Generative AI and Copilots

1. Introduction: So now it brings us to what is possibly the most exciting topic surrounding artificial intelligence at this time, generative AI. Now, generative AI is artificial intelligence that is capable of actually creating new content or generating content using models trained on existing data. So with generative AI, we can generate text, images, audio, video, and I'm sure if there is more content, more content types on the way, generative AI will be able to adapt to it, as well. Now, it's generally powered by machine learning models such as transformers and guns or generative adversarial networks and others. And I'm not going to get too much into that into the dynamics of what those are and how they work. But generally speaking, we just discuss what machine learning is. It's a machine that has been trained on data, and it can make certain decisions based on the historical data that it has seen. So now, generative AI takes it a step further, where it can now actually generate content based on that kind of historical data. So it has seen stuff that looks like this and you're asking for something new. I can pick parts of the historical data that it knows to then generate a new response for you. So examples of generative AI engines include hat GPT, and I think that's probably the most popular one. That's the one that kind of came on the scene and showed us what Gene AI is really capable of. We also have Doll E for images. We have music MLM, sorry, for audio. And I've actually used that one. Actually used the generative AI engine the other day to generate a song for me. So I spent hours working on this song on my keyboard, and then I was like, Hmm, let me see what the generative AI would come up with. And I was blown away at the amount of detail, the level of detail and awesomeness that came out of that engine. So generative AI is very powerful. I think it's here to stay, and we definitely once again need to appreciate how we can leverage this technology for our solutions. No, of course, we've all watched movies, and we've seen what the robots and AI engines in theory are capable of. And I think that is in the back of our minds when we think about generative Air and how scarily accurate things seem to be. But let us discuss what generative AI is not before we write it off as something that's dangerous. So generative AI is not fully autonomous. It relies on data that it has been trained on and different models that we human beings have built. It cannot think and it cannot reason by itself. And it's only as accurate as the body of knowledge is. So it's really only as accurate as our body of knowledge that we have given to the engine is. It is also free from bias. So AI models can actually inherit biases from the data they are trained on. So let us say, for instance, we talk about training the machine learning, right? If I train if I wanted a system that could detect dogs and cats or let's say dogs different from human beings. And then I gave it a bunch of pictures of dogs and cats, and I told this engine that all of these are dogs. And then I gave it a bunch of pictures of human beings and said, these are human beings. That means that anytime this system sees a four legged creature that looks like a dog or cat, it's going to think it's a dog because once again, that's the bias that I introduced in the training. So to bring that full circle to gen AI, it's only going to be able to generate new content or response based on what it knows or think it knows, things it knows, right? So you know, bear that in mind that, you know, the body of knowledge that it has been trained on really influences the content or the response you get from it. It's basically not always accurate, right? So accuracy, once again, is relative to the data it has been trained on. And I think the scenario I just shared also addresses the accuracy issue, and it is not a replacement for human beings. I don't think it is a replacement for human beings. You know, the whole craze of AI a whole part of it was that, you know, it's going to be replacing people in certain jobs and so on. And I agree, every time there is an advancement in some technology, there is going to be a need for a human to be maybe upskilled or move or shift in a different direction because now this new technology can do that job. And this has always happened. Since the Industrial Revolution, this has been happening, right? We had printing press. That would have replaced people who used to sit down and write out the news. Now, we don't need that because we have printing press, right? Um, horses and carriages used to carry people in the way of cars. So technology will always displace the current need for certain services and certain human operations, but it is not a replacement because you can always upscale change direction. And I like to think of it as my intern. I don't use it as my body of knowledge wholeheartedly. I still have to review what generative AI tells me and, of course, correct it as necessary. So don't worry, it's not here to replace you, but it is there for us to use to maximize our potential. 