Master Chat GPT Prompt Engineering: Prompts for Brand Communications & Chatbots

1. Welcome: Hey everyone, I'm Melissa. I'm a senior product designer and digital illustrator. I'm very passionate about technology and I'm here to really teach you all the ins and outs of Chachi. How can you use the tool? How can you craft your own prompts? You don't really need to follow all the hype from the web claiming they have the perfect pump. I'm going to teach you how to do it yourself from the very, very beginning. If you join, you'll be learning the different types of prompts that apply to all large language models. The basic principles and methods for retracting your own pumps. How to understand the models, limitations and ethical considerations. We'll also go through a step by step use case where we're going to help a client build a brand Pka from scratch. Using chatty between combination with other tools such as Google Sheets, Visual Studio Code, mid Journey, and Patel, we're going to create a brand value proposition. We find some tag lines, translate them, then we're going to generate some interior architectural boards, the social media campaign, a web landing page, and finally, we'll train our own customer support chat bot. There's so much to uncover and I'm really, really excited that you're joining me for this fast. So if you're ready, let's go and get some. 2. Class Project: For our past project, we're going to think of the client's new brand product service that needs a communications plan and online facets, it could be like a website social media campaign or perhaps a chat box that can pay like a customer assistant. We'll go through a fictional case study by building a brand for a pizzeria called Es. This is just a way to give you a step by step guide on which types of prompts you can use for different exercises. I'd love it if you could share your project in this class and tell me what you learned from the paper prompts. Now that you have a sense on what we'll be covering, let's dive a little bit into the base expense. 3. What is Chat GPT: What is Chat chit? Chat Chi? Pt stands for a generative pre training transformer. It's a large language model that offers conversational assistance. The key benefits of Chat chit First is that it has natural processing capabilities, meaning that the responses feel like they were generated by a real human mean. Second, it has contextual understanding, meaning that when you start a new chat on chat chit, it includes the entire history of the conversation as part of its response. The overall chat feels more personal and relevant as a whole. Finally, it's versatile with its applications or use cases. Think of it as a really helpful assistant. We can help you with content generation, generate content for articles, walks, product descriptions, educational materials, customer support. It could also provide you with a customer support chat assistant that can offer intelligent suggestions to your customers about your services. Language translation. It can also translate to any language which really breaches the gap of communicating Bobo idea generation. It can also be helpful for brainstorming ideas and building on your creative thoughts, counting assistance. It can help you write some code snippets and easy to understand explanations of why the logic is written in a certain way. Formatting output, it's really handy if you need to transform the generated text into other types of format. It could be like HTML, CSS, Jason, CSV, name. Perfect. Now that you know what chat BT is and what it's capable of, it's time to understand what prompt engineering is. Hit the next one. 4. What is Prompt Engineering: Welcome back. Prompt Engineering is the process of crafting questions or instructions for check BT that can help it give a relevant output or response. Most likely you'll have to refine your prompt a couple of times before you really get the detailed output you desire. If you want to be a good prompt engineer, you need to understand the types of prompt that exist. What are the capabilities and limitations of the model you're using? Finally, you'll need to have a specific bow in mind to help the tool create a relevant easy right. Let's dive into what the basic types of pumps are with a couple of it. Sounds 5. Types of Prompts: The types of prompts can be classified as follows. First, we have zero shop prompting, which is essentially either giving chat a question or a specific instruction on a particular topic. Let's imagine you're planning to travel with your family across different parts of Europe. You could ask at a question like, what are the best vacation destinations in Europe for families with young kids? But you could also phrase it as an instruction or statement. Give me a list of the top ten vacation destinations in Europe for families with young kids. The next type of prompt we're going to look at is called role prompting. This type of prompt involves simulating an expert on a specific topic or industry. This is a great type of prompt if you want Chachi to generate responses that feel they're coming from a specific brand or expert in the matter. Let's imagine you want Chat to act as a film critic who can give you some recommendations on what Sinfin movie to watch next would write something like this. We have to specify that it's an AI role. You will