Prompt Engineering for Everyone with chatGPT and GPT-4

1. Course Introduction: Will your job be eventually taken over by an AI? Yes, most probably. Will it happen in the near future? Probably not. Then maybe you're asking yourself, why should I bother learning a new skill? If everyone will use AI in a few years, what difference could it make to give you a straight answer, if you learn from engineering in a few years from now, it won't make any difference at that point. Why? Because it's probably going to become a mandatory scale for everyone to have. So it would simply be just another drop in the ocean. There would be nothing special about it in the future. Some people choose to believe that using an AR model requires nothing but a few simple sentences. And somehow, magically, the future AI models will become so smart that they'll be able to read our minds and understand what we want. I think these expectations are a bit unrealistic. When we interact with modern AI models to solve problems or generate ideas, we use plain natural language in our inputs, but it's like writing. Although you may be able to sketch some simple ideas or write a certain number of pages every day. That doesn't make you a good writer. Think about it. Averaged skills cannot produce more than average results. Have a look at some of the free modules in my course. I think you'll understand pretty fast that prompt engineering and artificial intelligence in general are not just another hype. The good news is that we're still very early. Now it's the perfect time to sharpen your skills and be prepared for the future by going beyond the average. Whether we like it or not, AI is bound to change the way we work and live in ways that we can not yet imagine. So stop missing out on opportunities. The moment you'll begin studying prompt engineering, you will immediately start improving your productivity. The tools are already available. Some of them are even free to use. The concepts that I'm teaching in this course are not specific to a particular AI. Although I make a lot of references to chat GPT or GPT for the skills that you'll learn in my class are actually applicable when interacting with any model in general. Whether you want to solve problems, summarize data, come up with ideas or create stories, images, or any other type of generated content. The skills that you are building in this course are going to be very useful. And they are going to be useful not just now or tomorrow or in the near future, but in the longer perspective. Although the models will evolve and the technology will certainly get better in the future, the core principles will remain applicable. I've optimized this course to give you the most important information in the shortest possible amount of time. Because I know your time is precious. There is a clear learning structure in place. We'll start with the basics and gradually move on to more advanced topics. There'll be lots of examples and quizzes also. You'll be able to verify your understanding of the subject. Don't have access yet to check GPD or GLUT4, Not a problem. You get free access to chat, GPT-3, what digital playground that I've built. It's like having a dedicated laboratory where you can immediately does the concepts that you learned need more examples for real-life scenarios? I've got you covered. There are more than 250 prompt examples that you can download and use immediately. But compared to the rest of the course, to be honest, these are not that valuable anyway, I'm teaching you how to build prompts, so you won't really need those examples in the end. This is not a technical course. I won't be showing you how to build an AI or fine tune an existing model. I have created this material for any level of technical ability. So there are no technical prerequisites. If you're looking for more technical aspects of artificial intelligence, I'm working already on other courses, but this one is for everyone. In case you'd like to share anything about your learning experience or have any questions regarding the material. You can reach me on LinkedIn or Twitter and we can chat. Thank you for taking part in this. I hope you will enjoy the experience. 2. What is Prompt Engineering?: Alright, now that we've covered the introduction, Let's talk a bit about the main subject. What is prompt engineering? Simply put, prompt engineering is the art and science of designing effective inputs and fine tuning parameters for AI models to get the desired results. It's a mix of language skills, problem-solving, and applying logical thinking. In many ways, prompting is similar to writing code. It requires a similar mindset and approach. You have to understand the problem and be able to clearly define it and break it down in individual tasks. Prompts play a crucial role in working with AI models. They are like the steering wheel and the gas pedal of a car that helps you steer and control the AI's capabilities and guide it towards the right destination. With a well-crafted prompt, you can unlock the true potential of AI models and make them work for you in amazing ways. Think about it in a matter of months. Now, most people will be using this technology like chatbots and virtual assistance in everyday jobs and activities. But not everybody will have the same proficiency in doing so. Just like with any tool available out there, the average user will get average results. So before we dive deeper into prompt engineering, Let's briefly touch on some popular AI models you might come across today. The first example is GLUT4, which is the latest version of OpenAI is groundbreaking generative pretrain transformer, which excels at understanding and generating human-like texts. Meet journey. It's another example. This is a powerful model focused on creating images from prompts. Another one is stable diffusion. Also a generative AI system that is primarily used to generate detailed images based on textual descriptions. So basically, we are using prompts to create images, but it can also be applied to other image related tasks. And there are many other less popular generative AI models. As artificial intelligence research continues to evolve, will definitely see more cutting-edge models emerging in the near future. These models all have their unique strengths and use cases, but they all have one thing in common. At least for now. We use prompts to harness their capabilities effectively. Because there are so many topics to cover and so much knowledge to explore in this course, we will only focus on text-based generative AI systems. Prompt engineering can be applied to a wide variety of tasks, from writing emails to creating artwork, generating code, or even composing music. By understanding how to craft prompts and fine-tune AI model parameters, we can guide these powerful AI systems to assist us in our everyday tasks and professional activities. E.g. GLUT4 can be very useful in text-based tasks like summarization of large datasets, classification of data, text generation, language translation, conversation, or even rewriting or spell checking content. So why should you learn prompt engineering? As ai continues to shape our world, knowing how to work with AI models becomes increasingly valuable. Ai is rapid development has brought about some concerns, particularly when it comes to jobs and skills. Automation and AI in general have the potential to replace certain job roles and tasks for sure, which can be both exciting but also unsettling. On one hand, AI can free us from repetitive tasks and allow us to focus on creative and strategic work. But on the other hand, it also has the potential to make some jobs obsolete or requires us to adapt our skills to remain relevant. It's not really about the product itself, whether that chat GPT, GPT-3 for Google Bard, or any other product. Artificial intelligence is here to stay. The recent open letter signed by 1,000 researchers where they advise for a break in the training of ai models. I think that's the best proof that AI should be taken seriously. By mastering prompt engineering, you'll be able to enhance your productivity and believe it or not, even your creativity. Discover new ways to tackle complex problems, adapt to the changing job market and stay ahead of the curve, and also develop a valuable skill set that can be applied in various professional settings. Remember, this course is all about making prompt engineering accessible to everyone regardless of your background or technical expertise. I'm going to use examples and real-life scenarios to show you how to solve real life problems using AI. In short, learning prompt engineering is an investment in yourself and your ability to thrive in the age of AI. In conclusion, we've covered what prompt engineering is and why it is important in the age of AI. Now, it's time to talk a bit about the resources and additional tools that you'll find in the contents of this course. See you in the next module. 3. A short talk about the additional resources of the course: Welcome back everyone. Before we start exploring the basics of prompt engineering, I want to take a short moment to tell you more about the additional resources and tools that I have included in this course. Apart from the videos that I hope you will enjoy watching, that are two very important tools in this course that you should consider using. The first is the testing playground, which is a free interface where you can chat with chat GPT and experiment with your prompts. Here is the link to the playground. And you can also find it in the resources attached to the last lecture of the course. The other item I wanted to tell you about, it's a collection of prompts that I've built and which can be very useful as a starting point. This can also be found in the resources attached to the last lecture in the last section. A few words on that. By the way, you should consider the prompt in that collection as just a starting point, as I said, on which you can build more complex ones. Solving an actual real life problem using AI can readily be accomplished just by using a simple prompt. So you'll probably have to tinker a bit, experiment and refine them until you get the right result. Be creative and try to define the problem first. Okay? That's what I wanted to share about the additional resources in the future. I'll be adding more material in that section. So take a look at it once in awhile. Now it's time to dive deeper into the nuts and bolts of crafting effective prompts. In the next module, we'll explore the basics of prompt engineering and I'll share some best practices to help you get started. So let's begin our journey and get to work. 4. Important Definitions and Key Concepts: Hello everyone. This is module 1.4. I know I've promised to begin our talk about the basics of prompt engineering, but there's one more important topic to address. Until we go further to the more detailed chapters in this course, we need to clarify some basic vocabulary. Otherwise, you may not understand correctly the more advanced topics in this course. So I've prepared this lecture in which I will explain in simple words, some of the definitions that you should be familiar with. These are the terms that I'm going to explain. So let's begin. Large language models. A large language model is an AI model that understands and generate natural language by learning from huge amounts of data. They are very good at understanding context, detecting patterns, and making predictions. And in general, for almost any task involving natural language. Generative pre-training transformers, GPD. But GPD is a type of AI model designed for natural language processing tasks. Gpt is a specific type of a large language model. Actually, it basically creates an output by predicting the next word in a sentence, which enables it to generate human-like responses in a variety of contexts. And by the way, this is just the technical classification which is really not necessary, is not relevant for the average user. Generative AI. A genetically of ai is a specific category of AI tools and models like chat GPD that generates something like text, video or images, e.g. this is different from other types of AI systems that are designed to make decisions or create recommendations or categorized on data. They can be very useful in scenarios like content creation, design and problem-solving. A token. What is a token? Well, GPT models in particular, understand and think at a token level. A token is basically a chunk of texts that goes into the AI model to be processed. Depending on the model, this chunk can be a single character, a word, or a combination of words. What is AGI? Artificial general intelligence refers to an AI system that has the capability to understand, learn, and apply knowledge on a level comparable to human intelligence. It's the kind of AI that we usually see in the Hollywood movies. As far as I know, at the current moment I'm recording this video. Artificial general intelligence has not yet been achieved anywhere in the world. But the ray is to get that is faster than ever. Another important topic is ai alignment. Ai alignment is the process of making sure that the goals, values, and behavior of AI systems are aligned with those of humans. This is crucial for creating AI systems that are safe, trustworthy, and beneficial to society. Because we don't want to end up like in the movies. Reinforcement learning from human feedback is a technique used to train AI systems by providing them with feedback from humans. So it's actually a way of aligning them to our values. This feedback helps the AI system learn to make better decisions and adaptive behavior to better align with human values and goals. Chad, GPT and GLUT-4 are examples of AI models that were trained using this technique. So we can say it's a method that's used to achieve a better alignment of an AI. What about ai fine-tuning? Ai fine-tuning is a technique where a pre-trained model is further train on a smaller set of data specific to a particular domain. These tuning will improve the model's performance on the respective topic. It's like taking a person with standard education into a specialized course for a particular domain. Chat GPT. Chat GPT is an AI chat interface made by a company called OpenAI. Gpt is basically just a website based on an AI model called GPT 3.5, durable. Although during this course you might notice that I'm calling chat GPT an AI model. This is inaccurate but more convenient. So I apologize in advance for this intentional mistake. And finally, GPT for this is opening eyes latest and the most advanced large language model available inside the same interface called GPT. That's all still there. I hope that wasn't too much information to process. And don't worry, you can make a bookmark to this module and come back anytime when you might need to see some definitions. Again. That's it for this module. See you in the next lecture where we will start discussing about practical prompt engineering. 5. Understanding Prompts: Inputs, Outputs, and Parameters: Hello everyone and welcome to Module 2.1, where we're going to discuss the building blocks of prompt engineering, the inputs, the outputs and the parameters. Don't worry, they may sound technical, but you'll see that these concepts are pretty straightforward and easy to understand. So let's dive right in. First. Let's talk about the inputs. Inputs are the starting point of any prompt engineering journey. There are simply the task questions or information that you provide to the AI, helping it understand what you want. Think of it like a conversation you're asking the AI a question or giving it a task to perform, e.g. you might input the following text. Here. The input guides the AI in understanding that you want it to translate a specific sentence from English to French. So it's essential to provide the AI with well-defined inputs in order to get the desired results. Remember as a general rule, garbage in, garbage out. A mediocre input will probably result in a mediocre output. Keep that in mind. Next up, we have the outputs. These are the responses. The response is generated by the ai based on the input you've given. In our previous example, the output would be the translated sentence in French. Obviously, the quality of the output largely depends on the clarity and how specific the input is. Finally, let's discuss about parameters. The prompt parameters are the settings or knobs that you can tweak to customize the AI's behavior. You can think of them like dials on a studio where you can adjust the sound settings to get the perfect sound. There are some common terms related to prompt parameters that you should get familiar with, like temperature. The temperature controls the randomness or creativity of the AIs output. Higher values will make the AI more creative, while lower values will make it more focused and deterministic. Another one is maximum tokens usually sets the maximum length of the AI's response. If you want shorter responses, you can reduce the max tokens value. Don't worry about using these parameters though in most scenarios where you'll be interacting with an AI system, there are no settings in the interface itself that you can actually tweak. So basically you're fine tuning these parameters by the way, in which your phrasing your problems. So it's all in the prompt. Now, let's see an example of using parameters. Here is the input. This is going to be our prompt. As you can see, the prompt contains the task and the required parameters, and everything is in plain English. Parameters as creative as possible, limit the length to 50 words. So by tuning the parameters, we got creative and concise story about whiskers, the cat. Alright, now you should have a better understanding of the basic structure of a prompt, the inputs, the outputs, and parameters. In our next module, we will explore how to craft simple prompts using this fundamentals. Remember, prompt engineering. It's all about communication and experimentation. So don't be afraid to try new things and keep refining your prompts until you get the results that you're looking for. Happy prompting. 