2. Generative AI Ethics: Alright, so, of course, we all know the great phrase. With great power comes great responsibility. So there are some ethical implications when we are using AI because AI, especially the way it has blossomed in more recent times, has the potential to unlock all kinds of opportunities good and bad for businesses and individuals. So we want to make sure that we use it in a way that does not promote discrimination, and we should always be as fair as possible whenever we're using AI technologies, make it accessible to all persons in different age groups, cultures, and, you know, everybody don't have any no gatekeeping, pretty much. We should also make sure that we're not going to use it in a way that violates, sorry, human rights. We should not bring any harm or subordination to any other human being or community, whether it's physical, economical, social, political, et cetera. Throughout the life cycle of AI systems, the quality of the life of the human beings should be enhanced, it should not diminish, right? We also want to make sure that we do not um, impact the environment in any negative way. So the AI systems life cycle should not increase carbon footprints as much as possible. And, of course, we should try to just reduce the overall environmental impact of anything that we do with this data and these systems. We should also make sure that we're accountable for everything. So when you are using the system, or when we're implementing a system, we need to determine who is responsible for the actions and decisions and the decisions made by the AI systems. It's also good to give a disclaimer that, hey, this information did not come from me. It was generated by an AI system, so at least persons will be informed of the potential inaccuracies in what you are about to present. And finally, right to data, privacy and protection. So, of course, we are talking about AI systems being trained on data. What we don't want is to give it to more sensitive and personal information such as the AI system is going to be sharing details about anybody else that should not be, right? So we always want to make sure we protect human dignity, autonomy and agency. And throughout the life cycle of the system, any data that is to be collected should be consistent with the international law and in line with values and principles. So some best practices, we always want to be aware of biases in accuracies and ethical concerns. We always want to validate AI generated output, especially in critical applications, ensure that there's compliance with privacy laws as generative models often use large datasets, and we cannot always vet where the data is coming from. So always make sure you're in compliance and make it clear when the content is AI generated to avoid confusion and misinformation. Avoid using generative AI for malicious purposes like deep fakes or misleading information and, you know, mocking persons voices, letting them say things or making it sound like the person saying something that they never said. You want to avoid those kinds of situations, B as ethical as possible and use this technology properly. Once again, with great power comes great responsibility. 3. Generative AI Overview: Alright, so we've kind of discussed this before, but let's go a little more in depth in exploring what exactly is generative AI or gene A or short. So we know that artificial intelligence is designed to as best as possible imitate human behavior, and we try to program machines with some algorithms that would mimic our decision making and even some of our senses like hearing or speech or vision, right? So, generally speaking, that's what AI is. It's supposed to take some information and be able to execute tasks without our explicit intervention and even learn from these tasks over time. So generative AI describes a category of those capabilities where we can create new content using the same intelligence or the artificial intelligence. So a typical example of interacting with a generative AI solution is usually through like a chat application, and an example of a chat application is at Microsoft copilot or GPT. And generative AI applications accept natural language input. So basically plain text, English. So we are developers, but you might also you might not be an out and developer. You might be like a business analyst or, you know, a non technical person. The fact is generative AI is designed to take natural language the way we speak naturally with our natural a way of communicating, not talking computer language here. And then it can process that and return an appropriate response. And these appropriate responses are usually responses that are generated on the fly based on what you've asked. And these responses can come in the form of natural language. So it's responding to you in a chat like fashion. You asked the question, so it's giving you a response. It can be an image, and it can even be code because you could say, Hey, help me with this or generate code that does something for me. So it's a very powerful breakthrough in artificial intelligence, and it is definitely here to stay, and it is good to appreciate the power that is in our hands. So when we come back, we're going to look at some of the language models that usually power these generative AI. 4. Copilots and how they work: Now let's explore what co pilots are. So a co pilot is a generative AI assistant integrated into applications, and it's usually done through a chat like interface. It provides contextualized support for common tasks in whatever application it's in. So the most accessible co