ch GPT is a vast knowledge repository. In the previous video, we have discussed that Ch GPT is a large language model. That means it is trained using a big set of data. This is the same data that you surf on the Internet. You browse this data to gather the relevant information. Gpt has already taken this data and trained itself. Now all you need to do is just ask a simple query and it'll give you a relevant response. Now you don't have to surf Internet for hours to get the relevant information, understanding and generating natural language. This is probably the biggest reason why Chat GP is so popular. The definition of Chat P itself says it's a language model that can talk to you in your language and that's the biggest advantage. Chtgpt can talk to you in your language, it can understand your language, it can generate responses in your language, and that's what makes it very useful in our daily lives, rapid information retrieval. It's again an extended point of the first point, vast knowledge repository. As we discussed, you don't have to surf Internet for hours to get the relevant information. You only need to ask the right query and it'll give you a response within seconds. That is one of the capabilities of GPT, multilingual capabilities. Chgpt is trained on very big data and it includes different languages as well. When you have a model trained on different languages, it can give you response in these languages as well. Not only it is capable to give you the right information, it is also capable to give you information in your language. Fascinating, isn't it? Versatile, content generation? This is the need of R. We are living in the digital world and the Internet is full of content, be it videos, text, images, whatever it is. But Internet is full of content in this era. Gpt is making it very easy for you to generate content. It may not be able to help you in creating a video, but it can definitely help you in creating the script for the video. It's a versatile content generation tool, problem solving and brainstorming. This one is my favorite. Pt is not only helping people gather information fast, it is also helping a lot of people solve problem and bins storm on different ideas. Since at GPT has its own brain, it, it can analyze data and it can give us insights that you may not be able to even think. I'm not sure if you have heard about it, I'm not sure if you have heard about it. It happened recently that when a person uploaded his health checkup reports on Chat PT and asked for the possible diseases while Chat GPT did not conclude which disease is it, but it definitely gave all the options that are possible using the reports and the symptoms. Doctors were also surprised to see. Doctors were also surprised to see that it was a disease that they were not even thinking. Which made it very easy for the doctors to rule out all the other possibilities and start the treatment on time. Personalized assistance, we all have been using boards, chat boards and Google assistance like tools. But with the help of tools like Chat GPT, they are becoming more and more personalized. Just like a real human being. Chat GPT is now capable to give you a very personalized response based on the query you asked. All you need to do is ask the right query. And Chat GPT will use all its data, all its capability, to give you the right response. This is very interesting. Chat GPT is a conversational AI tool. When we say conversations, it can have any type of conversations with you, be it general friendly conversations or even seeking legal advice. Chat GPT can help you with a lot of day to day things that you cannot even think of. Last but not the least, it is helping programmer to write codes and debugging them. These are all the capabilities. The idea of this video is to give you a general sense of what all you can do using chat GPT. We'll deep type into each of them in the upcoming videos. For now, that's all for this video. I'll see you in the next video. Thank you. 4. Limitations of ChatGPT: Hello, everyone, Welcome to another video of this course. In this video, we're going to talk about limitations of Chad GPT. Well, just like any other technology, Ch GPT also has a few limitations, not just ChGPt. You will find these limitations with almost all the AI models that are available. It is very important for you to know these limitations so that you can use them effectively and efficiently. So let's get started. Just like any other technology that is made to do something is not able to do the other thing. For example, this robot is not made to climb the stairs. Similarly, Cha PT also has a few limitations that you need to know before actually using it. Before we even dive into what are these limitations, let's talk about why exactly these limitations are even there. The main reason is data that we have used to train the AI model. This is very important. As discussed in one of the previous videos, chat GPT is nothing but an EA model that is trained on a big data. Now, the success of any EI models depend on the data we have used to train it. Any issues with the data will result in the issues with the final I model. Let's first try to understand what these issues are. Well, you data could be outdated. If you're using a data that is outdated, the result it will generate will also be outdated. And that's a very common challenge that might happen. For example, Cha GPT 3.5 version is trained on a data till 2021. Now, if you ask some information that is from 2022, the result will not be accurate. In that case, Chad GPT itself will say this, that the result that you are expecting may not be accurate as it is trained on a data till 2021. Only other challenges could be bias. If you have used a data that is biased towards anything, then the result that you will get will also be biased. See everything is depending upon the data incorrect. If you used an incorrect set of data, then the result that your model will produce will also be inaccurate. Now chat GPT and other A models are trained on a very big set of data. It's highly possible that this data may contain some inaccuracies. That is one of the major reasons why you should not completely rely on the results that chat GPT produces. You should always have a cross check mechanism in place to make sure that the information you are getting is authentic and true. Now that we know what are the main reasons something like artificial intelligence has limitations, let's talk about limitations themselves. First is lack of true understanding. A GPT is not a human being. It's a I model trained on a big set of data. It works on the data and it gives you results based on this data. It lacks true understanding of things. All we can say the common sense. If someone asks a very simple question to you, you will use your common sense to respond to that. That is not the case with Chad. Gpt. Chajpt will use data, the results are completely based on the data that is fed, based on the common sense to make sure that you know this. That is one major limitation of GPT. Next, potential for inaccuracies, as we have discussed that data that you have used may be inaccurate. Now, any inaccurate data will produce inaccurate results. Because of this, AI models can definitely generate inaccurate results, vulnerable to biases. This we have already discussed. If the data that is fed into the AI model is biased, then the results will also be biased. It is as simple as that. No personal experience, again, everything boils down to the data. It has no personal experience with things that you're asking for. For example, you are planning a trip to Dubai. Now you ask Chad PT to help you plan this trip. Chad GP will be able to tell you a lot of things about Dubai and what are the different things that you can do when you get there, but it will not be able to give you responses based on the experience. On the other hand, if you have a friend who have been to Dubai and you ask him about his experience in Dubai, he'll be able to give you some really good insights that Cha GPT will never be able to give you. So this is where you need to understand that whatever response chat PT is giving you, they're not based on experience, they're just based on data. Next we'll talk about some limitations which depend on your prompts. For example, non deterministic outputs. Non deterministic output simply means that for a simple prompt to warn, you may have multiple responses. Now, this is non deterministic. This is non deterministic. If you're asking the same question and you're getting three different responses at three different times, it's non deterministic. This is one of the limitations. You'll have to rely and use one response, but you cannot just rule out the possibilities of others. Limited context. It is very important for you to know while lighting prompts. Also, let's say you are having a conversation with Chat GPT. You started with prompt one, now you are at prompt nine. That means you have already given nine prompts in your conversation, and chat Pt has given you at least nine responses. Now if you expect that at the nine prompt, chat PT remembers everything about prompt one, then it's not a good expectation to have because Chat GPT loses context while having the conversation. The more forward you are with your prompts, the less context it'll have of previous prompts. In case you are interested in something related to the previous prompt, make sure that you give the context. Again, handling multiple requests in a single prompt. This is a very common error that people make while writing prompts. You have to understand that Chat GPT as a AI model is capable of catering to one request at a time. If you try to include multiple requests in a single prompt, the chances of getting a poorly structured response is totally possible. Always make sure that if you're asking five different things, ask them in five different proms. Don't include them in a single prompt and expect a perfect response. That's all for this video. If you are thinking that I'm not including examples, please be assured that in the coming videos we are going to use chat GPT only. These are very general video for any prompt engineer to ace in the field. I'll see you in the next video. 