6. Crafting Simple Prompts: Techniques and Best Practices: Hey everyone and welcome to Module 2.2. In this section, we'll explore how to craft simple prompts, different techniques to increase effectiveness, and some best practices to follow. What makes a good prompt? Well, there are three key aspects to consider. First of all, is the clarity. The prompt should be clear and concise. This helps the AI understand your request more effectively. Think of it like trying to explain a concept to a complete stranger. As you don't know anything about that person, you have to use clear and concise wording to avoid any kind of confusion. The next important aspect is contexts. Provide enough context to guide the AI's response, but not too much to make it overwhelming. Just like in a conversation, too little contexts may lead to confusion and misunderstanding, but too much contexts may also confuse the audience. And then we have creativity encouraged to be creative and explore different solutions. Obviously, sometimes we need creative responses, but sometimes not. It largely depends on the problem we are trying to solve. So let's look at some examples to get a better understanding of these concepts. Imagine you need the AI to help you draft an email to a client. This is an example where probably you will need a bit of creativity in the response. Otherwise the output might look like more like a template. Instead of creating a certain connection. Instead of just asking, write an email to a client, which is simple but two general prompt for this problem, try to be more specific. You see this prompt provides clarity, contexts and encourages a creative response. Now, let's say you need a catchy social media posts for your bakery. Again, in this scenario, creativity is really important. Instead of asking, right, the social media post about my bakery, which again is very generic. Try something like this. You see this prompt sets the tone, provides context, and lets the AI's creative juices flow. Now that we've seen some examples, Let's talk about a few best practices for crafting simple prompts. Start with a clear action verb. This helps the AI understand your intent. Be specific about your desired outcome. This helps guide the AI's response. Experiment with different approaches. If the AI isn't generating the desired output, try rephrasing or providing more context. To wrap up this module, let's do a little practice activity. I want you to come up with a prompt for the AI to write a haiku about a rainy day. Remember the three key aspects, clarity, contexts, and creativity. You can practice and experiment with this little homework using the playground interface. If you come up with a very funny response, feel free to share it with the rest of us in the comments. That's it for Module two point to keep practicing, your skills are getting better with every new prompt you create. Next up, we'll dive into evaluating and refining prompts, stadium. 7. Evaluating and Refining Prompts: An Iterative Process: Hello and welcome back. This is module 2.3 about evaluating and refining prompts. By now, you've learned about the basics of prompts and how to create simple ones. In this lesson, we'll explore how to evaluate and refine your prompts using an iterative process that's easy to implement, an accessible to everyone. So let's begin evaluating and refining prompts. It's like taking care of a plant. Imagine you have just planted a seed and you're excited to see it grow, you'll need to water it, give it sunlight, and maybe even talk to it in a kind voice. Similarly, when you create the prompt, you'll need to nurture it, tweak it, and learn from it. The first step in refining a prompt is to review its output. Let's say you're using an AI model to generate the recite for your food blog. Your initial prompt could be something like create a vegan lasagna recite using eggplant and mushrooms. After inputting the prompt into the AI model, you receive an output. Now ask yourself, is the output accurate and relevant to the prompt? Does it cover all aspects of the prompt? Is it creative and engaging enough? Is this what I'm looking for? For our vigor lasagna example, let's say the output is a short recite with just a few ingredients and minimal instructions. It's accurate but not as detailed or engaging as you'd like. Don't worry, this is just the starting point. We'll make it better. Take note of the observations you may have, as there'll be essentials for refining the prompt. Once you've reviewed the output, it's time to modify the prompt. This could involve changing the phrasing, adding keywords, or providing more contexts. In our example, because we are talking about a blog article, it's probably a good idea to make it more detailed and engaging for the audience. So let's modify our vegan lasagna prompt. As you can see, by providing more context, we can help guide the AI model to generate a better output. In our example, we've added step-by-step detailed instructions and reach tomato sauce to make the prompt more specific. By the way, in many cases, step-by-step actually works like a magic formula. Sometimes, especially when dealing with a complicated problem. Adding step-by-step into our prompt will make the AI more focused and logical in its response. Now that you've modified the prompt, test it out. That's the process. Inputs, the revised prompt, and check if the output has been improved. If it's still not quite there, don't worry. Remember, this is an iterative process. Continue refining the prompt by going through the steps we discussed, reviewed the output, modify the prompt, and test it. With each iteration, the output should become more accurate and relevant for your problem. If you want to have a better control on the process, try changing your prompt in small increments. So if you make changes, make small changes with every iteration. Evaluating and refining prompts is an essential part of prompt engineering. It's a dynamic and iterative process that helps you fine-tune your AI generated content. As you gain experience in working with AI models and crafting prompts, you'll get better at guiding the AI towards generating the desired results in time. Once you get used to that specific area model, you will probably need less iterations to get to the right output. It's like developing a new reflex. You'll constantly get better with practice. So let's review the key takeaways from this lesson. Evaluating and refining prompts is crucial to achieve the outputs we are looking for. Review the output focusing on aspects like accuracy, relevance, engagement, and whatever characteristics you want to improve. Modify the prompt by adjusting phrasing, adding keywords, or providing more contexts depending on what you're trying to achieve. That's the device prompt and iterate the process until the output meets your expectations. Remember that prompt engineering is an interactive and dynamic process that requires practice and patience. Congratulations on completing module 2.3. You are now ready to fine-tune your prompts for various applications. But before we get to the more advanced concepts, in our next module, we'll talk about some good principles to follow when interacting with AI. See you there. 8. Basic Principles for Interacting with A.I.: Welcome to Module 2.4. In this section, we'll explore some simple but essential principles to follow when working with AI to create prompts. By following these guidelines, you will be able to generate better prompts and achieve better results. The first principle is focused on the topic. Understand the problem you're trying to solve first. Otherwise it might be difficult to clearly formulate the prompt. Stay on track, and avoid deviating from the task at hand. Sometimes results you'll get may steer away from the initial topic. Make sure you bring it back in focus. Make sure you're as specific as possible in your prompts. This will increase the probability of getting the results you're looking for. If you're unsure, don't hesitate to ask the AI for suggestions. Sometimes the answer is right in front of you because the model may actually suggest a better approach and give you a different perspective on the problem you're trying to solve. Remember that having a clear goal in mind is crucial when working with AI, make sure you know what you want to achieve and keep your prompts focused on that objective. The second principle is assumed nothing. Do not assume that AI knows something that may seem basic to you just because it has been trained on a huge amount of data does not necessarily mean it knows everything. Do not assume the AI understands the context. Sometimes it helps a lot to provide more context hints or simple examples in your prompt actually always double-check the output for validity. The problem with generative AIs is that sometimes they give false results. This phenomenon is actually called hallucination. Just like with people, because the answer is wrong, but it might be very convincing. Sometimes this may not be so important when creating entertainment content, but when the task requires very accurate information, the impact of these may be significant. There's not an actual intention of lying behind the model. Really consider, is that to be the right answer. It's basically our job to question and validate the output. As a conclusion, ai can be incredibly powerful, but it's not perfect. It can also be incredibly stupid. Sometimes, approach your interactions with AI systems with a healthy dose of skepticism and always verify the information it provides, especially when accuracy and precision matters. The third principle is start with simple prompts. Begin with straightforward and simple prompts. Make sure you are precise and clear in your intentions. Use simple language. Gradually add complexity as needed. When the result begins to resemble your intended output, you can start adding contexts, details in the question, or try to refine the format of the output. Like e.g. asking the AI to answer in a specific language or using a specific style. So when working with AI models, it's best to start with simple prompts and build upon them. This approach will help you understand the models, capabilities, and limitations better and make it easier to refine your prompts over time. The fourth principle, this is iterate and improve. Find the simple prompt to begin with and build from there. It might take a number of iterations to get what you want, but that's okay. Keep track of previous prompt iterations so that you can go back and reuse them if needed. Sometimes when the AI interface, it's built like a chat features such as width, chat, GPT or GPT for the interface itself will take care of that and we'll keep the whole history for you. But that's not always the case. If the chat history is not an available feature, then it's basically your job to keep track of all the previous prompts. The main takeaway from this principle is that AI works best when you embrace an iterative process. Test and refine your prompts. And don't be afraid to make changes or go back to a previous iteration if it works better. Now the fifth principle, practice makes better. The more you practice prompting, the better you'll become. As I said before, it's like developing a new reflex. Strive for improvement, not perfection. Remember, in most cases you do not need a perfect solution. Be aware of the tradeoffs and try to strike a balance between cost and benefit when interacting with AI. Remember, time is probably your most valuable resource. The whole point in using AI in most cases is to save time in doing a specific activity. If you end up spending more time than it would take you without the AI system, then it kind of defeats the purpose of using AI. Isn't it? As with any skill, practice is key to mastering prompt engineering. Keep experimenting. Learn from your mistakes. And remember that sometimes good enough is better than chasing perfection. By following these principles. Focusing on the topic, assuming nothing, starting with simple prompts, iterating and improving and practicing regularly, you'll be well on your way to becoming a proficient prompt engineer. Mastering this skill will allow you to unlock the full potential of AI and transform the way work. Now that we've covered these principles, Let's move on to the next module to continue expanding our understanding of prompt engineering and its applications. Stay tuned and see you in the next module. 9. Role Prompting and Nested Prompts: Welcome to Module 2.5, which is about role prompting and nested prompts. Sometimes providing more context into your prompts can be accomplished by using a method called roll prompting. In the first part of this section, I'll give you some examples and explain how rolled prompting can help you achieve better results with your prompts. First, I want to start by explaining an interesting phenomenon in human psychology. Sometimes we tend to assign the AI with human traits or characteristics. We basically visualize the model as a human person. In psychology, this is called anthropomorphism. And although it might seem a bit childish to employ, it turns out that in some cases, it may actually be an advantage when creating prompts when interacting with AI. Anthropomorphism can help users relate to the AIM or easily and create a more engaging experience. Okay, now let's explain the concept of role prompting or all prompting is a technique where you assign a specific role or identity to the AI to help guide these response and achieve more realistic results. By giving the AI or role, you can set the context and tone for each response, making the output more relevant for your purposes. This can be very useful when you're dealing with problems that require specific knowledge. Or you want the model to generate the output in a certain style. E.g. instead of asking what were the causes of the American Civil War, you can use row prompting like this. Choose an appropriate role for the ai based on your needs. E.g. you can ask the AI to pretend it's a historian, a scientist, or a teacher, depending on the type of information you're seeking and the kind of response you're trying to generate. Remember to take it slowly when adding more context in order to maintain a focus prompt and avoid confusing the engine. Start with something simple and more genetic and then gradually refine the context by setting a role for the AI or adding more information gradually to your prompts. Role prompting can also be successfully used in combination with another technique which is called guided iteration. Guided iteration is an approach where you work together with the AI in a back-and-forth manner to refine and improve a prompt. Let's imagine you are a researcher looking for an AI generated summary of a scientific paper. You can combine guided iteration, enrolled prompting like this. Oh, by the way, I think it's always a good idea to be polite to the AI. You never know. Okay, For the second part of this section, I want to tell you a few things about another useful technique that's called nested prompting. So basically, nested prompts involve embedding one or more prompts within another prompt. I know it sounds a bit like inception, but don't worry, I'll try to explain how it works. It's actually a technique that can be used to break down complex questions into simpler parts. It allows you to get more specific information or make the AI's response more focused and comprehensive. E.g. instead of asking DAI two, in one sentence, you can use nested prompts like this. In this example, the nested prompt that goes to separate but related topics about electric vehicles. First, focusing on their environmental benefits and then addressing the challenges faced by the technology and infrastructure. By combining these related topics into a single prompt, you guide the AI towards providing a more comprehensive and connected response on the subject of electric vehicles. Apparently, the two problems are not so different, but you'll see that the second version will result in a more detailed response. This technique can also be very useful when trying to generate more detailed and complex responses. E.g. if you're trying to create a blog article about a certain subjects such as electric vehicles. You could use a prompt like this one. In the first part of the prompt, you ask the AI to come up with some facts about the subject. And in the second part, you ask him to combine these facts and create the actual article. So use nested prompts whenever you want to obtain more specific and detailed information about the topic. In conclusion, understanding and utilizing role prompting and nested prompts can greatly enhance your interactions with AI systems. These techniques can help you generate more specific and useful responses, making, making it easier to obtain the information that you're looking for. Don't forget to combine these techniques with the principles that we've covered in the previous module. By doing that, you'll be able to handle various challenges when working with AI and get the results. You're actually looking for. State yearn for the next module where we'll explore more advanced prompt engineering techniques. See you there. 