pilot would be Microsoft copilot, which is integrated or integratable into a wide range of Microsoft applications and general user experiences. If you have a Windows machine and you're running Windows 11, especially if you're running P, you actually have a co pilot built in. And you'll see that co pilots are really there to boost productivity, creativity, and generally provide AI generated content and can help with different tasks and different content requirements. On top of all of that, as a developer, you can extend the co pilot by creating plugins, and these can integrate into business processes, and you can even create your own type of co pilot, which, of course, you would kind of be starting from a foundation or model, but you can train it with data or train it to be able to perform a specific task. So think about it. You can go and look at existing co pilots and see which one is missing based on your estimation and try to develop your own copilot. Now, let's discuss a bit more about Microsoft co pilot features, which are found in several applications, and there are different use cases for Microsoft copilot. For instance, there's copilot.miicrosoft.com, which can answer questions, create content, and search the web. So you can go to that URL. We'll be doing that in a few and you'll also find that there's a co pilot for Edge browser, which if you have edge, which if you have a Windows computer, you already have edge, you will see that there's actually a dedicated pin for copilot right there that allows you to basically do the same thing as if you were going to the website. And then you have copiloto Microsoft 365, which integrates copilot into your productivity applications, and Office 365 usually comes with your word processor, your PowerPoint, Excel, et cetera. So having a copilot there helps you to generate the documents more quickly than if you were sitting down and doing that task by your there are co pilots available for other business line applications. So there's co pilot for security based applications for Microsoft Azure for Power BI, GitHub, which I'm sure you've heard of Github copilot, and there are many other applications. 5. Exploring a Copilot: Alright, so let's get into a quick demo where we look at how Microsoft Co Pilot works. So I'm here at copilot.miicrosoft.com, and you can opt to sign in, and you would sign in using your Microsoft live account. And if you don't already have one, you can go ahead and create one, not pressuring you or anything, but let us go ahead and get started. So here, they'll say, What can I call you? I'll just put in my name, then they'll say, What kind of voice tone would you like? As you can hear you. And I'll just skip past that. So here, you'll see that, you know, it's giving me some insights, and this is because I'm signed in. So it's kind of going by what it knows about me or I would suggest based on who I am. Here we can add documents. We can also review previous conversations, and we can use the microphone to talk to the co pilot to give it a prompt. So let us say that I wanted to say, write me a story about Jamaica and why it is a great place to visit. Alright? Let's see what our co pilot says. Alright, so you give me a nice story about why Jamaica is a beautiful place, and then notice that it's kind of ending. What's your favorite kind of adventure? I like to go hiking. Let's see what it does. So then it goes on to continue the conversation. So this is what we said about context, right? So it knows the context of the conversation because it knows what I asked initially what the context is. It's about Jamaica. Then it says, What's your favorite adventure. I like to go hiking, so now it follows up with letting me know that I can go hiking in different parts of Jamaica. So yes, I am ready. And then they're just encouraging me here. Let's go. Alright? So this is pretty cool. This is copilot, and once again, you can come here and you can ask you questions. You can have a conversation. And I'm sure it's a bit more insightful if you are asking more pointed questions. And we're going to talk about prompts and what we call prompt engineering after this. So let us jump over to OneDrive. So I have OneDrive, and that allows me to open up Microsoft Word online. You can also open up Microsoft Word. If you have your lives ONDriV account, you can open up Microsoft Word and you can use the documents that I have shared with you. So I asked Chat GPT in another window to generate this same thing about Jamaica, right? It goes into a bit more detail. I see here that it kind of comes with these little tokens that in certain editors would mean, like, bold, and I think this is Mart Down. Yes, this is Mardwn. So that would mean he four in HM, that's on. But I'm not going to fixate on those little things. That's not really why we're here. What I want to show you is that in Edge, there is a co pilot. So we just looked at co pilot here in the browser. But in the edge browser sorry, we looked at the website here. Now, I'm in the browser, and I'm going to show you that there's a co pilot pane right here in the top right hand corner. So if I click that, you see that chat pane is appearing. Now, see what's possible with copilot in Edge. Alright, let me generate a paid summary. So copilot is actually looking at what I am looking at, looking at the document here, and then it's going to