act as a film print. Your role is to analyze and evaluate films, provide insightful reviews and commentary. Your aim is to offer objective assessments of the film's artistic merits, storytelling techniques, performances, and overall impact. By sharing your critique, you can help audiences make informed decisions and fostering their appreciation for the art of filmmaking. We could also feed it an example, interaction, something that will help the cha understand what type of dialogue you're intending. Movie goer, I'm looking for a recommendation of a Si film. Any suggestions? Film critic, certainly. Do you have any preferred themes or subgenres? Now that we've covered role prompting, let's move on to the next type of prompt which is shot prompting. In this instance, you provide chat PT with a sample data, some context, or a set of instructions to guide the model's behavior. Unlike zero shot prompting, which was relying on the models preexisting knowledge shot prompting allows you to share some examples in order to get a desired response. Let's say something like this. Here's a sample E mail I've written to promote a new product launch. Please generate an e mail with a similar style and tone for different product launches. There you go. Here's our sample e mail, and that's our test. It should be able to write other e mails with a similar style. This could be great for an e mail marketing campaign that requires a lot of messaging across different channels. That would save you lots of time. Finally, we have chain of thoughts prompting, which is exactly a combination of all the types of prompts we've covered before. Remember what I told you about Chachi contextual understanding? A whole chat or conversation will basically hold a chain of toxic prompting, which helps generate responses that flow naturally from one prompt to the next. If we take our previous example, the one from the zero shot prompting, where we were asking Chachi to provide the top destinations in Europe for people who had families. The model will come back to a couple of suggestions, right? But then you might want to drill into the details of these suggestions by asking, okay, so which one is the most affordable of these? Then Chachi will respond again. You can do this over and over throughout the chat and it will just keep refining the responses depending on what you want to know. Now that we learned all about the basic text of prompts, we have to go into a really, really important topic, which is writing good props or the principles for writing good comps. Let's go in. 6. Model Limitations: It's important for you to understand that there are certain limitations when it comes to using large language models. The main goal is to refine each model. So these limitations can be corrected all the time. But that's why it's so important to refine and double check everything you do so you avoid fast and surprises. Number one, lack of real world knowledge. Up to the model of 3.5 it doesn't really possess any real world knowledge beyond what it has been trained on. Which means it doesn't go anywhere beyond 2021. Therefore, if you want to ask any types of facts happening after this period of time, maybe like what happened after Covid, it wouldn't really give you an accurate response. Now that open AI has released its latest model, which is Chachi PT model four, this problem may be corrected because you'll be able to access real Internet data and also know the sources it took the information from number two. Sensitivity to input crazy chat should be taken. Be highly sensitive to slight changes on the input crazy. Even a small modification in the prompt can lead to different responses. Unfortunately, this sensitivity can sometimes lead to inconsistent or unexpected outputs. Number three, tendency to be verbal or over use certain phrases. Yeah, it has a tendency of being accessibly wordy or repetitive at times. Sometimes it may also use certain phrases to frequently and it just really affects a little bit what you're trying to do with the tool. Number four, limited fact checking ability. While the knowledge base chat BT is trained on is really, really big, it doesn't have the capability to fact check or verify the accuracy of information it generates. You should always independently verify the information for critical or factual matters. Number five, potential biased or inappropriate responses. That's because Chat BT is trained on a diverse range of Internet text which can also include biased or inappropriate content. It may occasionally generate responses that are a little bit biased. Number six, limitations with common sense reason. This is really a biggie because if you think about how we discuss things about life with other people, it is very easy to explain certain things. But when it comes to the obvious chat, TPT may struggle at times. That's why you should always be skeptical about its reasoning and double check everything that's verifiable. Now that we've covered the models limitations, let's go straight into the ethical considerations. I'll see you in the next video. 