10. Chain-of-Thought Prompting: Welcome to Module 3.1, where we'll explore the concept of chain of thought prompting in prompt engineering, chain of thought is an advanced technique that involves breaking down complex tasks or questions into smaller and more manageable prompt. This approach enables better control and guidance over the AIs output, ensuring more accurate and relevant responses. Chain of thought prompting becomes necessary when a single straightforward prompt may not provide the desired results. It can be particularly useful in cases when the topic is very complex or has multiple layers. Or the AI needs additional context to provide the relevant response. Or when a step-by-step approach is required to guide the AI through a specific thought process. Let's consider a scenario where you want the AI to suggest a marketing strategy for launching a new eco-friendly product line for a company. So this is going to be our example. Without chain of thought prompting, you might input a prompt like. However, the AI might return an overly generic or off target output, such as this one. Now, let's have a look at this example and see how we can use chain of thought prompting to guide the AI to achieve a more useful output. So we begin by providing context and setting the stage. Next, we identify the unique selling points of the product line. We then propose marketing channels and tactics tailored to the target audience and product features. And finally, we combine the outputs from the previous steps to form a cohesive marketing strategy. So by breaking down the problem into smaller prompts were guided the AI through the process and receive the more accurate and contextually relevant output. This chain of thought approach can be applied to various scenarios to improve the AI's understanding and performance. That's it for this module. Today we'll learn how to solve more complex problems using chain of thought prompting. Remember to experiment with these techniques and don't forget to take advantage of the playground tool for that. That concludes Module 3.1. In the next module, we'll explore the exciting world of multilingual and multi-modal prompt engineering. See you there. 11. Multilingual and Multimodal Prompt Engineering: Welcome to module 3.2, where we'll dive into the exciting world of multilingual and multi-modal prompt engineering. Today, I will guide you through the process of working with multiple languages and different modes of communication using AI. In this increasingly connected world, being able to communicate across languages is essential. Ai can help bridge language barriers by understanding and generating content in various languages. Let's explore how we can use prompt engineering to build multilingual AI solutions. One common application of multilingual ai is translation. Let's say we want to translate an English sentence in Spanish. Our prompt could look like this. The AI would then provide a translate it outward. Another interesting application is language detection. To detect the language of a given text, we could use a prompt like this. Multimodal prompt engineering. Thanks AI capabilities of step further by allowing us to work with different types of data, such as images and audio. A multimodal AIS system is able to process and generate not just text, but also images or audio, or even video content. This opens up new possibilities and gives you incredibly powerful tools to solve many different types of problems. Imagine we want to generate a caption for an image. We could provide the AI with a description of an image, or depending on the system, are linked to the actual image file and user prompt like this. Audio transcription is another area where multimodal AI systems can shine. To transcribe audio. We could provide a link to the audio file or the actual audio content and user prompt that instructs the AI to generate a written transcription of the audio. Multilingual and multi-modal prompt engineering is powerful, but it comes with its own set of challenges and limitations. Some languages might not be well-supported and AI generated outputs may not always be perfect. It's important to be aware of these limitations and work iteratively until you achieve the best results. As a side note, at the moment when this video was created, not every AI system has multi-modal capabilities. You will find out, e.g. the one that we are using for the playgrounds does not have such capabilities. To succeed in multilingual and multi-modal prompt engineering is essential to experiment and iterate with your prompts. Be mindful of the AI's limitations. Continuously learn and adapt to new techniques and advancements in the field. There's literally new AI tools appearing every single day. Today, we've explored the fascinating world of multilingual and multi-modal prompt engineering. By combining these techniques, we can create AI based solutions that are more versatile and useful across a variety of applications. As you continue your journey with prompt engineering, remember to stay curious, keep learning, and never be afraid to push the boundaries of what AI can do. 12. Understanding the Non-Deterministic Nature of AI: Welcome to Module 3.3, where we will explore the non-deterministic nature of AI and discuss strategies for managing the variability and uncertainty that comes with it. Let's dive in. Ai systems, especially those based on machine-learning, can produce different outputs, even when given the same input. So that means you will most probably get different results each time you run the exact same prompt. The difference won't be much, but it will exist. It might be a different way of saying the same thing or a different style or length of the answer. This behavior is called non-determinism. It is important to be aware of this aspect when working with AI as it can sometimes lead to unexpected results. Let's take a look at a simple example. Imagine we're using an AI text generator to create a headline for a news article. We provide the following prompt. Write the headline for an article about a new environmentally-friendly car. Given the same prompt, the AI might generate different headlines each time. As you can see, each headline is unique. Even though the prompt remains the same. This variability can be both a strength and the challenge when working with AI systems. To manage the non-deterministic nature of AI, there are a few strategies that can help test multiple prompt variations to find the most reliable and consistent results. Fine-tune AI models to improve their performance and consistency for specific tasks. And set appropriate expectations. Setting appropriate expectations is crucial when working with non-deterministic AI systems. Keep in mind that AI generated outputs might not always be perfect or exactly what you're looking for. It is essential to be patient and flexible when reviewing AI generated results. Sometimes is just a matter of providing more contexts or being specific in your prompt so that the variability of the answer will diminish. On the other hand, the non-deterministic nature of AI can also lead to delightful surprises and creative solutions. By embracing the variability, we can discover new ideas and perspectives that we might not have considered otherwise. This can be especially valuable in fields like marketing design and content creation, e.g. let's look at another example. Suppose we want to create a tagline for a new brand of eco-friendly shoes. We give the AI the following prompt. Write a catchy tagline for a suitable shoe brand. Might generate multiple creative options such as each one of these taglines showcases a different creative angle, highlighting the power of AI is non-deterministic nature to spark innovation. While non-determinism can bring benefits, it's essential to stay vigilant and mitigate potential risks. Always review AI generated content for accuracy, appropriateness, and relevance before sharing it with others. Additionally, be prepared to iterate and refine AI outputs to ensure they align with your goals and values. To wrap up, understanding and managing the non-deterministic nature of AI is crucial when working with AI systems. Here are the key takeaways from this module. Ai can produce different outputs even when given the same input, which is known as nondeterminism. Test multiple variations and fine-tune AI models to improve performance and consistency. Set appropriate expectations and be prepared for some variability in AI generated outputs. Embrace the creative potential of AI is non-deterministic nature to discover new ideas and perspectives. Always review AI generated content for accuracy, appropriateness and relevance. Be ready to iterate and refine the outputs to ensure they align with your goals and values. Thank you for joining me in module 3.3. I hope you now have a better understanding of the non-deterministic nature of AI and how to manage it effectively. As we continue to explore the world of prompt engineering, keep in mind the strategies and tips shared in this module. Good luck and happy prompting. 13. Human-AI Collaboration: Best Practices and Strategies: Hello everyone. Today we'll explore the fascinating world of human AI collaboration. We'll discuss how we can work together with AI systems to create a powerful synergy that can improve our lives and careers. To make the most out of our collaboration with AI, it is essential to understand what AI systems are good at, where they might need our help. E.g. AI can analyze vast amounts of data and generate content quickly. However, it may struggle with understanding contexts, emotions, or cultural details. By being aware of the strengths and limitations, we can delegate tasks effectively and know when our human touches needed. Working with AI isn't a competition, it's a partnership. Think of AI as your teammates, someone with a different set of skills that compliments your own. When we approach artificial intelligence with a collaborative mindset, we can create a win-win situation where both human expertise and AI capabilities are used to their maximum potential. Let's explore some strategies to help us collaborate effectively with AI in different job roles. E.g. in content creation. Let Ai generate ideas or draft content while you focus on refining, editing, and adding context. For customer support, use AI to handle common queries, allowing you to focus on more complex issues that require empathy and understanding. Or when managing projects, use AI to track progress and predict potential bottlenecks. So you can make informed decisions and allocate resources efficiently. As we work with AI, we might face some challenges or misconceptions. E.g. we might worry that AI will take our jobs or we might be skeptical about the quality of AI generated outputs to overcome these concerns. Remember that AI is a tool designed to enhance our capabilities, not replace us. By focusing on what we can accomplish together, we can turn these challenges into opportunities. Let me share a few real life examples of successful human AI e-Collaborations. An editor using AI to generate a draft article, then revising and polishing it to ensure it meets editorial standards. A salesperson using AI generated product recommendations to guide customers toward items they're more likely to buy, leading to increase sales and customer satisfaction. A project manager leveraging AI insights to identify and address potential bottlenecks resulting is motor project execution and more efficient resource allocation. And these are just some examples. The possibilities are limitless. As we've seen, human AI collaboration holds immense potential for enhancing our productivity and creativity by understanding the strengths and limitations of AI. Adopting a collaborative mindset and employing different strategies, we can build successful partnerships with AI systems that may help us thrive in our careers, but also improve our personal lives. Now that we've explored human AI collaboration, let's move on to the next module where we'll dive deeper into advanced prompt engineering techniques. See you there. 14. Generating Ideas Using "Chaos Prompting": Hello everybody and welcome to module 3.5. Are you an artist? Do you work in a creative role? A few modules ago, we talked about the non-deterministic nature of AI. Remember? I said that it might represent a challenge because it might lead to unpredictable results. But I've also said that sometimes randomness and unpredictability can work in our advantage. In today's module, it's time to embrace the unpredictable and use the power of randomness to ignite our creativity. Chaos prompting is a unique approach to prompt engineering that allows you to generate unexpected and unique outputs from AI models. It's a journey that begins with a spark of inspiration and leads you through a labyrinth of ideas. Each more intriguing than the last. For a creative mind, this might just be the right way to trigger new ideas compared to what we've seen in the previous modules. This technique, it's a bit more untraditional because it's quite the opposite of generating precise and correct answers. So what exactly is chaos prompting and how can we use it in our interaction with AI? Chaos prompting is a technique that involves crafting prompts that intentionally introduce elements of randomness, ambiguity or contradiction, rather than seeking precise and predictable responses, chaos prompting encourages the model to produce outputs that challenge our assumptions, spark our curiosity, and open the door to new possibilities. The beauty of this method lies in its ability to generate a wide range of creative ideas. Whether you're a writer, e.g. seeking inspiration for a story or an artist exploring new themes, calles prompting can serve as a catalyst for igniting the AI's creative potential. So, without any further introduction, let me present you a typical chaos prompting session. We start by generating some random elements. Okay? I think we, we may have too many. Let's reduce the list a bit, shall we? Now it's time to build and add complexity. Let's mix it all together now. Well, pretty cool, isn't it? At this point, depending on the kind of creative work we might be doing, we could use this output as a seed for a story. Or we might even turn it into something visual, e.g. taking the previous output into an image generation AI model like meat journey might result into something similar to this. So now that we've seen an example, let's discuss some guidelines which might be useful in this process. First of all, embrace the unexpected chaos prompting. It's all about welcoming the unknown. Don't be afraid to ask open-ended questions, experiment with unusual combinations of words, or explore abstract concepts. The goal is to create a space where the AI model can surprise you with is it espouses play with contradictions. One of the hallmarks of chaos prompting is the use of contradictions or paradoxes by introducing conflicting ideas or combining opposing concepts, you can encourage the model to think outside the box and generate novel interpretations. Iterate and refine. Like any other prompting technique, chaos prompting, it's an iterative process. Feel free to build on the AI models responses, ask follow-up questions, or take the conversation in a new direction. Each interaction is an opportunity to go deeper into the creative process and uncover hidden gems. Also, stay open-minded when working with chaos prompting, it's important to keep an open mind and be receptive to unconventional ideas. Some responses may seem strange or a complete nonsense at first, but they can often lead to valuable insights or spark creative ideas. Obviously have fun, above all, chaos prompting. It's about having fun and enjoying the creative journey. It's an opportunity to experiment, play, and collaborate with the model in a spirit of curiosity and wonder. In conclusion, chaos prompting is a powerful technique that can enrich our interactions with AI systems and inspire us to think creatively. By following the guidelines I've mentioned, you can unlock a world of imagination and innovation that's going to surprise you. I bet you won't believe how amazed you'll be above the creative potential of an AI once you start using this technique. As you continue to experiment with chaos prompting, I encourage you to share your experiences, insights, and discoveries with the community. I'm really curious to see the kind of results your prompts might produce. I hope you enjoyed this lecture. See you in the next module where we will talk about using AI in software development. 