generate that summary for me. I can now ask more questions tell more about the blue mountains. And it generated a nice little fact list about Blue Mountains and note that it's not just generating stuff, yes, it is probably saying parts based on its own generative powers, but then you also see that it's citing a source. So it's letting me know that this part is from that source on that website. Accessibility that's coming from that source on that website. So not only is it generating content, but it's also going across the Internet and finding sources, that complement what it is saying. And then afterwards, I can like or this, generally speaking, I want to give feedback so that the system learns whether or not it was accurate or if it needs to adjust accordingly. You can also download response or read it aloud if necessary. Alright, so I'm going to run one more experiment. You don't have to run this experiment, but I'm going to try it because we're all developers here more than likely we're all developers. So I'm going to see if my co pilot can actually look at this block of code and suggest some changes, right? So, look at this now. It's already in the context of you know, looking at that document and generating a summary. So every question that I ask is kind of within that context. I don't want that, so I'm going to click for a new topic, and then I'm going to say suggest ways to optimize this code. Let's see if that would work. Alright? So it did generate some stuff, but contextually, it is not really what I was hoping for and it's not really helpful within the context. So this is some code from my test driven development with Aspeed on a core course, where we built a testable API and application. So I just went here. It's on GTubtsFreeF access. I just went here and I said, so just we optimize the code. Now, it's seeing or it either sees blazer web assembly or thinks that it sees blazer web assembly, or it just knows this prompt and it's just answering based on the last time we got the prompt that, hey, somebody wanted to know how to speed up Blaser web assembly. So that's not quite what I wanted, but it is what it gave me. So I'm just showing you the different scenarios how the copilot may be useful and may not necessarily be the best for the situation. 6. Generative AI Prompting : Now let's prompting. And I'm saying copilot prompting, but the general principles can be applied in generative AI chat where they receive a prompt and should give a response. So responses usually depend on the language model being used and the type of prompt that you provide. So the language model, remember, represents a body of knowledge. Is it trained on a large amount of data, so it knows a little about everything or is it trained on a smaller dataset, but it becomes very specific in its knowledge about whatever that situation is. So that does determine the body of knowledge that it draws from. But on top of that, the way that you ask the question does determine how it responds. So the prompts are usually the ways that we tell the application what we want to do. So just now, when I typed, Hey, give me the summary, well, I use one of the preset prompts, but that was a prompt. Give me a summary of the document that is a prompt. So when you are generating or putting in prompts rather, you want to start with a specific goal for what you want copilot to do, provide a source to ground the response in a specific scope of information at context to maximize the response appropriateness and relevance and set clear expectations for the response. And then you can iterate based on the previous prompts and responses to refine that result. So that's where that chat component comes in. So you want to be very clear and succinct while giving enough context. So let us so an example of a prompt that would give us a good response would be this one, summarize the key reasons for traveling to Jamaica for an onshore traveler. Format the document as no more than six bullet points with a professional travel agent tone. So right there we're kind of following the rules. I'm telling it what I want to be done. I'm telling it what the context is, you know, who the target audience should be, how the document or the response should be formatted, as well as the tone, the overall tone that I would expect it to be brought across in, right? So when we do send a prompt, usually the co pilot will augment the prompt with a system message that sets conditions and constraints for the language model behavior. So those would determine the style of the model's responses. We also have a conversation history. So when the prompt goes, the history kind of goes with it to say, here's what was being discussed before, so take that into consideration when you are responding. And then, of course, the current prompt, which might be reworded with, you know, additional data or scope. But you don't have to worry about those things. Those things are happening in the background. 