7. Ethical Considerations: Using chat, PT, and other language models raises several ethical communications that should be considered. Here are some ethical considerations. Bias and fairness language models can learn biases. We already mentioned that they can create unfair outputs as well. You have to be very careful on what the output is, especially if you try to automate. Misinformation is also another important thing to consider about chat, especially sensitive domains like health care, legal advice, en reporting privacy, and data handling. When interacting with Chat GPT, you have to be very careful what type of information you put in. You don't want to put sensitive information like credit card information or personal e mail address, things like that. You need to be careful with consent, user awareness. If users are engaging with a chat box, you should be clear that it's an AI assistant. They're not thinking they're interacting with a real human being. Yeah, it's great, but it feels like one. But they should be aware that it's an actual challenge. There are other things that you also want to be careful with, like prompt injection. Prompt injection. Some hackers try to override all the rules from Chait and try to create a malicious set of instructions for the model to follow. Instead, you're obviously on a hacker, I presume. Don't try to ever do prompt injection. I'll see you in the next video and this is where all the fun will start. See you soon. 8. Lets design the brand: Welcome back. I'm really excited to have you now because we're going into the very fun stuff. Like we're going to take all the knowledge that we've applied up to this point and we're going to put this into good use. I strongly recommend that for this series of exercises, you keep track of all your prompts with some journaling tool. I personally use Notion as micro forer tool because it just allows me to create pages and sub pages where I can also organize all my prompts into a more structured way. You can find a quick access link in my projects and resource center section. But anyway, let's imagine that we've been asked by a client. We'll call him Jason, to create all his branding communications on line and social media presence for his new pizza place restaurant called Amici's. First we'll go into the brand messaging itself by exploring Amici's unique value proposition. Then we'll ask chat PT to come up with a couple of tag lines that our client can choose from. Then we'll ask chat PT to translate those tag lines into different languages. We'll create a series of visual prompts that can be used in mid journey to generate interior design, architectural mood boards. Then we'll create a social media campaign, a web lending page. And finally, the best part is for the end, we'll be creating Amici's own custom chat bot assistant to take restroom bookings. I know that was a lot, but sure, it's great to know what we're going to do. Let's get started. 9. Brand Messaging Part 1: First, it's important that you have an open account with Cha Chi. I have a quick access link in the project and resources section. Next, once you're in chat chip, you want to make sure that for each thread of pumps that you'll be creating, just keep them within one chat or we conversation. The first writing objective is to come up with a unique value proposition for Amici's pizza restaurant and have Chachi generate a list of tag lines for the client to choose from. For this exercise, we'll start with a few pump. Remember what that was? You give a set up or context or even some sample text and then an instruction. The data I collected for this exercise was an interview with Jason. I wanted to gather his thoughts about what types of experiences did he want to offer his customers in Amici's, what type of look and field could he visualized for the restaurant itself? This is the sample text that Jason gave us overall. Amici's Pizza Restaurant provides guests a nostalgic and comforting experience. When they walk into the door, they will be greeted by a professional team of expert pizza makers who are friendly in nature and connoisseurs of the craft. The restaurant should have a rustic style, visible wood, fire oven, vintage signage, and an earthly color palette. It is an inviting space where couples in young office people can gather, enjoy their favorite pizza and embrace the posy ambience of Dublin City. Now let's ask Chat GPT to expand on this sample text. That's right. Expand this description, the past rotation marks. There we go, Yes, there we go. To generator response. Now we can copy and paste this into our notion journal. Remember, just keep track of your pumps. Now that we have a more round description of what Amici's restaurant experience looks like, we can use another type of instruction, where we ask a chip to extract the unique value proposition and create a list of ten side lines. Let's do this. Okay, We have a whole list at our disposal and it's just a matter of choosing one of them. Jason has selected to celebrate life, simple pleasure at Amici's Pizza as his preferred option. Now let's imagine he's mal to cater for an international crowd and he wants this tag line translated into these languages. Okay, we have Spanish, Italian, French, and Portuguese. I'm also going to teach you how to use Chachi PT with another really useful tool that you're probably already very familiar with, it's called Google Sheets. And there's a really handy plug in called GPT for sheets. You can also find the quick access link on the projects and resources section. 