15. Writing Code with the help of AI Part 1: Hi everybody, and welcome to module 3.6. As I mentioned in the introduction of this course, I have designed this learning tool for just about anybody. And up until now, I've tried to keep our conversation as non-technical as possible. Well, in this module though, it's time to touch a bit of a more technical subject. In an area where prompt engineering might have a very good use case, writing code. Whether you have a technical background or not. I think you've heard already about people using AI models to write code and create simple applications, web pages or automation scripts. Actually, the developer community right now, it's pretty much divided in terms of capabilities of an AI model to assist in writing code. There are a lot of YouTube videos and all kinds of demos proving that you can start from a simple prompt and build an app in just a few minutes. Let me tell you the truth, though. 80 per cent of the demos you see on the internet are a bit exaggerated. Not that they are fake, but they usually don't show you the whole sometimes painful process of iterating back and forth, refining your prompt and correcting bugs in the code. It's true that ai can be an incredible tool for developers, but I want you to have clear expectations about what it can and what it cannot do for you. You should also know that using an AI model directly to assist you in writing code, it's not the only path available. There are also more specialized tools for this purpose, which rely on AI to assist you in software development. Some examples include good hubs, co-pilot, Amazon code whisper, or quadriga. Anyway, let's assume you're currently using chat GPT. You don't have too much experience in software development. Be smart in your approach. Let's say you've written some code in your experience, but you're not qualified as a business analyst or a software architect, not a problem. The model can assist you with that also. But instead of using a basic prompt, like you should use the full power of AI with a prompt like this one. This prompt will result in a more structured and comprehensive assistance from the AI. In my experience writing code together with an AI, there are a few observations that I've noted and I would like to share them with you. Remember the golden rule, garbage in, garbage out. If you haven't already define the problem properly and try to get the AI to write code for an ambiguous task. Well, you probably won't get the results you're looking for. It also feels like a superpower. Ai allows you to tackle problems you've never solved before and approach topics that you barely studied in advance. E.g. I. Was able to translate a simple app from Python to PHP in a matter of minutes, having only very limited knowledge of PHP. But it still need supervision and control. Having at least some minimum knowledge of the technology you'll be using. It's a must. The more knowledge you have, the more useful the AI becomes. It act like an amplifier to your productivity. Sometimes though, it makes childish mistakes because they are very unexpected. These mistakes in the code are also very hard to identify. They might be in places where you wouldn't even think to look. Trying to debug this kind of mistakes can be very time-consuming. Take that into account. The max token limit can also be an issue. Don't forget that all current AI models have limitations in terms of the number of tokens they can process. If you're engaged in a longer conversation with chat GPT, e.g. at some point, the AI might simply start to forget the beginning of the conversation. This happens because the maximum token limit has been reached. So in order to continue the conversation, the model needs to delete some of the previous messages. This can be very inconvenient when writing code because suddenly the model may start outputting lines of code that do not apply to your app or repeat the same mistakes that were previously corrected. The best approach to avoid this, divide your code in smaller modules and functions from time-to-time. During the process, submit the entire code of a module or even the entire file, repeating the initial task. That way you regenerate the context for the AI model, helping you to stay on track and avoid hallucinations. Sometimes, especially with chat GPT and GPT for models, the answer you receive might be truncated. So you may ask for a Python function that solves a particular, a particular task, but receive only half of the function in the output. When that happens, you can simply ask the model in your next prompt to continue the code and give you the rest of the function. Okay? Because I want to keep each module within a reasonable length. I have divided this lecture in two parts. So see you in the next module where we will discuss the typical process of using an AR model as a coding system. 16. Writing Code with the help of AI Part 2: Hello and welcome to module 3.7, which is the second part of the lecture describing the way developers can use an AI model as a coding assistant. If you haven't yet completed the previous module, I strongly recommend you finish part one first in order to get a better understanding of this process and the reasoning behind it. Those of you who are experienced in writing applications already used, tried and tested software development methodologies. Probably. I'm not going to address the topic of a software development lifecycle in this course, but I would rather present you a simple approach that you can customize and adapt to your own development process. Remember that writing the code itself, It's usually just the tip of the iceberg. In most cases, is the process behind creating the code that matters most. So here's a typical flow that we might use to write code with chat GPT or GPT for. First, define the app requirements. Begin by clearly outlining the apps purpose, target audience, desired features, and any constraints. I also recommend you to include at this point any nonfunctional requirements such as security performance or compliance requirements. This information will provide context to the AI, enabling it to provide more relevant suggestions and code snippets. Second, choose the technology stack. Determine the appropriate technology stack for your app, including the programming language development frameworks, libraries, and tools. When discussing these topics with chat GPT provide the context of your chosen stack to get more accurate the systems, it may come up with additional suggestions. Third, break down the project into smaller tasks. Divide the app development process into smaller, manageable tasks. This will make it easier to work with chat GPT and get specific help on individual components of the app. For initiate the conversation with chat GPT, providing context about the app technology stack and the specific tasks you are working on. Be as clear and concise as possible when asking questions or requesting code snippets to ensure the most useful responses. The fifth point is about seeking guidance on architecture and design patterns. You might already have a pretty good idea on your architecture, but it's always good to have a second opinion. Ask Chad GPT for advice on application architecture and design patterns that are suitable for your project. This will help you create a well-structured, maintainable and scalable application. Six, request code snippets and examples. When working on individual tasks, asked Chad GPT for code snippets or examples that demonstrate how to implement specific features or solve particular problems. They replaced any placeholder values that the model might provide with the correct values that your application is supposed to use. Be sure to customize and test the code snippets in your app to ensure they meet your requirements. Do not assume that everything is working perfectly. Troubleshoot and debug if you encounter any issues or errors during development, describe the problem to check GPT and seek guidance on potential solutions, debugging techniques or best practices to resolve the issue. Optimize and refactor the code as the model for suggestions on how to optimize and refactor your code for better performance, readability, and maintainability. It's really great at that. Review and test the app. Once the app is complete, review and tested to ensure it meets the initial requirements and functions as intended. If necessary. Ask Chad GPT for advice on testing strategies, testing tools, and best practices. Don't forget to test and ask suggestions also for the non-functional requirements. The model may help you e.g. identifying security vulnerabilities or misconfiguration in your application. And finally, deploy and maintain, deploy your app and seek guidance from Chad GPT on deployment strategies, server configurations, or maintenance tasks as needed. Don't forget to break down your work into smaller pieces of code. Otherwise you may receive less useful outputs from the AI model. Remember to be patient in this process, as it may not always provide the perfect solution on the first trial. It's also important to test any AI suggestions and make adjustments as needed to fit your specific application requirements. But wait a minute, ai can be very useful, not just for writing code. There are a lot of other use cases where we might use an AI model to improve our productivity as software developers. Some examples in code review. Ai models can be used to review code and provide suggestions for improvements. The model can analyze code for potential issues such as bugs, security vulnerabilities, or styling consistencies, and even provide recommendations for fixing them. Generating documentation. Everybody loves writing documentation, right? Ai can be used to automatically generate documentation for your code. Given a code base, the model can generate human readable explanations, comments, and documentation for your APIs, for functions, classes, or modules. Another interesting use case is natural language interfaces. The new class of software is about to emerge with the current capabilities of AI, we can already build more intuitive and natural interfaces where users rely just on natural language to interact with our code. The learning curve becomes easier and the overall user experience can be improved. Instead of using buttons and menus, you're just talking to the application. Another use case is requirement analysis. As you have seen in my example, AI can also act as a business analyst. It can identify, classify, summarize, and validate the requirements, and that can save a lot of time and effort for developers. The model could help identify ambiguities, inconsistencies, and missing information in Requirement Documents. Code, translation. You need to translate from Python to JavaScript or C plus plus AI can help you with that too. Test case generation is another interesting use case for an AI model. It can be used to automatically generate test cases for code. Given a code snippet that the model can generate a set of test cases that cover different scenarios and edge cases. Again, saving you a lot of time. Okay, that's it for this lecture. See you in the next module where we will discuss about some other cool prompting techniques. 17. Using Prompt Compression Techniques: Hello everybody and welcome to module 3.8. In this lecture, we'll talk about a very interesting technique that we can use with the newer AI models like GPT for. Now, I know some of you may not have access to GPD for yet and are currently using chat GPT or other models. But I believe it's just a matter of time until this new model or other AI models with similar performance will become widely available. So I think you should know how to use this method already. Okay? Remember when we talked in one of the previous lectures about the limitations of an AI model. One of the problems with the current models is the maximum token limit. Language models often have a maximum token limit due to memory and computational constraints. Well, it turns out there is a way to overcome this limitation. The method we will be discussing in this module is called prompt compression. Some people have also nicknamed this technique sugar tongue. And yes, it is very similar to the way we compress files in our computers so that they take less space on the disk. As mentioned before, the prompt example that I will be using here is built and tested on GPT for you can achieve similar results in chat GPT, but the compression level will be much lower enough talking. Let's have a look at this example. Okay? So looking at the output, it's obvious that no human can actually read that. We ask the AI to compress the text using a language of its choice and any characters it may prefer. However, the length of the resulting texts, in this case is 30 per cent of the original texts, which is already quite remarkable. But let's see, what did the same way I model can understand from that? Will it be able to decompress it and maintain the meaning of the text? Let's find out. Although the text is slightly different from the original, the meaning is still there. We might improve the accuracy further if we adjust the prompt and ask you to provide a lossless compression. One thing to consider though, is compatibility. Any compression technique arbitrarily chosen by an AI model will result in texts that can only be successfully decompressed by the same model. Which means you cannot compress with GPD for and decompress it with another AI. Now that we've seen this capability, think about its use cases, e.g. you could summarize a very long conversation with the model and later continue that conversation in another session and still retain the original context. Or you could use compression to gradually create a really long bronze that otherwise would be impossible to use. And there are many other possibilities. Keep this in your toolbox and use it when required. Well, it's time to wrap up this module and prepare for the following topic. See you there. 18. Problem solving and generation of visual outputs: Although a GPD model is different from an image generating model and does not usually have the capability to create images or any visual responses. I'm going to show you in this lecture a simple way in which you can produce diagrams or other useful visualizations for your data using a model like chat GPT. But first, let's talk a bit about how we can use an AI model for problem-solving. There are several methods that we can use in order to identify the root cause of a problem. In this lecture, I'm going to emphasize the usage of a GPT model for two very popular problem-solving techniques, the fishbone analysis or Ishikawa diagrams and the swot analysis. These two methods also happened to rely on visual representations. So I'm going to take advantage of that to show you how to solve problems and also how to create simple but effective diagrams using a model like chat GPT. Why use an AI for problem-solving? Think about it. In most problem-solving techniques, we are usually encouraged to ask for different opinions on the subject from different subject matter experts that we might have available. Problem-solving is best when approached as a collaborative exercise. The problem is, if you're just an individual or a small company, you may not have the luxury of being surrounded with subject matter experts for all the problems you're trying to solve. You may also lack the necessary expertise in some fields in order to play the role of the expert yourself. Your only option might be to hire expensive consultants, which in many cases may not be a feasible option for obvious reasons. Well, here comes artificial intelligence. A GPD model which has been properly trained on an extensive set of curated knowledge, can act as your personal subject matter expert and help you solve some of your most complex problems. It may not be the most qualified experts out there, but multiple studies and benchmarks have already established that it's probably going to perform a bit better than an average consultant. Combine that with some proper prompt engineering skills. Yeah, that's you. And you will have a team of experts ready to help you solve even the most complex problems. So let's begin our discussion with an interesting problem-solving technique which is called the fishbone analysis, aka Ishikawa diagram, bearing the name of its creator. This method is very simple, yet very effective in helping you identify potential root causes for a complex problem or event. The reason they call it a fishbone analysis is that the method relies on a visual representation of the outputs which resembled the skeleton of a fish. Let's watch together how we can use a model like GPD for to apply this method for a specific problem. So in this example, I'm asking the model to help me identify the root causes for the fall of the Roman Empire. I know this is a fairly simple example because we are analyzing unknown event for which we have lots of data already. But don't worry, you can apply the same process for any problem. The only additional step that you'd need to include is to provide specific data to the model