7. Understanding GitHub Copilot: Now we're moving on to the co pilots that I'm sure every developer is really interested in or most interested in, which is Github copilot. This is the world's first scalable AI developer tool that can help you write code faster with less work. You can think of it as an AI peer programmer. It draws context from the comments and code that it sees within your IDE and the project at hand and can suggest individual lines or even whole functions, which will, you know, speed up what it is you're trying to accomplish. It does help you to code faster and to focus on bigger problems so you'll feel more fulfilled at the end of the workday. And it's powered by the Open AI codex, which is trained on a dataset with a larger concentration on public source code. So that means all of these suggestions and everything that it's sending you is really coming from public publicly shared code and other projects that may or may not be similar to yours. But we all know, once you build one project, you can build several because a lot of the fundamentals are the same across many projects, and it does draw from all of these fundamentals to suggest to you what you can do in your program. So Github co pilot started the wave for AIPerPgrammer applications. So there are several others. You'll hear other names like Cursor, and there are several other tools, but the fact is that everybody's going to prefer a tool for their own context. But Github Github Copilotor is the foundational application that got this whole new wave started. No, it is the thing that Github copilot is just another copilot to help you write documents. I mean, blocks of code are stored in documents, but it is actually more than just another editor assistant. It does have features that make it a great assistant through the entire development cycle. For instance, it does feature a co pilot chat, which is a chat interface that focuses on developer scenarios and natively integrates with Visual Studio and Visual Studio code. There's also a co pilot for pull requests, which can draw from your code changes and your description for a pull request and generate tags that best suit that pull request. And this is very useful. It can be very tedious when you have to do pull requests to sit on and think of every little detail. So this can be very useful to help to speed up that process. And there's co pilot for Git CLI. So, I mean, the best of us, we are going to forget some commands. We are going to write commands incorrectly. We're going to forget parameters. Having the copilot right there will kind of remind us that, Hey, you need this. Hey, I suggest you put this if that's what you're trying to accomplish. So it is more than just something to help to finish your lines of code. It does help with various parts of the development cycle. So now that we've explored at a high level what Giub copilot is, let us look at a demo. 8. GitHub Copilot Sign Up: Alright, so before we get into signing up for Github co pilot, I just want to take a step back and discuss with you for 30 seconds what Github is. I'm sure we all know what Github is. At this point, I'm sure we know. But if you're not sure what Github is, it is one of the largest, if not the largest Git hosting providers, and it is free for individual use, and it does offer corporate and enterprise plans. However, as an individual developer, it is free for you to go to github.com. Go ahead and sign up very easy process, and you can get your account easily. So I suggest that if you don't already have a Github account, you go to github.com and create that account. Now, we're here for Github, C pilot, which is that AIPAPgramming tool, which is powered by what I guess, data that it has been trained on based on open or public repos on GitHub. So that means even if you're signing up as a company, you don't have to worry. The copilot is not using your private repos to inform its training, right? So Github co pilot, you can go to github.com slash features slash C Pilot. Or quite simply, if you're on github.com already, you can just click product and go to GitHub copilot. So once you're on that page, you'll want to get started. Now, it is not free, but they do give you a 30 day trial. So you can get started with copilot, and you'll see here that they are the individual plans, and you will be required to provide some payment information to create this account. So you can go ahead and start free trial, and you can authenticate with whichever account you need to. And once you have completed that step, you can choose which plan you want. So do you want to pay $10 a month or just pay $100 for the year? So I can appreciate that $10 a month is a bit easier to work with. So you can get access to get up copilot. You don't have to pay the same time. So they will take payment information, but you have up to 30 days to use it for free. Once that period ends, then you will be charged. So that means at least for this course, if you don't plan to use Gitub copilot beyond this course or immediately after, ensure that you cancel before the 30 days are up. So keep it for our demos, and maybe for the rest of this course, also keep it and test it out and see if it's something you really want to invest in. But if not, make sure that you cancel before the 30 days are up. So you want to fill out that form with your personal information and then go ahead and provide your payment method, whether it's by debit, credit card, or papal. And once you have confirmed all of that, you can go ahead and save and continue. And once everything is verified, you finish up with some general contact information and you submit your application. So once you have done that, now you have the different policies. So you have Github Sorry, co pilot in github.com. So