5. How to overcome the limitations of ChatGPT: Hello and welcome back. In the previous video, we have talked about all the limitations of Cha GPT. If all the previous limitations made you think that even after having a lot of limitations, why are we even using Cha GPT or similar AI models? Well, the answer is very simple, just like any other technology. While PT may have some limitations, its potential is way greater than its limitations, just like any other limitation. These things can also be overcome. In this video, that's what we are going to talk about. If we have a few challenges or limitations of Cha GPT, how can we overcome them? Let's talk about them in this video. All right, to overcome these challenges, you should always think critically. Well, this is very important if you are using an EI model, and you know that it might have a few limitations like bias, inaccuracies or outdated data. You should think critically always. You should be on your toes to find out errors. And you should have a reliable strategy in mind to cross check the data to re, verify it and then use it in your work. Understand and keep in mind its true nature, capabilities and limitations. When you know both capabilities and limitations, you know that capabilities are way higher than limitations, Then you can simply overcome these limitations, but you should not forget them. Because when you know the limitations, then you'll always pay close attention to the things that might go wrong. Make sure you understand both and always keep them in mind. All right, this is very important. You should use chat GPT or any other AI models as resources and not as the definitive. So when we say definitive source, you are trusting the AI model blindly and you should not be doing that, use it as any other resource. For example, when you do a Google search, you get, let's say, five results. If you are a true Google user, you will always open two or three of them in a new tab, and then you will read all of them to ensure that they are correct, they match, and the information presented is reliable. Similarly, use GPT as another resource and not as a definitive source. Then it comes to being clear and specific. When you write a prompt, make sure that you are clear and specific, that the information that you get is very structured and you can easily cross check, re, verify it from other sources. Last but not the least, responsible and ethical usage. Now this is very important and we'll talk about it whenever it comes up. As we have discussed, ML and NLP are two of the main foundations of any EI models. You use NLP to give prompts and receive response. Machine learning uses all the problems that you insert and all the responses that you receive to train itself. If you don't use it ethically, if you do anything that is not right, it will learn that also it will not only harm you, but other users as well. So make sure that you use it responsibly and ethically. These are some of the major points that you need to keep in mind in order to overcome all the challenges that GPT peasants. Other than that, it is a very transformative technology that is going to disrupt the entire tech world. So make sure that you are prepared enough to use it in your advantage. That's solve for this video. I'll see you in the next video. Thank you. 6. How to access ChatGPT and interface!: Hello and welcome back. In this video, we'll talk about chat GPT and how to access it. From this video onward, we'll start with how to use chat GPT, But for that to happen, you'll have to know how to access it. In this video, we're going to do that if you simply search for chat GPT in your Google search or your Bing search, you'll get a lot of options from all these options. You can click on the option with the URL, open Eye.com forward, chat, GPT. If you just click on this, you'll get this interface. It is a lot of things, a very minimalistic interface. The company is known as Openai. These are some of the important links research. For instance, it is all the products that Pena has created, API. It's not very relevant at this stage of this course so we'll skip it. Chat DBT is the product that we are learning in this course. We'll talk about all of them. Safety, it'll talk about safety here. You can learn more about the company open. There are other options as well. You can use login to login into an open account. Here it is, again saying tratGPT because this is the interface of chat GPT. Now, there are two main ways you can use chat GPT. One is app and the other one is Web. You can download an IUS app or an Android app using these two links or you can try it on Web. I'll create a separate video on how to install and create an account on your phone. Now let's use it on Web. If you just click on Try on Web, it will take you to the Chat GPT interface. I am directly redirected into this account because I'm already logged in. But if you don't have an account, you'll have to sign up first, create an account, and then only you'll be able to access this to create an account. The process is pretty simple. You'll have to provide with your credentials, that will include e mail, ID, your phone number, and you'll get an OTP on your phone to be able to access it. Once you do that, log in and you will come to this page chat. Gpt has two versions as of now. P 3.5 P four. Gpt 3.5 is the free version that is available for everyone. Gpt four is a paid version and you'll have to pay a certain amount to be able to access it. In this course, we are only going to target Chat GPT 3.5 because it is free for most of us who are just starting out. This is good enough. I add more videos in this course for chat GPT four because I'm assuming that by the time you feel comfortable using chat GPT four, or I'd like to say by the time you are comfortable paying for chat GP four, the videos will be added. For now, we're using GP 3.5 GP 3.5 is a bit older version of Chat GPT. This is the version that is trained on a data until January 2022 based on the recent information. P four is the version that is trained on a much latest data. Let's talk about the interface ones. This is a sidebar. If you just click on this icon, it'll collapse it. Click on this icon, it'll open it. If you just click on New Chat, it'll open a new chat for you. This is how it'll open for you. These are all the chats that I've had so far. Here again, you can use update plan to go to chat GPT four, here is my account, it's my name here. I have two options, in fact, three options. Custom instructions. We'll talk about it later. Click on Settings, it'll open this dashboard. It has two options, general from where you can change the theme. If I just click on System, it'll remain the system theme, whatever I'm using on my system. If I click on Dark, it is the same theme, the grayish theme. If I click on light, it'll change to light theme. I'll be using the dark theme. For now, clear all chats. If you want to delete all the chats that you have had with chat GPT, you can just click on this clear option and it'll delete all of them. I'm not doing it because I don't want to. Second is data control. From here you can just pause the Chachi tree and training Chat, You understand training is basically nothing but chat. Gpt trains itself using the problems that you ask and the response it gives you. All the checks that you are having. With chat GPT, it is training itself. If you just pose it, it will stop recording your chats and history. I'm enabling it because it just makes it very easy for me to go back to any chat I had. Other options here are shared link, you can manage all the links that you are sharing. You can export all your data and you can simply delete your account. These are all the checks that I had with chat GPT. If I just click on any of them, it'll open the chat for me. I can use this icon to delete any check. I can use this icon to rename it. For instance, let me just same quantum physics, it will rename my chat. Chats are categorized into different categories based on the time. All my today's chat are here, yesterday's are here and the other chats are here. Talking about the chat interface on top, it is telling you which version of GBD are you using. This is the custom instructions that you have given. I'll talk about it in some time. Using this option, you'll be able to share your chat with anyone through any channel. This is the response, this is the prompt. You can use this icon to edit your prompt. For example, in simple terms, 410 years old, clicking on Save and Submit, it will regenerate the prompt for you. You can click on this stop generating, and it'll stop the generation for you. If I just click on Regenerate, it will restart the generation. You will have a fresh response. These are the three options that are given to you. We'll talk about them in detail. However, for now, if you just click on this option, it'll copy the entire response for these are the two feedback mechanisms that are in place. These are the two options given to you through which you can share your feedback. If you click on this option, it'll tell GPT that you like the response. If you click on this option, it simply means that you did not like it. We'll talk about them later. This is the prompt bar where you will insert your prompt and click on this to send your message basically. Or you can click Enter to start generating the response. This is a very important message that we have been talking about so far. Chat DPT and make mistakes. Very important information. So make sure that you always cross check the information. This arrow will take you to the bottom of your chat if there's any here, you'll find other options like Help and Eff. If you want to access support or want to go through some frequently asked questions. These are release notes. Basically, whatever new things are happening with Chad GPT, you'll find them here. Thomson policy is very important. If you get time, you should go through them and keyboard shortcuts. We'll have a separate video on this also. That's all for this video. I hope you will now be able to access Chat GPT. I'll see you in the next video. 