describing the event in sufficient detail to provide more context. In this particular case, I'm going to use a very direct and simple prompt because I'm not looking for a precise or very specific answer. As you can see, the model returns the most probable cause is divided into several categories. We can then further explore these root causes in a greater depth if we want to get a better understanding of the issue, e.g. okay, now, let's see how we can better visualize the answer. The easiest approach to create a more visual response would be to specify that we want the output to be returned as a table. Right? We've got a table, but still, this doesn't really look like a fish bone. So is there a way to create a more visual representation of the results? Yes, there is. The metal I'm going to demonstrate next relies on an open-source technology called dot language. This is a very simple but powerful way to create graphs. And it turns out that GPT models like chat, GPT and GPD for it can actually speak the language. So you see where I'm going? Let's go back to our example. And there you go. All we have to do now is copy the code and use a free visualization tools such as this one to generate our diagram. From here, we can also modify the looks of the graph or export the result as an image file. Going back to our problem, now that we've identified the root cause is the next step is to ask for solutions. And that's it. We just saved the Roman Empire. Would this be a good idea? I don't know. Great. Now that we've discussed about the fishbone analysis, let's move on to our next popular method which is swat. While the swat analysis is not a problem-solving technique, in a traditional sense, it aids in problem-solving by providing a structured approach to evaluate the current situation. The name comes from the four elements that it can identify, the strengths, weaknesses, opportunities, and threats. Let's see how we can apply the swap method to our problem. As you can see, this time, I'm using a more detailed prompt. That is because swot analysis has to be very specific on the context and current objectives in order to produce meaningful results. Let's see what we've got. Normally, a swat analysis results in a visual representation of the items divided into four quadrants. Can we achieve something similar? Obviously, now that we have a current understanding of the situation, the next step would be to ask for an improvement plan. But in this case, I would like to do something a bit more interesting. Apart from the fishbone analysis and Swat, there are several other problem-solving techniques that you should be aware of. I encourage you to explore the ways in which these other techniques, such as the five whys, the Six Thinking Hats, mind-mapping, decision analysis or PESTLE analysis can help you solve a complex problem. And also how chat GPT or GPD for can be used to assist you in applying these methods. Actually, this would make a pretty good practical assignment for you to apply the concepts that I've explained in this lecture and familiarize yourself with some of the most popular problem-solving techniques available. So your assignment is to choose any problem you might be facing right now or one that you faced recently. Describe it in a prompt and work together with the AI model to identify potential causes. Man, Why not? Maybe even potential solutions? But please choose a method different from the ones that we've discussed in this lesson. If you think the method you've chosen would benefit from a visual representation, that would be a great opportunity to exercise generating visual outputs with the dot language. That's it for this module. See you in the next lecture where we will discuss about other advanced prompt engineering techniques. 19. AI-Assisted Questioning: Building on our exploration of problem-solving and generating visual outputs in the previous module. Let's have a look at another important aspect of problem-solving, AI assisted questioning. This technique allows you and the AI to work together more effectively in solving complex problems by encouraging the AI to ask the necessary questions to better understand the problem and provide more accurate responses. As we saw in the previous module, AI models like check GPT can be used to identify root causes and potential solutions using techniques such as fishbone analysis and swat analysis. In this module, we will learn how to use the AI assisted questioning to complement those problem-solving techniques and further refine our understanding of the issue at hand. The key to effective AI assisted questioning is to ensure the AI asks the right questions. This not only helps clarify the problem, but it also ensures that I can provide the most relevant and accurate solutions. Here's how you can use chat GPT to perform a IOCs that questioning and solve complex problems. Step number one, identify the problem. Start by describing the problem you want to solve or the topic you want to explore, be as specific as possible to give the AI are clear context of the issue. Step number two, start asking for questions. Prompt the AI to generate questions that will help you better understand the problem. Step number three, answer the questions. Now it's time to give the contexts that it needs, wants the AI provides a list of questions, respond to them in detail to give the model more context and information. Step number four, iterate and refine. As you provide answers, the AI might generate additional questions or dive deeper into specific areas. Keep engaging in this back-and-forth process to refine the understanding of the problem. And finally, step number five requests the solutions. Once the AI has gathered sufficient information, ask you to provide potential solutions or recommendation based on the data provided. You will definitely get better results by using this method. In the same way we utilize chat GPT for a fish bone and swat analysis. Ai assisted questioning can be combined with those techniques to create a more comprehensive problem-solving process. So let's say this process in action with an example. Suppose you are experiencing a decline in sales for your online store. And one to identify the root causes and potential solutions. You could start by describing the problem to the AI and asking it to generate questions to better understand the issue. The model might respond with questions like you would then respond to these questions providing the eye with more context and information. And it may follow up with additional questions, diving deeper into specific areas. The thing is you'll have to continue this iterative process until the AI has a solid understanding of your problem. And finally, ask the model to provide potential solutions or recommendations based on the information gathered. It might suggest actions like improving your marketing strategy, addressing common customer complaints, or offering new products to meet changing consumer demands. In summary, AI assisted questioning allows you to collaborate with the model to solve complex problems by encouraging the AI to ask relevant questions, you can ensure it has comprehensive understanding of the issue and can provide more accurate solutions. Sometimes in this process by asking the right questions, you will also get yourself new perspectives on the problem and be able to come up with out-of-the-box solutions by applying some lateral thinking. Okay, I think it's time for some practical activity to exercise what we've just discussed. Choose a problem or topic relevant to your personal or professional life. Use the AI assisted questioning process that I've described together, information, refine your understanding and requests solutions from the model. Don't forget to iterate and refine your questions to get the most valuable insights from the model. In the next lecture, we will explore more advanced prompt engineering techniques, enhancing your ability and using AI for real life scenarios. See you there. 20. Automating Emails and Social Media Posts: Hello everyone. In this lesson we'll dive into how prompt engineering can help automate emails and social media posts, making our digital lives more efficient and engaging. We live in a connected world and communication is more important than ever. Aim prompt engineering can help us craft more effective messages, saving us time and ensuring our content is engaging and relevant. When it comes to crafting prompts for emails, it's all about the context. Here out a few key points to consider. The recipients relationship to you. The purpose of the email, and the desired tone. Let's say I want to schedule a meeting with a colleague. I could create a prompt like compose an email to schedule a meeting with a colleague discussing a new project idea. The AI will then generate a suitable email that I can use or edit as needed. Using templates is another great way to automate emails. We can create a prompt that includes the template structure and then customize it with specific details for each unique email. I'll let you experiment with this idea on the playground and test it yourself. See what you can come up with. Social media is a powerful platform to share our ideas and engage with others to create problems for social media posts. Consider the following aspects. The platform, which might be Twitter, Instagram or Facebook or any other platform, the target audience. And the purpose of the post. Let's say I want to promote an upcoming workshop. They could create a prompt like that will probably generate a tweet that captures the essence of the event while appealing to my target audience. In this lesson, we've explored how prompt engineering can be applied to automate emails and social media posts. By understanding the context, purpose, and audience, we can create engaging content that save us time and keeps us connected. In the next lesson, we'll learn how prompt engineering can be used for content generation in blogs, articles, and reports. See you there. 21. Content Generation: Blogs, Articles, and Reports: In this lesson, we'll explore how prompt engineering can help create engaging and well-crafted contained in various formats. So let's get started. When it comes to content generation, prompt engineering can be an invaluable tool, will focus on three main types of content. Blogs, articles, and reports. First, let's talk about blogs. Blogs often require a more casual and conversational tone. With prompt engineering, you can guide the AI to generate blog posts tailored to your target audience. To craft a blog prompt, start with a clear topic and a few relevant keywords. E.g. remember, the purpose is not to replace human creativity. Adding that human touch on top of the content generated by AI will make a huge difference. Next, let's discuss about articles. Articles are typically more formal and structured than blogs. With prompt engineering, you can create well-researched and informative articles on various topics. To create a prompt for an article provides a specific topic and requests for evidence-based information. E.g. remember to always check the data for accuracy and make sure it doesn't provide outdated information. Most of the AI technologies out there, including chat, GPD, and GPT four, have been pre-trained on a specific set of data and can only answer based on that data alone. So if the training has included only data available until a certain year, the AI will have absolutely no clue about newer events happening in the world. On top of that, most AI models do not possess the ability to access the internet. That means their outputs may not be up-to-date and the model may not even mentioned that. Finally, let's look at the reports. Reports are more detailed and comprehensive than articles, often presenting data and analysis. Prompt engineering can help generate well-organized and insightful reports. To generate a report, your prompt should include the topic, scope, and any data or sources that you'd like to be incorporated. E.g. you may then adjust the tone and style to match the format that you need for the report. In this lesson, we explored how prompt engineering can be applied to content generation for blogs, articles, and reports. By leveraging AI, you can create engaging and informative content for a wide range of purposes. In the next module, we'll discuss using prompt engineering for task delegation and project management. Stay tuned. 22. Task Delegation and Project Management: Hello everyone and welcome to module 4.3 of our course, prompt engineering for everybody. In this lesson, we'll explore how prompt engineering can help with task delegation and project management. Are you ready? As we begin, let's talk about how AI can assist us in delegating task effectively. By leveraging prompt engineering, we can create intelligent ways to analyze workload, prioritize tasks, and assign them to team members based on their skills and availability. Now, let's discuss how prompt engineering can help in task prioritization. By inputting the right information. Ai models can sort tasks based on urgency, importance, and deadlines, enabling us to focus on what matters most. Here's an example of a prompt that can help in task delegation. By providing relevant information, the AI model can suggest an efficient as distribution among team members. Next, let's explore how AI can aid in project management. Prompt engineering can help us track project progress, identify potential bottlenecks, and generate status reports, making project management smoother and more efficient. Consider this example of a prompt for generating projects status reports. By inputting relevant data, the AI model can generate the concise and informative project status report to keep everyone in the loop. Ai can also be a valuable tool for enhancing team collaboration. Through prompt engineering, we can create ai models that help facilitate communication, scheduled meetings, and manage shared resources, ensuring that teams work together seamlessly. Here's an example of a prompt for scheduling meetings. The AI model can analyze everyone's schedules and recommend the suitable meeting time, making it easier to coordinate and collaborate. While AI can be incredibly helpful in task delegation and project management, it is essential to recognize its limitations. As humans, we must still apply our judgment, intuition, and empathy to make informed decisions and guide our teams effectively. As we wrap up this module, let's recap what we've learned. Ai can assist in dust delegation by analyzing workload, prioritizing tasks, and assigning them based on team members skills and availability. Prompt engineering can help in project management by tracking progress, identifying potential bottlenecks, and generating status reports. And these are just some examples. Ai models can enhance them collaboration by facilitating communication, scheduling meetings, and managing shared resources. It's also essential to recognize AIs limitation and apply human judgment, intuition, and empathy in decision-making. Now that we've explored how prompt engineering can be used for tasks delegation and project management. You are ready to apply these principles in your own work. In the next module, we'll discuss how prompt engineering can be utilized in job roles that are currently affected by AI or that might be affected in the future. Stay tuned and see you in the next module. 23. Customer Support: Enhancing Human-Agent Collaboration: Hello everyone. Today we will explore how prompt engineering can help enhance human agent collaboration in customer support will discuss how AI can complement human support agents, handle repetitive tasks and inquiries and improve response times and customer satisfaction. Ai tools can be a valuable addition to customer support teams. By using prompt engineering, we can design virtual assistance that work alongside human agents to provide faster and more accurate support. Let me give you an example. Suppose we have a virtual assistant that helps answer frequently asked questions. We can craft a prompt like in this case, the area assistant will then provide a concise and informative response, allowing human agents to focus on more complex inquiries. Ai driven virtual assistants can efficiently handle repetitive tasks, reducing the workload of human agents. For instance, considered a prompt like, how can I reset my password? The AIS system can respond with a step-by-step instructions, saving human agents diamond effort. This way, customer support teams can devote their energy to more challenging and engaging tasks. One of the main advantages of a IOCs that customer support is the ability to reduce response times. Prompt engineering allows us to design a system that can answer multiple inquiries simultaneously. Imagine a prompt like in this case, the AI assistant can quickly provide a detailed response, ensuring customers receive the information they need without having to wait for a human agent to become available. It's also important to strike a balance between AIS systems and human expertise. While AI can handle many tasks efficiently, there are situations where human intervention is necessary. Complex issues. Empathetic understanding and nuanced communication are areas where humans are very good. Well-integrated customer support system should allow for seamless handoffs between AIS systems and human agents, making sure that customers receive the best possible support. So by embracing collaboration between AI tools and human agents, customer support teams can achieve optimal results. Prompt engineering enables us to create ai systems that compliment the skills and expertise of human agents, leading to enhance the efficiency and customer satisfaction. In summary, prompt engineering can greatly benefit customer support teams by complementing human agents with AI tools, handling repetitive tasks and inquiries, improving response times and customer satisfaction, and overall balancing AIS systems and human expertise. That's all for this module. In the next section, we'll explore how AI driven personalization and efficiency can improve the retail and e-commerce industry. Stay tuned. 