you can use Copilohat inside of Github, and this can help with pull requests and other preview features. You have the copilot for the CLI, you have the chat in the IDE, chat in the mobile Github app, and you have several other things. So you can just go ahead and save if you're okay with those. Oh, I'm sorry. So I have to select suggestions matching public code, and I can allow that and then save the setup. Now the next thing would be to install the co pilot extension. So there is support for Visual Studio Code, Visual Studio, Jet Brains, and Neo VM. Alright? So when we come back, we're going to look at how we can set it up using Visual Studio Code. 9. Using GitHub Copilot: Alright, so we're back in Visual Studio code, and we want Github copilot. So I went over to the extensions tab, and you'll see here that it is recommended. It's recommended for me. Of course, if you don't see it in the recommendations, you can always search copilot or Github copilot. So I'm going to click on Github Copilot and also make sure that anytime you're installing an extension, you verify the author. So this is coming straight from Github. And I can go ahead and install after it has been installed, it's going to ask me to verify who I am with my Github account. And well, I'm already signed into Github on my machine and within the context of Visual Studio code. So that part was kind of done automatically, but you might need to do that by yourself, alright? So you'll see this little I guess, that's a little icon with goggles, a little face with goggles. That same copilot emblem now appears in the bottom right hand corner of your Visual Studio code. And from here, it's letting you know that it's ready. You can do a chat. You can view logs, you can do several things. So I want to chat with copilot. So I could actually use the little emblem and say, copilot, Get up copilot chat, or I could just click on this one and say chat with copilot. Now, obviously, you won't always have this co pilot page. So you'll want to use the little context menu as much as possible. So you can go ahead and open up that chat and let us see, help me to create a new minimal API project using the.net eight CLI. Let's see what that does. So it's going to let me know that it can so it gives me a step by step. Outline of how I can do that. So it says, open the terminal in Visual Studio code. So let's actually follow those instructions. So I'm going to open the terminal control and apostrophe just in case you've forgotten. I went to change before I do anything. I went to change over to my projects folder. Projects CDU sorry, CD CrivePjects. There we go. Not bad. All right, so now I'm in the context of where I can create that project. So now I can just copy. And, I mean, you're probably saying, Okay, so far, this is much better or different from using GPT task for instructions. The real benefit here is that it's here inside of is your studio code. So productivity wise, you don't have to be jumping in and out, jumping in and out, but, I mean, that's probably minimal when you consider that one is relatively free. One is going to cost money, right? But let's continue exploring what the copilot can do for us. So now I have my minimal API project created. Of course, they gave you a template. So if you wanted a different name, then you go ahead and change that name. But for now, I'll just see the end to the minimal API project, and I can say code that, so it opens up. In Visual Studio code for me. So now, I do have this project, and I'm going to switch over to this new Visual Studio code window. So this is the one I want to work with now, right? Alright, so I have my solution explorer with my minimal API project. Let's see what else we can do here. All right, so let's get our hands a little dirtier now. So I close the chat and you can always reopen it here. Alright, you get a little icon. So I don't want the chat right now. I don't even want to see the explorer. Let's focus on the code. So I'm going to try to create an endpoint that says Hello world. Now, you're saying, Alright, how do I engage co plot? Well, you just start writing code. So you're going to get code suggestions from the IDE naturally speaking. Sure. But then if I do this, you'll see that if I say map, get oh, look at that. I'm getting an auto completed line for me. So I'm going to press Tab to use that auto completed line, and it fills in the rest for me. Now, what if I wanted to generate a line of code. So let us say, endpoint to return list of numbers 1-10. So this is a comment, right? So based on that comment, co pilot is looking and saying, well, if that's what your upcoming code is supposed to do, then maybe this is the code you want, I can press tab, and I have an endpoint called numbers that is just going to return an array of numbers 1-10. And it's that simple. So just keyword wise, whenever we see those suggestions appearing, that's called ghost text. And you can always accept it by pressing tab. You can actually remove it by pressing Escape. And, of course, there are times when it will try to auto complete what it thinks you're about to do. It's always trying to anticipate your next move, which may not necessarily be accurate. Now, in a situation and let us say I want an endpoint to let's not say return a list of numbers, but return a random number 1-10. And then I go on and then it's now making