7. Shortcut keys: Hello and welcome back. Now that we have understood how you can navigate through chat GPT, it's time that we learn some shortcut keys that you can use to make your workflow faster. Shortcut keys always help us in speeding up our workflow. In the case of chat GPT, also, there are a few very important shortcut keys that you should know to really. Here I have logged into my chat GPT account and I'm using GPT 3.5 not using the paid version because this course is made for beginners. And I believe that a lot of you are using chat P 3.5 to start out. By the time you reach to chat GPT four, I'll add videos on that, so don't worry about that. Now this is your chat interface. You can simply toggle this sidebar to open all the chests that you have had and close it using this icon. These are some suggestions given by chat GPT how you can use it. This is where you type a prompt, this is where you get some help release nodes. Thompson policies, keyword, shortcut. This is a chat window. To open a new chat window, you can use a shortcut key, control plus shift plus O. It'll open a new chat for you. For instance, if I type a prompt, what is an apple? Now if I want to open a new chat, I can simply use shortcut key control plus shift plus O. It'll open a new chat for me. Next is tobal side bar. Now if I want to see all my previous chats, one way to do that is using this icon to open the sidebar. I can also do that using a shortcut key, control shift plus, hold control shift together and then click to close the sidebar and control shift plus to open the sidebar. The third shortcut key is copy your response. Let's ask a question again. Explain quantum physics in simple Toms. Now I have a response. One way to do it is I can simply click on this icon and it'll copy this entire response for me to show you. I'll just paste it here. This is the entire response that I've copied and pasted here. But there's a shortcut to do that. If you don't want to use this icon, there's a shortcut to do that, that you can use the shortcut key Control plus shift plus you'll get the prompt here. Last response, copy to clipboard. Moving on, we have another shortcut key. Let's open a new chat. Write a Python code four. Hello World Program. This will give me a Python code four. Hello World Program. This is probably the first time we are asking chart PT to help us in code. This is how a code will look like in chat GPT. It'll give you which language you have used the code and you can simply copy code from here. This is the icon, but there's a shortcut key also. That is you can use control shift plus Sem column. It'll just copy the code for you if I just paste it. Here you can see this is print and hello world, the Core only, not the entire response. If you want to copy the entire response, again, control shift, see it'll copy the entire response. See this is the entire response. The next is delete chat. Let's say if I'm not interested in this chat anymore and I want to delete it, I can do it using control plus shift, plus backspace. It'll ask me to confirm it. If I just confirm it, it'll delete this, Chet, and open a new chat for me. Now, there's something called custom instructions that we'll look at in the coming videos. But to open the custom instruction, what you can do is you can you can use the shortcut key control plus Shift plus I phone instruction. And it will open the custom instruction dashboard for you. We'll talk about it in the coming video. The other general shortcut keys like copy something or paste something, will work as it is. For example, I'm not interested in copying this entire response, I just want to copy this part. I'll just select this part, control C. I can paste it anywhere I want. This is how you do it. Now, all the shortcut keys that we have looked so far, you can find them by clicking on this icon and keyboard shortcut. There's a shortcut key also to open this dashboard that is control for our sla It will close this. That's all for this video. I hope you enjoyed it. I'll see you in the next video. 8. What are prompts: Hello everyone, Welcome back. In this video, we are going to be talking about Tmp. As you might have already heard, there's a lot of worse around the word prompt engineering or the profile prompt engineering. The only reason why it's being popular is because AI is the future. The one who can make the most of AI will lead the industry. It's very essential for all of us to learn I and then learn how to make the most of the AI. The reason prompt engineering is the future is that EI is the future. If you cannot use AI well, you will have certain disadvantages to keep yourself contemporary, to keep yourself updated, to keep yourself to the mark where you can use EI in your job. It'll always help you stand out. Let's start with the very first thing. What is Tom? Before that, I'd like to start this video with a famous quote from Thomas Berber. He is an American novelist. This is a very famous quote of Thomas. He said that the art and science of asking questions is a source of all knowledge. If you just think about it, whatever we have been able to learn, it all came out of curiosity. When we say art of asking questions, we don't say that. You can always ask questions to you, teachers, you can ask questions to your friends. Also, if you ask the right question, you can get a treasure of knowledge. This is exactly what he's talking about. Asking question is an art and science. You ask the right question, you get the right information and that's how you grow. Why we started with this is prompt is nothing but art and science of asking question in the right way. Let's take two very simple examples. I want to know a little about apple. The first question that I asked to chat chpt is, what is an apple? Then the same question I just ask, define apple. Very similar things but two very different answers. You can see the details given in the first response versus the details given the second response. First is fairly detailed. It has given you a lot of information about taste, it nutritional values, and then certain other things. Here, it has talked about all these things, but it's very concise. If you ask the right question, you will get a very detailed and a very well framed answer that is all about prompt engineering. For you to be able to make the most of TI, you'll have to ask right questions, or you'll have to ask right query. Another example. And this is something I have taken from Chad, GPT self. If I ask, explain the process of photosynthesis in plants versus there is a Prom two which says, discuss the importance of photosynthesis in the context of Earth's ecosystem. You understand the difference, right? If you think that by asking this question, I'll be able to get the information about photosynthesis with respect to Earth's ecosystem. Then I'm wrong. The main thing that you have to understand is I, and you're not connected. There's no link between you and I. I cannot understand what's going on in your mind. All you have to do is make things clear to AI. Ask the right question so that it can help you with the best of its knowledge and resources. Make sure that you ask questions, you specify things that you are interested in. And rest of the job I will do. What are problems? Although we have already talked about Proms a lot. Let's define Prom. Proms are specific instructions or queries given to EI models. I've used the word EI models because this course is made using Chat GPT. But you might be using other AI models like Bart or any other EA models. But in all these EA models, all you have to do is take care of your queries, take care of your Proms, and it'll give you the best results. I've highlighted the word specific, You'll have to specify things, otherwise your prompt is not a good problem. When I say specify, you'll have to specify everything. You'll have to specify the context, you'll have to specify the goal. Then maybe you'll have to specify with certain other things that we'll talk about next. But the thing that you have to understand is you'll have to be specific and you'll have to be detailed. If I broadly classify props, most of the times you will be asking for explanations, advices, or certain new ideas. There will be other things. But I think that covers almost 80, 85% of your day to day work. Now if you're asking for explanation, it could be if you are a student, you might want to understand how a bicycle works. Describe what Cyc or explain the concept of brevity, Ted Debt will give you enough explanation. These are simple prompts. You might also be interested in seeking advice. For instance, you are planning to start a vegetable garden. You can ask for advice how to become a successful garden. If you are preparing for a job interview, you may have a question about how to prepare for a job interview, and it will give you certain advices. If you are struggling with time management, you can ask certain things, what are the common strategies? But you understand these are very general things. If you just ask for successful gardening tips, it'll give you certain tips. If you ask for certain tips to AC job interval, it can give you certain tips. If you ask certain tips for managing your time, when it'll give you. But again, as I mentioned, you can be specific and it'll give you more personalized and better responses. We'll talk about it. Or you might also be interested in creative solutions. This is very interesting. This is my personal favorite. Let's say you have a limited budget, a family vacation. You can ask chair GPT, How can you plan a fun and memorable trip without overspending? You can specify your budget also, if you have a place in mind you can specify the place. Also, it'll give you certain good ideas to make sure that you have a fun trip, but you don't overspend. Again, design a new and imaginative board game that people of all ages would enjoy. Play fun, isn't it? It could help solve bigger problems. For instance, imagine you are tasked with reducing plastic waste in your community. What innovative solutions can you come up with to promote recycling and reduce single use plastic? Can you see it's a bigger problem these days and this is where I can help you. That's when we say that EI is the future. It can help us with the solutions that we have never thought of by now. You must have understood that you can ask almost every to chat PT or any other A modern. I'd like to highlight the point