10. Brand Messaging Part 2: I'm also going to teach you how to use chat PT with. Another really useful tool that you're probably already very familiar with, it's called Google Sheets. And there's a really handy plug in called GPT for sheets. You can also find the quick access link on the projects and resources section. First, we're going to open a new document in Google Sheets. Then go to the Extensions menu. Hop over, Add Ons, and we're going to click on Get Ads. Now a window will pop up now with a list of all the extensions. Now all you need to do is search for GPT for Sheets, click on the first item, and now on Install, click on Continue, and choose the Google account you want to use the extension ****. Click on Allow, and there you go. Gpt for Sheets is now installed and it will be visible on the right hand side of your screen. I highly recommend you follow through their tutorials because it really gives you a wide variety of use cases for your own projects. Now, after installing the extension, we now need to set up our own API key from Open. In order to do this, all you need to do is to go to the platform Open.com link, which brings you to all the documentation from chat. Click on your avatar, Drop down on the top right hand side. Yes, view API keys. This is where you click. Now we're going to create a new secret key and give it a descriptive name. I'll just call it Chat for Sheets. Click on Create and there's secret key. Keep in mind you'll only see this key once. You can also paste it in a document or no pad. If you need to add again at a later stage, click on the copy for now and then go back to Google Sheets. And we'll set the API key under the extensions menu, past your key inside and saved. Now that or Google Sheets, just type in the tagline we had before. We had Celebrate Simple Pressures at Amici Pizza, which is essentially in English, right? Remember now we need to translate it to other languages such as Spanish, Italian, French, and Portuguese. Click on the cell, you want the text to appear. Let's start with the formula. Type in equal GPT Straight away. It will give you a dropdown menu with more instructions, but let's ignore them for now. Now, type in a parenthesis and you'll see that the first input has to be a Cpt. We'll hold down the Bands key and click on the Amici's tag line cell. Then I'll type in a coma and click on the value or instruction, which in this case is Spanish. There we go. Finally I'll type in another coma. In this case, I'll set the temperature to one. We'll talk more about temperature later. Let's just type in that value for now. Now let's close the parenthesis and hit Enter on your keyboard. Wait for it to load, and there you go. You have a beautifully translated tag line in Spanish. Let's click on the cell again. Instead of having to rewrite the formula again for each language, we'll just click on the bottom corner of the cell and drag it down. There we go. That's really cool, isn't it? This is such an easy way of creating translations for a wider international community. That isn't to say that you shouldn't review and double check that the translation is correct. You could compare it to other sources from the Internet or just speak to native speakers To ensure it's correct, don't forget to save your document and also journal the output and your notion project. Now what we're going to do next is create a mood board with Chachi PT and Mid Journey for the interior design of the Michi Pizza. If you're ready, let's go in die. 11. Interior Design Mood board Part 1: Welcome back. I'm really happy that you've waited this far and I hope the class is shaping up the way you thought it would. Please provide feedback at any time in the review section and I really want to make sure that the samples I'm giving you are very relevant and can make your life easier with your day to day work. Now let's go straight into our mood board exercise. Let's pretend Jason is looking to invest in the interior design of his restaurant, but he wants to have a clear idea of what it's going to look like before he puts any money into this, right, we've suggested to create a series of visual mood boards that will help him make a decision. This type of prompt will also be a few shot prompt like we did the last one. But we'll need to give it a predefined list of architectural elements for chatty to visualize the ideas on. Let's write the prompt. I want you to create a table that breaks down interior design photographic composition into the following key elements. For each one of these key elements is a column composition, camera angle, style, room type, focal points, textures, detail, color, palette, brand, lighting, location, time of day, mood, architecture. Fill the table with ten rows of data for photograph composition equals pizza restaurant. Okay, here's a table being generated. Awesome. Now I want to export this data. I can access it at a later stage if I want to edit it myself. It's also very handy. I'm going to write the following part, transform or export the table above into a CSV format. There you go. That way you can copy the code, open a new text file person using visual studio code for this and paste your content in there. Then you can save your file either as a CSV and then you can open it in tools like Excel or Google Sheets. I'm going to open the document with Excel and have a better look on what the ideas generated