that suggestion. Of course, I can press Tab to accept it. But if I'm not quite sure that this is exactly what I want, I can always hold on Control and press Enter to get the suggestions pan. So it's now loading other suggestions, and it's saying, Alright, so here's another way I can do it. I can Well, here's one way. This is suggestion one. Here's another way Suggestion two. So this is kind of a very simple task. This probably won't give enough suggestions based on what I'm asking to do. Right? So don't judge it too hard, but if you have a more complicated function that could be written in several different ways. So here's even suggestion five, where it's changing up the way that the endpoint is, and then it's changing up a little how it goes about finding what is random, and then it's even generating calculator like methods and endpoints for me. Wow, that's really cool. So what if I was to accept suggestion five, then look at that, it automatically fills in all of these other methods, right? So you see, it can speed up what you want to do because at the end of the day, you would have had to sit down and write out all of these in a bigger application in a more applicable scenario, of course. There is code that you have to sit down and write manually, and copilot knows this because it's trained on the fact that people have been trying to write this kind of code before you. So it's saying, well, you know, this is probably what you're trying to do, let me help you with it. Another way that we can interact with copilot and we can do that. I'll just remove this Bt code is I can do, oh look at that. Is suggesting reverse sentences after reverse words. Alright, why not? Let me accept it. So I can do an inline chat as well. So if I do Control I, I get that little prompt right here to ask copilot something. So a simple endpoint, and I'm kind of running out of simple endpoints, now. A simple endpoint that registers a user with email and password. Let's go ahead and see what it does, and it goes through and it generates, and I can accept this or I can discard it. So I'll accept, and we just generated code using the chat. Alright, so let's look at another way that co pile can work. So I just went ahead and generated a new method block called login, and there's an error here. I can quickly call on co pilot to fix this error by going to that little light bulb, and this shows code action. So let me highlight the erroneous code and then go to the light bulb. There we go. And then I can rewrite using copilot. I can fix using co pilot. I can explain using copilot. So let's try fixing it using copilot, and then it's going to generate the suggested fix, which is not in the code once again until I click Accept, or I can press escape or discard to not accept it. So in a nutshell, that is how Github copilot works. It's right there inside of your IDE, helping you along, making suggestions. And generally speaking, anytime you see that little sparkle, I don't know if you'll see that little sparkle, you might see it in the CLI, you might see it in the editor. But once you see that, it means that this is a Github copilot powered feature. So feel free to use it, see how best it can help you. And yes, this is Gen AI for.net developers. This is GitubcoPallot, within the context of a.net application. But clearly it's there for Visual Studio and Visual Studio code, which means that it will help you with whatever type of code you are writing in the moment. So feel free to use it for your JavaScript and for your SQL statements that you might have to write in those IDEs, at least, but it is there to assist you once you have it. 10. Building a Copilot: Alright, so we see that we have co pilots that are pre trained for specific scenarios. So we just saw Gitub copilot, which is trained for coding scenarios. We've seen that we have Microsoft copilot, which is more of a large language model which knows a little bit about everything. And then we have more specialized ones for Azure and security and so on. But once again, you as a developer might need to develop your own. So Microsoft provides two tools that help us to develop or extend existing copilots. Co pilot Studio, which is designed for low code development scenarios where, you know, you're not necessarily an IT person, but you are technically proficient, and that's usually high end well, technical business users or developers who don't necessarily want to write too much code for a particular scenario. You can use that to create conversational AI experiences, and you do have Azure AI Studio, which is a pass offering from Azure, which is a development portal for more professional software developers where you get total control over the language model you want to use. So we'll be taking a look at this later on in the course. But for now, just know that this co pilot feature is very powerful, very extensible, and you have the power to create your own. 11. Conclusion: Alright, so we're at the end of another section. And in this section, we review generative AI and some principles that surround it. We looked at different development tools and how co pilots can be used to help us to make our regular tasks faster. We also looked at Github copilot, so we took a developer look at how copilots help us to be better developers. So thank you for joining me in this section. I'll see you in the next one.