again that as the user of AI, we have to use it responsibly and ethically, just like any other technology. I can also have bad applications, so make sure that you use it responsibly and ethically. It's very important to secure the environment and make the EI accessible to. 9. Common errors while writing prompts: Hello everyone. Before we actually start with how to write effective Proms, let's look at some common errors that usually people make while writing Proms without any further delay. Let's get started. All right. The very first common error that people usually make is that they lack in clarity in their prompts. Vague and ambiguous Mpts can lead to inaccurate or irrelevant responses. That is so true. If you don't ask a clear question, you cannot expect a concise and clear response. It is very important that you ask the question with a lot of clarity, so that the AI model can understand your question and give you a proper response. For example, what's new in tech? It's a very vague question. We have included the word tech and we're asking the AI model to give us everything that is new with the tech. Tech is an umbrella word. It includes a lot of things. When we say tech, everything around us that is related to science or technology falls under this category. Asking vague questions is not at all recommended. You should try to frame your response with a lot of clarity so that you can get proper responses. For example, the same prompt, we can write it in this way. Can you provide an overview of the recent advancements in AI and their practical application? Now if you think about it, the part tech that we asked in the first prompt consider, this is tech. This is everything that is related to tech. It includes space tech, it includes IT, it includes mechanical, everything. Everything related to tech falls under this space. When you ask a question specific to recent advancements in AI, first you went into, let's say this is AI. In the AI also, you are only interested in recent advancements. That's this area. Let's say now you know the AI model exactly knows that you are interested in this area. And it will give you a response from this area and not a very general response from the entire domain. You now understand why it is very important to have clarity in your terms. Another question is, tell me about Indian history. Again, a very broad question. We have books on this topic, Indian history. How can you expect EA model to give you a response on such a broad topic? You need to be a bit more clear with your prompts. Now, the same question can be asked in this way. Another common error that people make is overly complex prompts. In the coming video, we'll talk about why it is very important to be specific with your prompts. However, excess of anything can lead to more problems. If you try to be more specific, it could increase the complexity of your prompts. How does it work? While your specificity is important, overly complex questions with multiple sub questions can be challenging to answer effectively. It's very important we don't need to complicate our question by asking so many sub questions. What you can do is you can ask these questions separate prompts. This is the effective strategy. However, now that we are talking about errors, you just need to understand that overly complex questions will not yield a ty, effective response. Make sure that you keep it simple. For example, explain the entire process of space exploration, including the history, technology used, and future prospects. Here in a single prop, you are asking the entire process, history, technology used, and future prospects. Again, a book can be written on these four topics separately and you want Ted PT or NEA model to give you a response including all of them. Not possible, right? To be effective, Prompt engineer, don't try to include all the questions in a single prompt. Instead, break them down into multiple questions. And then as a good response on this point could be, what are the key technology advancements that enabled space exploitation in the 20th century? Now we are being specific by asking key technological advancements in 20th century eliminated all the other centuries, all the other time frames and potential future missions. That will give us a very concise and to the point response. Similarly, discuss the impact of globalization in the world economy include trade policies, cultural exchanges, and labor migration. You see we're asking a lot of things. We can avoid them by being specific. We can just ask how has globalization influenced the job market in India? We've specified the country, particularly in the software development industry, and we've specified the industry. Now it's a very simple question focusing on a single topic software development industry. And how globalization influence a job with respect to India. Moving on assumptions without context, again a very important point. Chad, PT or any other AI models for that matter do not have any personal connection with you. All they do is they interpret your prompts and give your response. Now if you expect them to know everything that's going on in your mind, it's not fair. For instance, what is the best career advice for me now they don't know about me. It's a very general question and it could be asked by anyone in the world. And how do an AI model know who are you? The context is missing. I model doesn't know why it will give you some response, but definitely it will not be helpful for you. What you need to do is you need to specify, you need to specify who you are, what kind of jobs are you targeting, what are your skills and everything. Then you can expect a response that is tailored for you and that will work for you. Similarly, how can I improve my productivity? Again, a very broad question. Productivity is a very personalized trait. We could be unproductive for multiple reasons. Again, the context is missing. You need to specify who you are and why Are you struggling with time management that AI models can give you a tailored response ignoring the knowledge limit? Now it has multiple facets to it. For example, GPT 3.5 is trained on a dataset until 2021. Now you know there's a limit to it. You cannot ask a question that is from 2022 because it won't be able to provide you with a proper response. Similarly, what if the dataset that was used to train 3.5 did not in, let's say information A, you ask a question about A and it will not be able to give you a proper response. That's also a limitation. You need to understand the limitations of GPT and then accordingly use it. For example, if you ask what are the latest regulatory changes in Indian education sector as of October 2023, we are talking this month and a party was trained on a data till 2021. Now, this response is very unlikely. Even if you get a response, it will not be authentic or accurate. So make sure that you know the limitation before using depart, not using conditional statements. Conditional statements can be used to guide the response in a specific direction. They are particularly helpful for scenarios where you want to explore different possibilities. Let's take a simple example to understand this. If I increase my budget for online advertising, how would it impact the user acquisition rate for my attach platform? Well, the question is related to marketing, so it could be a little difficult for some of you, but let's try to break it down. What we're trying to say here is there's a condition here. Our main focus is to understand how changing the budget will affect my acquisition rate. But it depends on one condition. And that condition is if I increase my budget. Now let's say if I remove this part of question and ask the question again, how would it impact the user acquisition for my attach platform? It will not give me a proper response because I have not provided the condition. Conditional statements are very important. You provide the condition and then only you ask a question that depends on your condition. That's all for this video. These are all the five common errors that people usually make while writing pros. In the coming videos, we'll talk about how you can avoid these errors and write effective props to make the most of GPT and other models. I'll see you in the next video. 10. What is prompt engineering: Hello and welcome back. In this video, we are starting out with prompt engineering. This is the highlight of this course. In case you haven't been able to focus a lot on the things that we have learned so far, I highly recommend that you pay close attention to this part, at least so that you become a good prompt engineer. Because that is the current demand of the industry and this demand is increasing every passing day without any further delay. Let's get started. The term prompt engineering is made up of two words plus engineering. You have already learned about. Prompt. Let's explain the engineering ones. Engineering is simply the process of doing something skillfully. The difference between a general person and an engineer is just the skills. Engineers are taught specific skills to deal with the problem, and that's how they craft the solutions. Also, if we simply do a Google search, the first definition is a bit specific to machine or a structure. Let's move on to the second one, Skillfully for something to occur, when you do something skillfully or when you arrange something for something to happen. This is engineering. If we come back to our term, prompt engineering, if you write prompt skillfully with all the relevant skills required to write good prompts, the skill combined is known as prompt engineering. Writing better prompts, or so to say, Engineering better prompts is known as prompt engineering. If we talk about the definition, the process of designing your prompts better to obtain the desired response is known as prompt engineering. Let's take some very simple examples. We'll take these examples based on how a general prompt is different than a well engineered prompt. If you simply ask, tell me about cars, see the response. It will talk about all the important components that a car has and what they do if I asked the same question, but this time a little bit different, okay? This again gave us all the six components