24. Retail and E-commerce: AI-driven Personalization and Efficiency: Welcome everyone to Module 4.5. In this module, we'll explore how AI and prompt engineering can assist people working in retail and e-commerce, making their jobs more efficient and helping them deliver better customer experiences. One of the ways AI can help sales teams is by providing product recommendations. Using prompt engineering, we can create personalized suggestions for customers based on their preferences, shopping habits, and other relevant factors. Here's an example prompt. Ai can also play a crucial role in inventory management and demand forecasting. By analyzing historical sales data and external factors, AI can help retail employees make informed decisions about stock levels, reducing the risk of overstocking or running out of popular items. Here's an example. Obviously, for this prompt to return the correct results, we should also provide relevant sales and event data. In the prompt. Customers appreciate personalized experiences and AI can help retailers tailor their marketing and communication to individual needs. By analyzing customer data prompt engineering can generate targeted promotions, discounts, or product suggestions that are relevant to specific customers, e.g. in this module, we've seen how AI and prompt engineering can assist retail and e-commerce professionals in various ways, including providing AI assisted product recommendations for sales teams, utilizing AI for inventory management and demand forecasting, enhancing customer experience with personalized offers and communication. So by embracing AIs, potential, people in retail and e-commerce can create better customer experiences, improve efficiency, and ultimately thrive in their roles. 25. Creative Writing and Brainstorming: Using AI to Generate Ideas and Drafts: Hello and welcome to Module 5.1. In this module, we'll explore how AI can help us generate ideas and drafts for creative writing and brainstorming. We'll look at practical examples and discuss strategies to make the most out of AIS systems without compromising our creativity. One of the biggest challenges in creative writing is coming up with ideas or story outlines. With AI, we can generate various ideas by providing a simple prompt, e.g. by using this prompt, the AI engine will provide us with story ideas that we can use as a starting point. We've all faced writer's block at some point. Ai can help us overcome this by suggesting new directions or perspectives. E.g. this prompt will encourage the AI to generate unexpected twists, giving us fresh ideas to break through the block. Ai can also help us refine and polish our work by providing a piece of our writing, we can ask the AI to suggest improvements or alternative phrasings. Here's an example prompt. In this case, the engine will return a revised version, which we can use as an inspiration to improve our writing. When using AI to assist with creative writing, it is essential to maintain a balance between AI generated ideas and our own creativity. Remember, AI is a tool to enhance and compliment our creative process, not replace it. In this module, we've learned how AI can be a valuable partner in our creative writing process, from generating initial ideas to refining our work. In the next module, we'll explore how AI can help us with research and information curation. Stay tuned, and let's continue discovering the fascinating world of AI and prompt engineering. 26. Efficient Research and Information Curation: AI-Powered Summarization & Analysis: Welcome to Module 5.2. In this module, we'll explore how ai powered tools can help you save time, find relevant information, and analyze content effectively. We'll also look at some example problems that you can use with AI engines like Djibouti for it to make your research process more efficient. One of the first step in conducting research is finding relevant sources by using a power search queries, you can quickly discover the information that you need. E.g. this prompt will help the AI engine locate articles that match your criteria, saving you time and effort. Keep in mind though the possible time limitations that we've discussed in the previous modules, the model may not have the latest data available. Once you have your sources. Ai can help you extract key information and summarize the content. Here's an example. With this prompt, the AI engine will generate a concise summary of the article allowing you to quickly understand its main points. Ai can also help you analyze data and derive insights from it. E.g. by using this prompt, the engine can analyse data and present you with clear trends, making it easier for you to understand and interpret the information. To recap. Here are the benefits of using AI powered research and new information curation tools. Time-saving, ai can quickly identify relevant sources and summarize content, allowing you to spend more time on analysis and decision-making. And hence the accuracy. I can process large amounts of data and minimize human errors in information gathering and interpretation. Customize insights. Ai generated summaries, analysis can be tailored to your specific needs and interests. Let's imagine you're working on a project to improve your company's sustainability initiatives. Here's how AI can help you streamline the research process. First, use AI to identify relevant articles and sources on sustainability best practices. Second, use AI generated summaries to quickly understand key points and save time. And third, analyze data trends and insights using AI to identify areas where your company can improve its sustainability efforts. In this module, we've learned how ai powered tools can make research and information curation more efficient and effective. By using AI engines like GPD for you can save time and hence accuracy and gain customize the insights that will help you excel in your professional endeavors. As you move forward. Remember to experiment with different prompts and strategies to find the best ways to harness the power of AI for your research and information curation needs. And always keep in mind that AI is here to support your creativity, not to replace it. Happy researching. 27. Enhancing Communication Skills: AI-Assisted Proofreading and Writing: Hello everyone and welcome to Module 5.3. Today we're going to explore how AI engines like GLUT4 can help us enhance our communication skills. We'll focus on proof reading, writing, and adapting our style and tone to different audiences. First, let's talk about how AI can improve grammar, spelling, and style. Ai powered tools can quickly analyze your texts and provide suggestions for grammar, spelling, and style improvements. This can save you a lot of time and help you feel more confident about the quality of your writing. E.g. you might input a sentence with a few mistakes like this one. The AI could then offer the corrected version. Hey, I can also help you craft more persuasive and concise messages. Let's say you're writing an email to convince your colleagues to adopt a new software tool. You can provide the AI with your initial draft and some key points you'd like to emphasize. The AI might then generate the suggestion for you. Lastly, AI can help you adapt your communication style and don't for different audiences. Imagine you need to rewrite a formal report for a more casual audience. You can provide the AI with a sample paragraph from the report and asked for a more informal version. By leveraging AI engines like GPT before you can improve your communication skills, save time, and adapt your writing to suit various situations. So use these tools as a helpful guide and allow your own unique voice to shine through. That's it for this module, we'll talk about time management and prioritization in our next topic. See you soon. 28. AI-Driven Task Management and Decision Making: Hello everyone. In this section we'll discuss how AI can help us manage our time and prioritize tasks more effectively. We'll also explore how AI can assist in making data-driven decisions and optimizing resource allocation. So let's dive in AI. It can be a very powerful tool for prioritizing and scheduling tasks. By providing a list of tasks and their attributes, we can use AI to generate an optimized task order based on factors like deadlines, importance, and estimated effort. Here is an example. Ai can help us spot inefficiencies in our workflows. By analyzing our work processes and habits, ai can identify bottlenecks and areas that need improvement, e.g. as you can see, it's all about the way we formulate the prompt. Ai can also assist in making informed decisions based on available data. It can analyze complex datasets and provide insights to help allocate resources more effectively. Here's an example. Okay, To sum up the contents of this module, ai can be a very valuable partner in time management and prioritization. By leveraging AI driven task management and decision-making tools, we can enhance our productivity and make better-informed decisions. Remember that AI is here to support and compliment our skills, not replace them. So let's make the most of it and create a more efficient and fulfilling work environment. In the next section, we'll discuss how AI can help us with professional development and lifelong learning. Stay tuned. 29. AI-Powered Professional Development and Lifelong Learning: Welcome to Module 5.5. This time we will explore how AI powered personalized learning paths can help you continually grow your professional skills and stay updated with industry trends. We'll dive into how AI can assist you in identifying skill gaps, crafting personalized learning plans, and discovering relevant content. Let's start by discussing how AI driven assessments can help you identify your skill gaps and learning opportunities. Ai engines like GLUT4 can analyze your professional background, e.g. accomplishments and goals to provide a tailored evaluation of your strengths and weaknesses. E.g. you can input a prompt like. Then, the engine will provide you with a list of areas to work on helping you focus on your learning efforts. Once you've identified your skill gaps, AI can assist you in creating personalized learning plans. It can create a list of relevant resources such as articles, courses, or webinars to help you learn and grow. For instance, you could provide the following prompt. The AI can generate the tailor learning plan, which includes a step-by-step resources, timelines, and milestones to track your progress. It's all about using your imagination. Be creative. In today's fast-paced world, staying updated with industry trends and insights is crucial. Ai powered content discovery tools can help you stay informed by filtering through vast amounts of information and delivering content that's relevant to your interests and professional goals. You could input a prompt like and the system will provide you with a list of articles or resources that are highly relevant for your query. Keep in mind though, as we discussed earlier, you may not get the most up-to-date data depending on the training input for the model. So take that into account. In conclusion, ai powered personalized learning paths can be an invaluable tool for professional development and lifelong learning. By using AI engines like GLUT4, you can identify your skill gaps, create, tailor learning plans, and stay informed about the latest industry trends as you continue to grow and evolve in your career. Remember that AI is here to support you in achieving your goals, enhancing your productivity, nurturing your creativity. To recap, we've covered how AI can assist you in identifying skill gaps and learning opportunities through AI driven assessments. Crafting personalized learning plans with AI curated resources. And also staying updated with industry trends and insights through AI powered content discovery. In a nutshell, embrace this superpower to support your professional development and lifelong learning journey. And watch as your skills and knowledge continued to flourish. See you in the next module. 30. Ensuring Fairness and Reducing Bias: Hello and welcome to module 6.1 where we'll discuss ensuring fairness and reducing bias in prompt engineering. Today we'll learn how to create inclusive prompts and consider the impact of our language models on a diverse audience. First, let's talk about the sources of bias in AI systems. Artificial intelligence models like the ones we use in prompt engineering, are trained on vast amounts of data from the Internet. As a result, they might learn and reproduce the bias is present in this data. It is crucial for us to recognize and address this bias is to ensure that our AI generated outputs are fair and inclusive. There are various types of biases that can emerge in AI systems such as gender bias, racial or ethnic bias, socioeconomic bias, or geographical bias. We should be mindful of these biases when crafting problems and evaluating AI generated outputs. Here are some strategies that we can use to reduce bias in our prompts. Use inclusive language. Avoid gender-specific pronouns and other exclusionary terms, e.g. instead of asking the AI to generate the list of famous male scientists, you could ask it to generate a list of famous scientists. Balanced examples. That's another rule. When providing examples in your prompts, make sure they represent a diverse range of individuals and perspectives. Also, test with diverse Inputs. Test your prompts with a variety of inputs to evaluate their fairness and inclusivity. After crafting your prompts, it's important to evaluate the AI generated outputs for potential biases. Here are some steps that you can follow. Review the content, go through the AI generated outputs and check for any bias or exclusionary language. If you find any issues, revise your prompt, or adjust the AI model's parameters to obtain more inclusive outputs. Ask for feedback. Share the AI generated outputs with a diverse group of people and ask for their opinions on potential biases are problematic content. Their perspectives can help you identify and address issues that you might have missed. And lastly, iterate and improve, continuously refine your prompts and AI model parameters based on the feedback and insights together, remember that prompt engineering is an iterative process and it's essential to learn from your mistakes and make improvements over time. As we work with prompt engineering, we must also consider the ethical implications of our AI generated outputs. Here are a few questions to ponder. Are we unintentionally reinforcing stereotypes or harmful biases? Are we being fair and respectful to all individuals and communities? Are we creating content that could be harmful or offensive or divisive? Asking these questions can help us create more responsible and ethical AI applications. In module 6.1, with discuss the importance of ensuring fairness and reducing bias in prompt engineering with explore strategies for crafting inclusive prompts, evaluating AI generated outputs, and incorporating ethical consideration in our work. By being mindful of these factors and continuously refining our prompts, we can create AI applications that empower and serve diverse audiences. 31. Responsible AI and the Future of Work: Welcome to module 6.3, where we will discuss responsible AI and its implications for the future of work. As we embrace AI technologies, it's crucial to ensure that they are used ethically and responsibly. Let's explore how we can achieve this. The first step towards responsible AI is to establish guiding principles that shape AI development and use. Here are some key principles. Human-centered, ensuring AI serves human needs and values. Transparency, making AI systems and their decision-making understandable. Accountability, assigning responsibility for AI systems, behavior and the outcomes. Fairness, reducing biases and ensuring equal treatment of users. And nevertheless, security. Protecting AI systems and user data from unauthorized access and use. Ai should be used to complement human work rather than replace it. For instance, in prompt engineering, AI can help you generate creative ideas while you as a human can refine and finalize the output. This way, ai can enhance our skills and help us become more efficient and effective in our jobs. Let's look at an example prompt that demonstrates this collaboration. The AI system can provide various suggestions and you can select the most relevant and appealing options, customize them to your team's needs and plan the activities accordingly. As AI technologies evolve, it is essential to stay up to date with the latest advancements and best practices. This will help you harness the full potential of AI and ensure responsible use in your personal and professional life. It can be a very powerful tool for collaboration, enabling us to work together with our digital partners in a variety of domains. By fostering a co-operative relationship between humans and AI systems, we can tackle complex problems and drive innovation. In conclusion, responsible AI, it's about using AI technologies ethically, ensuring that they compliment human work and promote collaboration by adhering to guiding principles and continuously learning and adapting, we can shape the future of work in a way that is beneficial for everyone. This brings us to the end of module six. Thank you for joining me today and see you on the next section where we will conclude this course and summarize key takeaways. 