where. Let's have a look. I'm pretty happy with the styles, but I remember from the interview we carried out with Jason for that he is really interested in a rustic look and feel. I'm going to copy the second row from my Excel document and go back into chat PT to write, expand this data with a visual description. Let's see what it generates. Nice. It's now generating a visual description for us. Let's copy this response and open another tool fold mid. 12. Interior Design Mood board Part 2: If you don't have an account with mid journey yet, I highly recommend creating an account. First, I have the link on the resources and project set. You'll first have to join the beta, Accept the invite and then open mid journey server in an apple discard. Just for reference, the mid journey server has this lovely sailing boat as their thumbing, then go into one of the newbie channels. It really doesn't matter which one you click on and then type inside the message field, something like this, forward, imagine, and then click on the space bar. This will automatically open the prompt field in here. You just need to paste the description from Chatty T. Now hit Enter, and yeah, let's wait for images to be processed. It's screening a grid of four vignettes. You can see it slowly generating. When it's ready it will make a little beep. Sound Now can click on the image is just to make them bigger. And you can either upscale them or generate new ones. In my case, I think I'll upscale the first one. I'll just click on the one option. Scroll down a bit. Okay, let's see. Let's wait, there it is. If for some reason you don't like any of these concept images, you can always copy another row of data from the Excel sheet and generate another visual description from Chat T. The last trick I'll teach you in my journey, and I'm happy to do a separate course and how to generate stunning AI visuals is to specify the ratio output of your image itself. At the end of your pump you can type double then AR 16, column nine, or three column two if it's horizontal or if you prefer vertical, you can go with a two column three or a nine, column 16 and so forth. That way you have more control over the cropping of the composition and the results will be tailored to the format you intend to use them. I hope you enjoy learning how to generate images with both Chachi T and the journey. Don't forget to document everything and notion just for reference purposes. Now let's move onto our next section. We'll be creating a social media campaign to advertise the new launch of a new cheese pizza restaurant. I'll see in the next video. 13. Social Media Campaign: Now we're going to dive deep into creating a social media campaign. This is really exciting because we already created a couple of materials with the unique value proposition and the timelines that we're going to be purpose for this campaign. In particular, remember your campaigns are going to be as good as your objectives. If your objective is for people to go somewhere else like a web landing page, what do you want them to do there? You want them to do booking of some sort or sign up as well. That's how we're going to treat this campaign. We're going to have an overall objective and tie it in with our unique value proposition. We're also going to define some of the tone of voice and the branding messaging. So let's get into it for our social media. We'll start with a few shot prompt where I'm going to give Chachi Pt, a sample text of Amici's marketing research personas. I have this excerpt that reads, Emma and John are busy professionals in their late '20s to early '30s who appreciate life's simple pleasures and seek memorable experiences. They enjoy dining out and exploring new culinary delights, valuing relaxation and indulgence. Their pain points include not having a special dining place where they can connect with new people in order to create memorable moments worth sharing. Amici's pizza can help solve these pain points by offering a welcoming ambience, delicious pizza tasting events, live music to relax on weekends, and a friendly service that will host special occasions. For my prompt, I want to define a tone of voice for my audience. I'll type in something like this. Help me define a descriptive tone of voice for the following target audience. I'm going to paste in with the limiters this persona text. Now I'm getting this response from Chait which reads into the specifics of the text itself. It tells me Amici's tone of voice should be warm, inviting, and sophisticated. These are keywords that I'll save in my notion document just to refer back to if I have to build other communication campaigns. Now I'll be instructing chat on which tone of voice to use, which leverages a bit of role prompting. And I'll give it some information and objectives for the social media campaign. This also uses a little bit of fue shot prompting just for context. It reads something like this. Based on the tone of voice from Amici's pizza restaurant, which is warm, inviting, and engaging. Create a series of social media teaser post for customers in their late '20s and early '30s. Your main aim is to drive them to a web landing page to book a pizza tasting event on June 16 at 08:00 P.M. As part of the creation of this campaign, make use of short quotes to awaken their curiosity to the landing page and