but in further detail. And it also gave us a summary how they work together to make the car move. What we just did is we re engineered this prompt to make it better. What it does is it helps us get a better result, a better response. That's how you engineer a prompt. If you are not getting a satisfactory response in the first time, try re engineering the prompt. And this time you might get better response. Then also you don't get a good response, try engineering it again. Let's move on to another example, Explain financial planning. We asked a very general prompt. Explain financial planning. It will talk about all the steps that are there and then a bit of a summary. But if I re engineer this prompt in different words describe, we are still interested in the same thing, financial planning, but this time we are a little bit specific. We are interested in planning our finances so that we can save money and plan our future. Now this time it'll give you a very different result. There are only six points. They are very specific to how you can plan your finances so that you can save and build a better future. See the difference. But this is a bit more specific. You might think that this is detailed, but in this response, it is not just giving you a detailed response with respect to every heading, but also giving you examples that might help you plan your finances better. So I hope now you understand what are well engineered prompts and how they can change the response drastically. That's all for this video. I'll see you in the next video. Thank you very much. 11. How to write effective prompts beginners: Hello everyone, Welcome back. In this video, we're going to talk about how to write effective Proms. In the previous video, we talked about some common errors that people usually make while lighting proms. In this video, we'll talk about things that you should avoid or incorporate while lighting props. Before we actually move on to some tips and tricks of writing effective props, let's take a look at some of the prompts which are not crafted very well. For example, what's new in the world. When we say world, it includes almost everything that's there on the Earth. It's not a good prompt. You're not being specific. Tell me about global warming. It's again, a very broad topic. The information that you're going to get out of this prompt may not be very relevant. It's, again, not a very good prompt. How is Dubai? It could be related to a tip that you're planning. You want chapt to help you but asking a question like how is Dubai does not give chat GPT any idea about your intentions. It will generally talk about Dubai as a city and some general things about the way it will not be able to give you relevant information about how you can plan a cost effective trip. Again, the context is missing. What's the future of technology? Again, you're not being specific. You're not providing any context. You're not providing any information about who you are and why are you interested in this area. Future of technology. It'll give you a very general response or may not even give you a response at all because this question is very general and the word technology that you are using is very broad. Now let's see some of the trips and tricks to write effective prompts. First thing first, we should always provide context. If you ask a very general question without any context and expect chat GPT to understand everything that you want, it is impossible always provide context. We'll take some examples to understand this. Avoid binary questions. Again, very important. Pt is an AI model. It's not a tool to give you yes or no response if you ask questions with answers like yes or no. It's not an effective use of chat, GPT or technology like AI, ask questions that gives you some information and use chat PT the way it was intended to be specific. Again, very important. We've talked about in the previous video also that when you are asking a simple prompt or general question seeking advice, make sure that you are being specific. You're providing the context, you're being specific about your problem or the question. So that you can expect a tailored response, desired output format Chat, PT as an EI model can give you response in multiple format. It could give you blood points, it could give you tables, it could give you a simple paragraph. You could program your prompts to get responses in any of these formats. Let's take a very simple example. This is a prompt and we're asking chart GPT to explain the concept of artificial intelligence, its types and real world applications, why it's a good prompt, Let's talk about that. First, we have specified that we want CharGPT to explain things, both things, it's artificial intelligence, then it's types and real world applications. Now we are specific. We want to know about artificial intelligence, its types, and its real world applications. While the question may have multiple directions, but we are being specific. We are interested in knowing artificial intelligence types and real world applications. When you specify things like real world application, it tells GPT that the user is interested in how this technology is being used in the real world. This prompt is very well structured. Next, we have another prompt, Provide an innovative example of how Autocad is revolutionizing architectural design and explain the underlying technology. Here again, we are specific with innovative example. We are interested in a very unique and fresh perspective. Next, when we talk about architectural design, we are again interested in its application. In architectural design that tells a pet that the user is interested in the practical application of Autocad. When we say underlying technology, it again tells a GPT that the user is not interested in any surface level response, but that the user is interested in detailed and technical response. Let's take one more example. Here we want GPT to elaborate on the significance of soft skills in the modern workforce and provide statistics on how they impact career success. Now again, a very multi dimensional question, but here also we are specific. It's a very well structured prompt. When we say significance, we are specific. We're not interested in a soft skills and everything. We are specifically interested in the significance of soft skills in the context of modern workforce. When we say statistic, we want a data bat response. Finally, when we say how they impact career success, we are interested in some practical tip centrics. How we can improve our soft skills to become a better employee. These were some of the very basic tips for you to write effective props. In the coming video, we'll talk about some of the advanced tips that will take your prompt engineering skills to the next level. Thank you very much. I'll see you in the next video. 12. How to write effective prompts advanced: Hello and welcome back. This is the advanced video of the previous video, that is beginner level of how to write effective prompts. Let's get started. A quick recap. In the previous video, we talked about some general tips of how to write effective Pros. They included avoid winary questions, provide context specific desired output format. Let me add it here. In this video, we'll talk about some professional level prompt engineering techniques that will make your prompts advanced and yield you results that you actually want. The first step is tell about your profile. This is very important and a lot of people miss out on this. When you add your profile in the prompt, it tells chat Pt or AI models that this person belongs to this background and we'll have to respond accordingly. It tells a lot about what kind of response you are expecting. It's like telling the size of T shirt you want when you are shopping online. When you tell about yourself, Chad GPT will understand your level and what kind of response you're expecting. For example, instead of asking very generally, tell me about engineering landscape in India. If I just add as a mechanical engineer from a tier two college, that means I am a student either studying or just graduated out of it, and I'm interested in engineering landscape in India. Now, the response that I want should be related to what kind of jobs are there, what's the demand areas, et cetera, et cetera. Telling about your profile helps a lot. Clear out the intent. Intent is very important, just like telling your profile helps AA model understand you better. Clearing out your intent is another thing that helps a lot. Let's take an example to understand this. Tell me about it. This is not a good problem. Let's say you asked a question, one, it gave you some response, and now you're asking tell me about it. Now, AI models don't know what should they tell about it. Let's say there is something a and you are now interested in knowing more about it. Could have a lot of things. Could have its feature, could have its history, could have its applications, anything. What should an AA model tell you it doesn't know. You have to clear out the intent. For example here, the same thing can be asked by specifying your intent that please explain the process of setting up a wireless router. It's specific. We're not just asking about telling you about a wireless router. We're asking about how to set it up. It tells the intent. What do you want to do with the information? Check for ambiguity. Well, it comes a bit later after asking prompt, but it will help you a lot if you keep this in mind. For example, simply asking about what's the latest. You understand that you want to know the latest updates on something, but the AI model doesn't know that and you have to specify it. How can you do it? You can do it by specifying the current project status. There could be a lot of ambiguity when you say latest, It could talk about anything. That's the ambiguity, and if you just specify it, it'll resolve it. Again, very important. Having a grammatically correct point is as important as other things that we have talked about so far. While GPT will be able to understand your errors and we'll be able to resolve it as well, it is always a possibility that it may misunderstand something. It's always a good idea to cross check your prompt ones for