32. An Introduction to the ChatGPT Plugins: Hello everybody and welcome to a brand new section of this course. In this chapter, we are going to address the most recent functionalities introduced in chat GPT and how those can be helpful in different scenarios and use cases. If you've been using the chat GPT interface in the last few months, you've probably noticed a new feature that was recently made available, the plug ins. What exactly is a chat GPT? In think of it as an extension to the core functionality of the AI model. To make a very broad analogy, it's like running a software on top of a particular system. These plug ins enhance the functionality and user experience by allowing Che ChipT to interact with third party applications. This gives the AI model the ability to access the Internet, to retrieve real time information such as the latest news or stock data. And also to interact with live systems in order to execute tasks or initiate transactions. Okay, But how do they work? There's a lot of technical aspects to discuss regarding the inner workings and architecture of these extensions, but since this course was never meant to be technical, I'm going to skip that part. If you're interested on the topic, you can find plenty of technical documentation on open AI's website. Instead, I will guide you on how to activate these plug ins, Explain what they can and cannot do and how to use them in real life scenarios. Oh, by the way, a short disclaimer. First, this section is not necessarily related to prompt engineering, but it's rather meant to be an introduction to the Chachi Pt plug ins and the other advanced features. That being said, every prompt engineering technique that we've discussed in the previous sections still apply because even when using these plug ins, you'll still be interacting with the model by using Prompt. Also, you should know that the plug ins are only available to paying open AI customers having a PT subscription. And as of September 2023, they show up as a beta feature, which means they are not yet available as an official production release. The reason they are already open to the public has to do with open AI's incremental approach in launching new features and their efforts to ensure AI safety and alignment. Keep in mind that most of these plugins are developed by third parties and not by open AI. They are, however, strictly verified by open AI through a review process to make sure that they are safe to be published. Even so, you should be mindful of what kind of information you share with them, because that information may end up being processed and potentially stored outside of open AI's environment. You shouldn't miss out on using these plug ins because they can substantially enhance the core functionality of Cha GPT as you'll see in the next modules. But first, let's see how we can activate and use these plug ins. The process is pretty much straightforward. Once you've accessed the Cha GPT interface, get the option to choose between GPT 3.5 which is the older model with a bit more limited capabilities but also very fast GPT four which is the latest and the most capable model. Although a bit slower than GPT 3.5 You'll notice that the previous model has no additional option for enabling any plug ins. That is because the plugins are only supported by the latest model, which is GPT four. All you have to do now is select the GPT four model and then click on the Plug ins option. Scroll down to the Plug in store and install the plug ins you want. After the installation completes, the new plug ins will show up in your quick access list and you can now enable them whenever you need. Just to make it clear, the plugins are installed on your open AI account and not locally on your computer. Once you've installed some plugins, you'll be able to enable a number of maximum three from the list of all available ones. The plugins you enable will then be active for this particular session and the AI model will be able to interact with them when required by your prompts. There are currently almost 1,000 available plugins, with many of them having overlapping functionality, and some even requiring that you authenticate on other websites to access that functionality. Some of them are free to use, while others may require a subscription and all that stuff can make things a bit confusing. It's not my goal to provide you with a detailed description of every available plug in. Instead, in the next modules, I will give you some examples of useful plug ins in real life scenarios. You'll get an understanding of how they work and you'll be able to explore and discover new ideas along the way. Therefore, see you in the next module. 33. Deep Dive into the ChatGPT Plugins: Ready to find out more about the Chachi pet plug ins and how you can use them to enhance your productivity in day to day activities. In this module, I'm going to teach you how to use some popular, but also very practical plugins available right now. Using them will definitely make your work easier and improve the already great experience provided by Chat GPT. As a reminder, there are currently almost 1,000 plugins available. So I'm not going to discuss about all of them in this course. I'll show you a few plug ins explaining how to use them. And from there you can explore on your own and experiment with others. Because once you understand how they work and how to use them, it becomes a very natural process. Let's begin with Wolfram. Wolfram Alpha is a computational knowledge engine developed by Wolfram Research. It's basically an online service which answers factual queries by computing the answer from structured data. Like a big database in general, Wolfram, it's optimized to give concise factual answers rather than long articles or web search links. The goal is to directly answer the user's query. The wall from plug in is just a clever way of integrating GPT four with Wafram's knowledge. To use this particular plug in in chat GPT, all you have to do is first to install it and then activate it also to make sure that the AI model actually uses the wall from plug in. Just add that in your prompt like this. Let me show you some key features of this Plug in. It can provide real time data based on your queries. That is possible because unlike Chat GPT, some of the Wolframs knowledge is actually updated in real time. Well, that used to be the case at least before October 2023. At the moment, open AI is gradually launching a new feature in chat GPT, which will enable the engine to browse the web for the latest information without any kind of plug in. But still, you could already do that using the Wolfram plug in for example. You could type something like, it will tell you the current weather in London. It has a vast collection of curated datasets and can do computations on them to generate answers. This includes data on weather, finance, geography, science, nutrition, and others. It can generate visualizations like graphs, charts, and maps to present answer in a visual way. For example, you could ask GPT to. It also has verified knowledge of real world data and facts. That makes it a very useful tool whenever you need to fact check an answer provided by Ch, GPT or any other large language model or any kind of information whatsoever. Because remember hallucinations, you can never completely trust the answer provided by an AI model, especially when logic, math, or factual events or entities are involved. But as it turns out, you can use war frame to verify any questionable answer. For example, probably the most important feature is that it can perform complex mathematical and scientific calculations. You can type in a function or query like it will show the graph and compute the result. As you can see, sometimes it may return a very long answer, but you can obviously tweak that with the right prompt. In summary, the war from plug in can help you validate any factual information whenever precision is important. It can also compute answers to factual queries using real time, up to date information. It's using its own curated data rather than crawling the web or predicting the most probable answer like a large language model. Okay, let's talk about another very useful plug in which is web pilot. This plug in gracefully solves one of the main problems of a pre trained model like chat, GPT, The inability to browse the Internet, and lack of up to date real time information about the world. I know open AI introduced this as a core feature in September, but still in my opinion, web pilot provides better results. Web pilot can be used to extract data from any link provided by the user and feed this data as an input to chat GPT for example. You could use it like this, although with some limitations. Sometimes web pilot is able to extract data from links. Even if the data isn't store in classic HTML format, it might be able to read most PDFs, docs, or Excel documents. You just have to provide a proper link to the document. Here's a protyp. Many of the currently available chat GPT plug ins don't really have a technical documentation or a description of their potential use cases. Their name is also not always representative for their actual functionality. This can create confusion for the user not knowing exactly what the plug in does or can be used for. There is an easy solution for that. Actually, you just have to activate a specific plug in in your GPT four chat session and use the following prompt to get a more detailed description. The third most useful plug in, in my opinion, is Wikipedia. It can answer general knowledge questions or supplement the knowledge of Cha GPT. For example, if you have a question about a historical event, a famous personality, a scientific concept, or any other general knowledge topic, the plug in can search Wikipedia for relevant information. It can also provide information on current events or offer insights on breaking news. If something significant has just happened and it's documented on Wikipedia, the plugin can be used to retrieve that information. When using the Wikipedia plug in, it passes your prompt as the query to search Wikipedia and then shares the results with you. Apart from these three plug ins, there are many others you can explore on the plug in store. I'm just going to give you some suggestions. For example, diagrams, show me. This is one of the best ways to visualize any kind of data. It can create a lot of diagram types including graphs, sequence diagrams, entity relationships diagrams, user journey maps, Gantt charts, and many more. It's also interactive, which means that once it generates a diagram, you can get a link to edit the diagram online if you wish to make any modifications. Aipdf's another interesting one. This plug in is designed to help you extract and understand information from PDF documents without having to manually sift through the entire document. It can either summarize the contents or allow you to ask questions about your PDF. The Earth Plug In offers several functionalities related to generating map images and obtaining coordinates of locations. Need help with a foreign language. You can use the speak plug in if you need to know if there's an AI based solution for a particular problem. There's even a plug in for that, it's actually called. There's an AI for that. You explain your scenario and the plugin will recommend potential AI tools that you could use. That's just the tip of the iceberg. Have fun exploring other plugins. You know the drill. Now, see you in the next module. 34. The Code Interpreter Plugin: Hi everyone. This entire section is dedicated to exploring some of the most useful plugins available to use in Chat GPT. This module in particular focuses on advanced data analysis, which is a very special kind plug in one that's been developed by Openai themselves and has a different and much more complex behavior than all the other plug ins out there. You may already be familiar with it under a different name, the code interpreter. It's been recently rebranded by Open AI to provide more clarity on its purpose. As you're already well aware, a large language model is nothing more than just a very complex prediction machine. It basically starts from the user's prompt and builds a response by establishing the most probable combination of words based on the patterns it has seen during its training on vast amounts of text that sometimes works like a charm, magically producing very convincing and accurate outputs. However, as we discussed in our previous lectures, in some cases the AI model may struggle in returning accurate results. I'm talking about the so called hallucinations and troubles with mathematical computation or logical reasoning. It turns out that word or token prediction alone is not enough to solve every real life problem out there. This is where the special plug in comes into play. If I may use an oversimplified analogy, think about the two hemispheres of the human brain. One hemisphere is logical while the other is creative. You can think of the code interpreter as the more logical part of ChagPT's architecture, as suggested by its name, This plugin has the ability to run code and return very precise algorithmic results whenever the model determines that such a method is more appropriate. When having this plug in activated chat, GPT will analyze your prompt and decide whether the task that has to be solved can be solved creatively by the AI engine alone. Or whether it should be solved by writing and executing a program in an algorithmic fashion. Sometimes it may employ both methods. For example, it may run a simple program to get very specific results and then combine these results with some creative AI ingredients. Probably the best part, it enhances the way you interact with the A model by allowing you to upload files and download the resulting output. Which is great. You can use it for analyzing specific documents, uploading or generating Excel or Word files, images, and many more. Okay, activating the plug in is very much straightforward. After initiating a new chat session, in the chat GPT interface, you have to select the GPT four model and then you'll have the option to enable the advanced data analysis plug in. Keep in mind that the plugin will only be active for the current session and obviously it will have no impact on other chat windows you may have opened. Once you provide a prompt, the AI will determine whether it needs to call the plug in depending on the task that you have provided. If the engine determines that the plug in is required, it will automatically launch it and use its output. In solving the task, you can wait for the task to be completed in the background, or you can click on the Show Work button and see the actual code. It builds in real time. Here's a Protyp. If you just need a task to be completed without chat GPT adding any additional comments, you can append this at the end of your prompt. Sometimes the plugin will provide a visual answer that will be displayed on the chat GPT interface. Sometimes it may provide you with a link to download the resulting file. If you prefer to download the resulting output, you can also specify that in your prompt, together with a file format you prefer. Now it's time to also learn about the limitations that unfortunately come with using this plug in because there are a few limitations to consider. First of all, it only generates code written in the Python language. And while it has several hundred Python packages pre installed, it cannot by default install any external Python packages. Any limitations of this language and the existing packages which are extensions to the core Python functionality will also limit the capabilities of the plug in. It does not have Internet access, unfortunately, you cannot use it in combination with Internet capable plug ins such as the web pilot for example. It also builds a temporary environment that is deleted at the end of the chat session. Any files that you may have generated will be forever lost if you haven't yet do them on your computer. Yes, while it allows you to upload documents, you can only upload files of a certain maximum size. Okay, This module was meant to be just a quick introduction into the purpose and capabilities of the advanced data analysis plug in. See you in the next module for some real life use cases for this technology. 