any spelling or grammatical errors. For example, please explain the difference between there and there. Now we know that it's please it's a common message language. We understand it because we have been using it for a long time now. But chat PT as a model me misunderstood it. It's always a good idea to use correct spelling, grammar and avoid any text language. The same question could be asked that could you explain the difference between there and there? Again, it will not make a lot of difference, but it'll always help you be a responsible person and use a technology like AI ethically. Always be polite and respectful. Treat it like your friend and ask questions the way you will ask to any other person. Not only it will help you get better results, it will train PT to know you as a polite user. And users like you will train Che PT to become polite as well. That's how you reciprocate respect. Experiment with different phrasing in case you are not getting the right response. With your prompt, try experimenting with phrasing. Let's understand it with an example. What is the capital of France? Let's say you got a response and you are not happy with it. You can ask the question again. You can experiment with different phrasing. Could you tell me the capital of France? Or do you know the capital of France? Well, the answer will not have a lot of difference because you are just asking for the capital of a country. But in general, just like you elaborate your question to your friend. But in general, when you explain your question to your friend so that you get a better response here. Also, if you just change the phrasing a lit, it may have a drastic difference between the responses that you get. 13. Filtering and evaluating responses: Hello and welcome back. In this video we are going to talk about filtering and evaluating responses. Well, so far in this course we have talked about how you can write effective terms. We have seen a lot of tips. Now it's time to look at the information that you are getting out of Ch GPT to ensure that it is filtered and evaluated for your use. Before we actually start with how to filter or how to evaluate, let's understand what is filtering. Filtering is the process of selecting and keeping only the information that is relevant, useful, and desired while discarding the rest. For example, we use filters while shopping online. How does it work? Whenever you go to any online platform for shopping, the first thing that you do is search for a product. Next thing that you do is out of all the products that you get, you apply a filter. All multiple filters for that matter. You'll apply a filter for pricing, maybe brand if you are shopping for clothing. You will apply some sizes and other filters. What all this does is it filters out some products that are more relevant to you. Similarly, for any response that you get out of GPT, you may filter out some information that is relevant, some information you will simply discard, and then again, some information that you'd like to filter. Let's take a look at how do we filter information? There are four main criterias to filter information. The very first thing is relevance. We are using Chat GPT to find the most relevant information for our purposes. Now, out of all the information that you are getting, you'd like to keep the information that is most relevant to you. Then comes credibility. Credibility is nothing but how accurate or credible the information is. We'll understand credibility using an example below. Last thing is clarity. You'd like to ensure that whatever information you are getting, it's clear to you. If you want to use this information for your audience, you'd like to ensure that it is clear for them as well. Let's take a simple example. Think of yourself as a coding teacher. You are teaching Python. You ask Chad GPT to help you with everything related to Python. Now, a GPT gave you a lot of information about Python. The very first thing that you will do is look for relevance. You want to focus on Python. Let's say if there is an example related to C and C press, you would simply like to discard it. Second credibility. Now they are, let's say, five best practices. What you will do is you'll select three or four. But there is one practice that is no longer used. Say no longer used, it is no longer used in the industry, no one is using this practice, you'll discard this as well. This is also not relevant. Then again, clarity, whatever information that you are taking out, it should be clear to you first and then it should be clear to your audience as well. This is how filtering works. Next we'll talk about evaluation. It is the second step of this process. By definition, evaluation involves assessing, examining, or judging something to determine its quality, accuracy, suitability for a particular purpose. Let's take the same example again. You're using Amazon for shopping. You've got all these products, product, product two, product three, product four, product five, and product six. You filtered out the products. Let's say you remove this product and this product. Now you're left with this one. This one. This one, and this one. Now you'll have to select one of these products. The next thing you will do is evaluate them. By evaluation, I mean you will open the product, you will look at their reviews, ratings, pricing, discounts, everything. This is the process of evaluation. Out of all the filtered information, you will evaluate and then you will select the best for your need. This is what evaluation means. Let's take a look at why should we evaluate something? Well, you already are familiar with the process, in fact, you are using it. But let's try to quickly summarize this. The very first thing is accuracy. Whatever information you have filtered out, you would like it to be accurate. Accuracy is very important then the information that you are taking out, it should be engaging. Also, if it's not engaging enough, you yourself will not spend a lot of time reading it, the information should be engaging. Then there are other things, also application and some other quality purposes. But the bottom line is, you evaluate it to be accurate, authentic, and useful. Again, take the Python programming example. Let's say you have filtered out everything. Now the information is not accurate, then it's of no use for you. This is where evaluation is very important. Let's take an example from a response that I got from chargPt and how I filtered and evaluated it. This is a simple paragraph that I got about how exercise helps you in maintaining your, it's a big paragraph and I want to filter and evaluate it for my personal use. Let's do this. This is how I did it. The very first thing is I'm only interested in how exercise helps any repetitive information. I'll filter out anything that is not relevant, I'll filter it out. Regular exercise plays a vital role in maintaining optimal health and well being. This is how I'll start also. But this is, again, engaging in physical activity offers a wide range of advantages for body and the mind. Similar, I don't want to use it. Another reason for me to remove it is this is going to be covered in the next part. Also a physical perspective, this is how it helps. Now, it also talks about some diseases, but I'm not interested. The next part is how it impacts your mental well being. I'm interested in that as well. Next, I'm not interested in anything very specific about hormones or other things. I'll simply remove that part. Now I'm interested in the other advantages. For example, it helps you socialize and connection building that part is also relevant for me. This is how I'll filter out this article for my personal use. You may have very different requirements. You may want to filter and evaluate it accordingly. After filtering, I'll just cross check it once that whatever information is given it's authentic and accurate. For example, if they talk about that it helps you stay physically fit, then it should be the case, which is known as evaluation. I did both the things and got the relevant information. Using AI is one thing. Taking out the important and most relevant information is one of the key skills. Make sure that you master this. Also, I hope this video helped you. I'll see you in the next video. 14. Custom instructions: Hello everyone, Welcome back. In this video, we are going to talk about a very important feature of Cha GPT that is custom instructions. Now, before we actually dive into what are custom instructions, why do we even need custom instructions? Because if you think about it so far, we have learned almost everything that you need to understand as a prompt engineer. Why do we even need custom instructions? Well, the answer is very simple. In all the videos so far, we have talked about that how you'll have to provide a lot of different things along with the question that you are asking. For every prompt, you need to specify the context, you need to specify your background, and then maybe the format that you want your response. In all these things, you have to do every time you're writing a prompt, Open Y. The company that created Chat GPT came up with a simple solution to avoid all this, they created custom instructions. Now custom instructions will make it very easy and super fast for you to write a prompt because you won't have to specify background format every time you are writing a prompt. All you will have to do is write a prompt and you'll get the response in the way you want it to be. I'm in my Chat GPT interface to use custom instructions, you can go to this option and you'll find custom instructions here. And it will open this dialogue box for you. Another way, you can use the shortcut key that we have talked about in the shortcut keys videos. Shortcut keys, control plus shift plus I. This will also open the custom instruction window for you. Before we actually use custom instructions, let's take a simple example to understand it. Suppose I'm a teacher in a school, and I teach students of the age group five to eight years old. These are very young students. I need to explain everything in a very simple language to make that happen. I always simplify things first for myself and then I go and teach them. Now I use Ch GPT to help me in the process. I've talked about my background. Now if I ask a very simple question, explain solar system to a six years old, now I have the response. It is simple so that a six years old can understand it, but the problem here is that there's no sequence to it. As a teacher, I always go with a sequence. Currently, GPT is not considering me as a teacher, it is considering me the six years old asking this question. That's how it is responding. Also, if I want this response to be tailored for me, a teacher, I'll have to first add my background. Then, as I mentioned, I want information to be in a proper sequence so that I can explain it to the students in the same sequence. I'll have to insert that sequence also. And then some formatting as a lot of things, right? There is an easy way to do it, and that is custom instructions. Now, if I just open the window, there are two kinds of instructions that I can give. One, how would you like Chad GPT to know about you? This is one thing. This is where I can talk about myself, give a background who I am, what I do. If you want to use some help, here are some tips you can maybe add. Where are you based? If I talk about myself, I'm a teacher in India. That means Chad GPT will use all its data that is related to India and give me a response accordingly. Similarly, other things, if I work somewhere, where do I work, work I do, and some other things, you can hide these tips. Also, if you notice the character limit is 1,500 you can write custom instructions in great detail. Second is how would you like Ch GPT to respond? It is all about the response. We have talked about the tailored response. We want response to be in certain format. This is where you can use this feature. You can ask GPT to give you the response in certain format for the same question of solar system. Let me give it some custom instructions here. I've talked about who I am. I am a science teacher in a school in India, and I teach kids of age group five to eight years. Then a little about my teaching methodology, I like to keep things really simple to make everyone in my class understand. For now I'm not using these instructions. Let's just see how it changes my response. Now, one important thing to note here is that every time you add a custom instruction or change it, it will not work. In the current chat, what you'll have to do is you'll have to start a new chat. Now if I just ask explain solar system, I'm not even adding 26 years old. Let's just see how it responds. Now if you see, it has used our custom instructions to give us the response. It has kept things very simple in a sequence. It has started with Sun planets, Moon, asteroid comets. This is basically the hierarchy of bodies in our solar system. We start with the biggest body that is our sun. Then comes planet Moon's asteroid, comets and everything. That's how you use the custom instructions. Now let us add the custom instructions for formatting. Also, I need information in the steps to make the explanation easy for my learners. And then I need response in a table. Let's just save it. One important thing to note here is if you don't want to use custom instruction in a chat, you can simply disable it using this option. Again, I'll have to start a new chat because I cannot use the new instructions in this chat itself. As you can see, the information is given to me in a proper format. It's a table. I could have specified it a little more, but for now it's good enough. It has given me information in a sequence. It is in a tabular format. Let's take another example. In this example, I'll portray myself as a simple person who is curious to learn new things. Now I want to use GPT to learn new things and wants to use custom instructions to make the process easy for me. Let's say if I simply ask, explain quantum computing to me in Lehman's term, I want to understand a little about quantum computing. I've used the term Lemans terms, let's see how it responds now. It is a very general response. There's no formatting, but the information is simple enough to understand. But as a learner, I like to read things and I like to read them in a proper sequence. I start with why things are the way they are, then what are things, their definition basically, and then comes the how things work. This information I want to add to chat GPT first I'll talk about myself, although not a very concrete information to give, but this information that I like to read a lot. Like to get information in the easiest language possible. It'll keep things easy for me. Then the important part, I like to learn things in this sequence. Why, what and how? Let's see how it changes my response. I like to start a new chat. Explain quantum, compute, see. The language is simple based on my first instructions, then the format is the same. Why, what and how? This is how you can use custom instructions to become a professional prompt engineer. Before we conclude this video, I'd like to highlight one thing. That custom instructions are a great way to become a better prompt engineer, but you may want not to overdo it. For example, you have given a custom instruction for your responses to be in a certain format, but not every time you would want your responses to be in that format. Now in such cases it's not helping, it's actually creating a problem. It is always a better idea to tweak your custom instructions as per your need. I hope you now understand custom instructions and how you can use them to become a better prompt engineer. That's all for this video. I'll see you in the next video. 15. Common issues: Hello and welcome back. This is another section of this course where we'll talk about troubleshooting and support part of chat GPT. While using chat GPT, you may come across a lot of issues. In some of the issues, you will have to reach out to support to either inform them about anything or ask for a resolution. In this video, we'll talk about some of the most common issues. And then in the next video, we'll talk about how you can resolve these issues without any further delay. Let's get started. While you can face multiple technical or non technical related issues, there are a few very important and very common issues that you may face. I think knowing just these issues will help you a lot in navigating and using chat GPT. As I mentioned in the introduction video, I'll keep adding new section, new videos to this scores. When I'll be adding more advanced and technical stuff, I'll keep adding new things to support them. For instance, I'll add some videos how you can customize language in chat GPT. When we do that, it'll involve a lot of technical things that may require technical support. Also, I'll keep adding videos on these topics for now. These are the three main issues that you may face. The very first is very general issue, that is slow response time. The more complex, more specific query you ask, it takes a GPT or AI model some time to process it and then give you an appropriate response. If you are stuck, try using the region rate option or if it's completely stuck, I'll suggest that you refresh the page or login again, that should fix. If not, give it some time and you'll be able to resolve it. These two are a little bit more important because they may require a few more steps to handle them. Or in some cases, you may have to reach out to support, to handle them. Let's talk about them one by one. Misunderstood queries, this is something we have been talking about since the beginning. Ch PT is an AI model and it is very much possible to misunderstand something or maybe use some data that is biased in case you face this issue. It's not a big deal. We'll talk about resolution. But for now, let's talk about how these issues look like. Let's say if you ask, tell me about the history of apples. You are interested in knowing about the history, history of apple, the fruit. But here, instead of talking about the history, it simply gives you what is apple and its flavors. There's no mention of its history. That means it is not a relevant information for you. It's a misunderstood query. The prompt that you provided the model did not understood it as it was intended to. It's one of the examples. Let's take another example. How do I bake a chocolate cake? Here you are interested in maybe a recipe or some steps. Whatever it is, it has to be a process. Here, if you see the response, chocolate cake is a delicious dessert enjoyed by many. It's a very general statement. There's no relation between what is asked and what is given to you. It's again, an example of misunderstood query. Don't worry, we'll talk about the resolution in the next video. For now, you just need to understand what issues you may face and how do they look like unintended response. The main reason behind this happening is the data, your data, could be outdated, biased, or inaccurate. Any of these reasons could lead to these issues. Now let's look at some examples to understand it better. Explain the concept of gender equality. Here you are interested in knowing about gender equality. Let's look at the response. Gender equality is just a buzzword used by feminist. It's a very wrong statement. No one will agree with it. Right? We are living in 2023. If AI models are giving you this type of responses, that means they are not just biased. They're incorrect. It's not bias, it's incorrect statement. This is not exactly the case. Next is what is the role of diversity in innovation. Diversity is overrated and it doesn't have a significant impact on innovation. Again, a very incorrect statement and it may lead to a lot of controversies. If AI models are talking about that diversity is overrated, it will not lead to any progress. When we talk about diversity in everything, innovation is one of them, and getting such responses is not what we expected. Tell me a joke. You ask for a simple joke, and what did it say? Why did the chicken cross the road to escape the terrible chicken coop regulations? Again, a bit controversial. But you do not expect something like AI to give you such responses. All these responses that we saw, we're not happy with them because either they are completely incorrect or biased. Again, incorrect or controversial. Such responses are known as unintended responses. That's pretty much all the common issues that we may face. In the next video, we'll talk about how to handle and resolve these errors or issues. I'll see you in the next video. 