35. The Code Interpreter Plugin Part 2: Hello and welcome back. As I've promised, it's time to explore some practical scenarios where the advanced data analysis plug in may be useful. In this module will cover topics such as optical character recognition, image manipulation, conversions from different formats, generating presentations, data extractions, and other interesting scenarios. Do you wonder what kind of magic this plug in is capable of? Let me show you some of the best use cases. First, optical character recognition using Tessera. Tesseract library is a great tool that can recognize and converted or handwritten text in images into machine readable text. In other words, if you have a picture of a page from a book or a handwritten note, tesseract can read the text in that picture and give it to you in a format that you can edit. This process is also known as optical character recognition, OCR. Here's an example. Upload your image file and use this prompt if you want. You can also ask the advanced data analysis plug in to reformat the extracted text in a more convenient way. Extracting tabular data from PDFs using Came Lot Pi. Have you ever tried to copypaste the data from a table from a PDF to an Excel file? You've probably noticed that it's a hit and miss thing. Most of the times it doesn't work as it should. It turns out there's a Python library for that, which can be used in the advanced data analysis plug in. Here's the prompt you should use after you uploaded the PDF. Here's the result. Converting PDFs to editable word documents. Talking about the PDF, sometimes we might want to edit the contents of the PDF document, but that's impossible unless you purchase a license for the Adobia Acrobat software or use some conversion services available on the Internet. Those services though, they are either limited in functionality or require a fee for the unrestricted version. That's not a problem anymore with the advanced data analysis plug in. If the PDF already contains selectable text, you could use the following prompt. Or use a prompt like this one in case your PDF is just a scanned document. As you have already noticed, sometimes the plug in may encounter different errors returned by the code it generates. The nice part is that it's able to understand the errors and make the necessary corrections to make it work. It's persistent if the code it generated the first time does not work, it's going to keep trying different approaches until it gets the right result. How to create Powerpoint presentations? Here's another interesting example. Are you still using the good old Powerpoint for your presentations? I do. You can use the advanced data analysis plug in to create Powerpoint presentations. Check out this prompt, That's just a basic scenario with a bit of creativity in your prompts and maybe uploading your own data, you can create great presentations in no time. I think that's enough for one module. If you want to see more, check out the next video. See you there. 36. The The Code Interpreter Part 3: Welcome back. In this module, we'll explore even more cool things you can do with the advanced data analysis plug in. Let's jump right in, face extraction from photos. The open source computer vision library, Open CV in short, is one handy Python library which enables real time image and video processing, object detection, facial recognition, and other computer vision applications. Let's see what it can do in the advanced data analysis plug in. Have you ever wanted to extract individual faces from a group photo? Simply upload your group photo and use this prompt. The plugin will automatically detect the faces, crop them out, and you could also have the plug in, save them for you into individual files if you want. Creating animated ifs from videos. We live in the age of memes. Ever wonder how you can take a memorable sequence of your favorite movie and turn it into an animated Jeff that you can easily post online. Here's an idea to accomplish that. Upload your video sequence and use this prompt. The plug in handles the video processing and conversion for you. If you want, you could try asking the plug in to add text or add any additional effects. Just be creative with your prompt advanced data visualizations. The Bock Library in Python allows you to create interactive and visually appealing data visualizations for modern web browsers. I know we talked already about creating diagrams and different data visualizations, but what if you could actually interact with your diagrams? Just upload some data and try a prompt like this one. It will generate the plot code and even host it for you to view it online. But you can also download the diagram using the link and share it or host it wherever you prefer. X KCD style diagrams want to convey a concept with a touch of humor and simplicity. Diagrams are no fun, but if you want something more free hand, you can have it draw diagrams in the stylistic hand drawn look of the XKCD comics. Yeah, there's a library for that too. Here's an example Prompt Visualizing GPX courses. How about planning your next outdoor adventure? If you're passionate about nature and outdoor activities like I am, you probably know already how useful can a GPS track be. Sometimes it doesn't only prevent you from getting lost, but it also allows you to monitor your progress during the activity. Unfortunately, GPS data is hard to read for a human at least, but not for an AI model. Coupled with a very smart plug in, here's a way you can visualize courses and elevation profiles. The plug in will crunch the numbers and generate the visualization automatically, creating Youtube thumbnails. You can even create Youtube video thumbnails. Well, probably not the best looking thumbnails, but it's a quick alternative. In place of more advanced tools, here's the prompt. It will mix together the image and text into a properly sized and formatted tum nail for you. Obviously, you could tweak the prompt to achieve better results. Epub to doc X conversions. Finally, converting books is a breeze. Simply use a prompt like this one. The plug in handles each step automatically to give you a clean word document output. As you can see, the possibilities are endless with this plugin. I hope these examples have inspired you to explore and get more creative in using Chat GPT and its plug ins. The advanced data analysis is using Python under the hood. If you want to have more control on its outputs or achieve even better results, I'd suggest you learn a few things about Python, thus making your first steps into the world of computer programming. Trust me, no matter what's your current job right now, it's definitely worth the effort. Okay, up next we'll talk about another new feature which can be really useful in chat GPT, the custom instructions. See you in the next module. 37. The Custom Instructions Feature in ChatGPT: Hello again and welcome to another exciting module. Let's talk about another brilliant feature of Cha GPT, the custom instructions. Although introduced only recently, this feature proves already to be really useful for those who want tailored responses to their prompts. In this module, we'll uncover ways to use it in order to command GPT to produce content in a specific manner, create structure, guide its thinking process and much more. First of all, let's have a quick look at how we can enable this feature in the Cha GPT interface. The option is actually located in the main menu. In the first box, you can provide the model with some details about you such as your location, line of work, age, or any personal preferences. By doing that, you're encouraging Chat GPT to customize its answers for you. But I personally wouldn't really use that one because in practice I would need to change it too often. The thing is you can only have one safe profile at a time. Which can be pretty inconvenient if you're moving back and forth between different personas. The second box, though, has a lot of potential. It basically allows us to fine tune the responses we get from the model. I'm going to show you some clever ways in which this feature can be used to improve our experience when using Chat GPT. For example, you can limit the length of the response to a certain size. Sometimes the answers produced by Chat GPT are extremely long and contain a lot of redundant information. There's an easy way to correct that in a semi permanent fashion by adding a custom instruction like this one and making sure you enable the instruction for all the future chats. From now on, you'll receive more concise and to the point answers using a certain answer style. Let's say we work on a bigger project that requires a longer conversation with chat GPT, maybe even using multiple chat sessions. But we also want to maintain the same consistent response style throughout our entire interaction with the model. There's an easy solution for that. Avoid using certain words. Generating text with an AI model is quick and very convenient. But I think you've noticed already that all AI models tend to repeat certain words like they're obsessed with them. Unfortunately, that's a side effect of the training data used in their learning process. The good news is that we can ask the model to avoid certain words, thus making the output sound more natural and less AI generated encouraging interactivity by asking clarifying questions. We talked about AI assisted questioning in one of the previous modules. Sometimes this method can be used to explore different perspectives of an issue and enable the model to gather the necessary information to better understand the problem at hand. And we can add custom instructions to accomplish that. In every conversation where chat GPT considers that the clarifying question might be useful, obtain a certain number of suggestions instead of a single answer. Another useful instruction that we could use can make chat GPT provide multiple suggestions for certain queries. Here's an example, of course you can customize it as you like, obtain code only answers avoiding lengthy observations. Everyone who's used GPT four as a coding assistant knows already that the answer it provides is not always to the point and sometimes includes a lot of useless explanations or too many comments. A quick and permanent solution would be how to avoid hallucinations, as we discussed in the previous modules, sometimes the outputs produced by I models like GPT four, may be unreliable. This has to do with the way large language models work by design, and the quality of their learning process. There are a number of techniques that we can use in order to increase the reliability of a model's response. But if you're looking for a quick improvement, here's the way you could use custom instructions. This one is my favorite. I'm actually using this instruction all the time. Define shortcuts in the prompt that trigger specific functions. To be honest, this is by far the most practical way of using custom instructions. Actually, all the customization methods that I've shown you so far are great, but they all have a major inconvenience. Once enabled, they will affect all future responses and they will limit your entire user experience, and that's not ideal. A much better approach, in my opinion, is to define shortcuts that you can actually use when and only when you really need to customize the response. Here's an example of how can you accomplish that. From now on in my future chat sessions, I'll be able to activate a shortcut just by adding a certain parameter in my prompt. Of course, you could also use shortcuts in the first box to define different personas that you can later activate with a simple parameter in your prompt. One thing though about using custom instruction is that you need to remember you're actually using them otherwise, especially if you customize too much, you will end up having a completely different user experience. That's why I strongly advise you to keep your customizations at a minimum and always acquire a taste for your preferences first that you know exactly what to change and how. Keep in mind that the length of the text you enter as a custom instruction or the total custom instructions you have will decrease your actual context window. The more instructions you add, the less memory will be available for the entire conversation. Try to keep it short. Always add customizations in small increments and make sure they do not overlap, as this may lead to unexpected behavior If you're in a chat session and want to have a quick view of the custom instructions without using the menus. You can also hover your mouse over the information sign at the top of the screen. It's incredible to think of the many applications for this feature. Remember, the more specific and clear your instruction is, the better tailored the response from Cha GPT will be. There you have it, a brief overview of the custom instruction feature of Cha GPT, but trust me, we've only just scratched the surface. Experiment with your own ideas and discover the vast number of possibilities. 38. Course Conclusion and Key Takeaways: Hello everyone. As we've now reached the final module of our course, I'd like to take a moment to recap some of the most important things we've learned together and share some key takeaways. Throughout our journey, we've explored the world of prompt engineering, starting with the basics and moving on to more advanced techniques and best practices. We've seen how prompt engineering can help us in our everyday tasks, in job roles impacted by AI, and even in our personal and professional development. We've also gone through some simple but useful case studies that show how regular, non technical users can harness the power of AI engines like GPT four, to elevate their professional skills and increase productivity while still preserving creativity. We talked about Chat GPT and GPT four in particular. But remember, these concepts can be applied when interacting with any AI model. Let's have a look at some of the key takeaways from our course. Prompt Engineering. It's a powerful tool that allows us to interact with AI models more effectively. Simple prompts can be transformed into more sophisticated ones to achieve better results. We can use prompt engineering for better productivity and inspiration. Ethical considerations, though, are crucial when working with AI power tools including fairness, reducing bias, and ensuring privacy and data security. As we wrap up our course, I want you to know that this is just the beginning of your journey in the world of AI and prompt engineering in general. I hope the concepts and techniques we've discussed will inspire you to continue learning, experimenting, and discovering new ways to leverage AI in your personal and professional life. Thank you for joining me in this exploration of prompt engineering. I'm truly grateful to have experience with you and I wish you all the best in your future endeavors. Never stop learning. And always remember that the power of AI when used wisely, can be an incredible asset to help you unlock your full potential. 39. Course Conclusion and Key Takeaways: Hello everyone. As we've now reached the final module of our course, I'd like to take a moment to recap some of the most important things we've learned together and share some key takeaways. Throughout our journey, we've explored the world of prompt engineering, starting with the basics and moving on to advanced techniques. We've seen how prompt engineering can help us in our everyday tasks, in job roles impacted by AI and even in our personal and professional development. We've also gone through some amazing case studies that show how regular non-technical users like us can harness the power of AI engines like GLUT4 to elevate our professional skills and increase productivity while still preserving our creativity. Let's take a look at some key takeaways from our course. Prompt engineering is a powerful tool that allows us to interact with AI models effectively. Here's an example prompt. Simple prompts can be transformed into more sophisticated ones to achieve better results. We can use prompt engineering to enhance our productivity and creativity in various domains such as writing, research, communication, time management, and learning. Ethical considerations are crucial when working with ai powered tools, including fairness, reducing bias, and ensuring privacy and data security. As we wrap up this course, I want you to know that this is just the beginning of your journey in the world of AI and prompt engineering in general. I hope the concepts and techniques we've discussed will inspire you to continue learning, experimenting, and discovering new ways to leverage AI in your personal and professional life. Thank you for joining me in this exploration of prompt engineering. I'm truly grateful to have shared this experience with you and I wish you all the best in your future endeavors. Never stop learning. And always remember that the power of AI, when used wisely, can be an incredible asset to help you unlock your full potential.