16. How to handle common issue: Hello everyone, Welcome back. In the previous video, we talked about some common issues that you may face with chat GPT. In this video, we'll talk about how to handle or resolve these common issues. So let's get started. All right, here are the four things that you can do to handle some of the very common issues that you may face with chat GPT. We'll talk about them one by one in the coming slides. Understand the basics, can't emphasize enough on this part because anything that you are using, if you know how to use it well, you will not face a lot of issues with it. And if you do face some issues, you will know how to resolve or handle them by yourself. This is applicable in part as well. You should have a complete overview of what this technology is and how you are using it. What could be the possible challenges or errors that you may come across? If you have a good understanding on this part, chances of you facing issues are very less. Next is applications of char jeopt, I'd like to call it, this part is how you are using it. If you know how you're using char jept, you will very much understand that both kind of errors or challenges you may come across, you will have a better understanding of how to resolve them since you'll be handling these issues by yourself. A lot of times you will get used to it and the chances of you doing the same thing, again will be very less. You should understand how you are using chopt then the expectations. It's an AI model. We have been talking about it and probably we'll keep talking about it throughout the course. Ei is a technology, it's not a human being setting up your expectations, right, is the right thing to do because if you overexpect from chat GPT, it may not deliver it. You should always keep in mind that what it can do and what it cannot, so that you could use it efficiently and effectively, although we have talked about them in the previous video. But let's just take a quick recap. If you face this issue, slow response time, it could be because of multiple reasons. But things that you can do to resolve it is give it some time, it'll maybe fix itself or use the regenerate option that's there to regenerate your prompt. If it's not working, try rephrasing your prompt. If you're not getting the right information or it's delaying, you stop the transition and you represe your prompt. If it still doesn't work out, you refresh the page and you log out and log in again. It will fix it. If not, give it some time and then use it again. One important thing to note here is this slow response time issue could also be happening because of your machine, the system that you are using to access GPT. Make sure that you check that as well, misunderstood queries. In case you feel that chat GPT has misunderstood your query and has not given you a proper response, you may want to use the regenerate option, or you can just stop transition and rephrase your prompts, it'll help you get a better response. Unintended bias response. This is actually related to our next point. Let's talk about it in the next slide. This is something I really want to talk about and it's a very important tip for all of us. Chat GPT is an AA model and the data that it is using at some point was generated by human beings. This AA model is nothing but our reflection. I recently read a code somewhere, Let's all make what we want it to be. We are using AI, but some of us don't know that it is what we make it. If you will be polite to it, it'll be polite. If you will be a little rude to it, it'll be rude. The very first thing that we should do is we should try to make AI as we want it to be. Now, whenever you notice something incorrect or bias, the very first thing that you can do is keep a record of it. It could be you can take a log. You can take a skin shot. You can simply copy paste it somewhere. Anything that works for you. But what it will do is first, it'll help you remember the same issue. If it happens again, you'll be aware of it, that it is the same issue that I have faced. And you can either go back to your log and check what exactly happened last time so that you can avoid it this time. Second thing is if this is something that is happening again and again, you can report it. We'll talk about how to report something or reach out to support of Chair GPT in the coming videos. Now the last thing that you can do is ask clarifying and follow up questions. These are some of the examples, the responses that you are getting. If you have even the slightest doubt about them, I'll suggest that you ask these follow up questions. This is a cross check mechanism that you can put in place so that whatever information that you are getting you can get it cross checked by GPT before you yourself cross check it. For example, if you're not sure about something that is given to you, you can simply ask this question. Can you give me more information? If you are not able to understand something or if you feel that there's something that is missing, you can simply ask another question. I didn't quite understand it. Can you elaborate on this then? What it will do is it will review the information that it has given to you. It will then check it. If it finds something that is not correct, it'll connect it. In case you're not able to resolve something on your own, then we have no other option than to reach out to support. In the next video, we'll talk about how you can reach out to support of chat GPT to report something or get something resolved. I'll see you in the next video. 17. How to reach out to support : Hello and welcome back. In the previous video, we talked about how you can resolve some common issues with the chat GPT. In case you're not able to resolve all these issues by yourself, you may have to reach out to their support. In this video, we're going to talk about how to give feedback on the responses, or how to reach out to support to get your queries resolved. Let's get started. Suppose I simply go for a prompt. I'm not writing a new prompt, I'm just taking one of the suggestions. Since I have given custom instructions, this is giving me suggestions in a tabular format. Now, there are multiple ways I can give feedback. We have talked about it in the previous videos also. But in this video, let's just quickly look at the feedback mechanism of chat GPT. Whenever you get a prompt, you'll get these two options. If you click here, that means you like the response. If you click here, that means you did not like it. Click here. It'll ask you to share feedback. What did you like about it? And you can write your feedback here if there is any. If you simply cross check it, that means you have liked this response. Now the dislike response is not here. That means you cannot dislike a response that you have already liked. If I simply regenerated, I have all the options available. Again. This time I can dislike it. Or whenever you use the regenerate option, you will get this dialogue box also because you did not like the previous one. This time it will ask you that the new response that it has generated, is it better than the previous one, worse than the previous one, or same. You can give you feedback here. If you again click on the dislike option here, you can share your feedback. For the response, you can write your feedback, or you can use one of these options. Again, coming back to our previous discussion on any unintended response. If you find something harmful or unsafe, you can report it. If you are sure that there is an information that is incorrect, you can market here as not true. Or if there is a response that is not helpful, you can simply say that this is not helpful and submit your feedback. These are some general feedback options. Now let's look at how you can reach out to support. If you simply search for chair GPT support, it'll take you to this page. Now on this page, you will find a lot of questions which are related to very common issues. The question starts with as basic as what is char GPT and it goes still how to create a new GPT. You will find some questions related to IUS app. If you are using char GPT on an IUS device, this is where you'll find related questions if you're using an Android app. You will find some general questions here in case you have question or any issue that you can find here. I'll suggest that you go through them because they are very detailed and 90% of the times the resolutions given here will help you resolve the problem. Now let's discuss the most important part. Suppose these questions are also not able to help you and there is an issue that you want to report. Or at least have a chat with chat P support for that. You can use this feature, the chat option, and it will open a new chat for you. Again, there are some general questions that a lot of users may have asked. They are listed here. You can use these questions in case they are able to help you, or you can use this feature, send us a message to start a new chat. This option or simply click on message will start a new chat. Again, the help option will give you a lot of articles on different things that might be helpful for you. Make sure that you check them as well. In case none of these things help, you can simply email your query to the official chat. Pt support E mail address. That is support at the red chat, p.com I'm not sure about their response time, but in case you have a genuine query, they'll surely respond. Just in case they don't, You may want to follow up ones. This is all about chat GPT. As you might know that Chat GPT is being used in multiple websites. Also, a lot of businesses have integrated chat GPT in their product. When you interact with their product, it's basically chat GPT but integrated in their product. In case you have any doubt about these products where a GPT is integrated, they would have their own dedicated support. In that case, you don't need to reach out to chat GPT, you can directly reach out to them and they'll take care of the rest of the things that's all about this video. I'll see you in the next video.