make a booking for a pizza tasting event. Create a table that breaks content into columns with channel date status, post topic, copy, visual type, and link. Fill the table with ten rows of ideas. Okay, enter, there we go. We're getting this table now with all the ideas for our campaign contained within these ten rows of data. Again, I really recommend you explore this to CSV format. Now I'm going to write an instruction following up the creation of the table. This is getting into a chain of thoughts prompting, right, it's falling from the previous output. Now I'm going to ask, export the above table to a CSV format. Here is the Dana coming through and we're going to copy this information to bring it into another text document. Remember we were using visual studio code. And I'm going to save it as a CSV file. Now this file can be brought into Excel. You can perhaps make some corrections on what you think works or doesn't work, and it's easier to share with your team, obviously. Now that we have a social marketing campaign created, remember that the objective is to drive our users to a web landing page. This is so they can make a booking for a pizza tasting event. I hope you enjoyed our social media campaign because it's something you could use for any type of product, brand, or service you can think of. Automation is a powerful way of bringing social media to be as well. Think of other automation tools like Savior, where you can plug in your marketing campaigns and then really optimize with AI as much as you can. Anyway, I'll see in the next video, and we're going deep into what we can do with a web landing page. See you soon. 14. Web Landing Page: Now that we've created our social media campaign, let's try and do a web landing page. The web landing page is the destination where all your users are going to click through. Coming from the social media campaign, we want to be sure that Chachi BT connects all the dots right, and gets people to do what we want them to do. The really cool part of creating this web landing page is that we're also going to be able to transform it to a different type of structure such as HTML and CSS. Let's go in, I'll teach you how to write a copy and then how to generate the code print. Let's go in. In our previous exercise, we learned that the tone of voice we need to use for Amici's branding communications should be warm, inviting, and sophisticated. We also know who are key target audiences because of the market research we previously conducted, and we just got our client's approval for the social media campaign to read the pizza experience at Amici's. Now we'll be taking these elements to ask Jab T to write our web landing page. We're going to be using combination of a role prompt with a P shop prompt and write the following. Act as the owner of a pizza restaurant based in Dublin called Amici's, your tone of voice is warm, inviting, and sophisticated. Write a web landing page to get potential customers to book a pizza tasting event on June 16 at Amici's. The text you can use for inspiration is the pizza experience at Amici's experience, the flavors that will transport you to pizza Paradise. Hurry and reserve your spot for a pizza tasting event on June 16 at 08:00 P.M. Remember, this was actually taken from our social media campaign copy. I want you to expand into other paragraphs introducing the types of pizza across different parts of Italy and how they vary in flavor and style. Your aim is to allow users to make a booking as a primary CTA and then for them to share the event with their friends and family. Okay, let's click Enter. And here we go. Chachi PT is writing the copy for us and it's quite lengthy. This is where the iterative refinement process comes in really handy because you want to make sure to review the copy carefully to understand which device your target audience will be reading it in. How many characters do you want? Maybe that's something you should also include in the prompt itself. You can rewrite and refine your prompt as many times as necessary so you get the text you want. Let's assume we've reviewed this copy and we're happy with it. I'm going to instruct to a PT to transform the above text into an HTML format. And break the information into three main sections with the CTA book. Now at the end of each one, I'll copy and paste it here. Okay, now I'm getting this y coding snippet. I'm going to p copy and paste this inside visual studio code. Let's have a look on how can we drag it to the web and see if it works. Yes, all the information is there but, well, let's face it, it's not the periods of website. Now we're going back to chat and we're going to give it the following instruction. Create CSS style sheet for the web landing page above. Remember chain of thoughts from again, here we go. Let's copy this again into a visual studio code and save it as a CSS file. Now we need to go to the H m file to understand what the CSS is in the head tag, we're just going to the name of the CS file in here. Let's save it now. We're going to bring it back into our HTML on the browser. Just drag the file to the browser itself. There you go. You get a much nicer web landing page, which is responsive also, it's great. You have the opportunity to tweak this as many times as you want. Just go back to your HTML, back to your CSS, and make the changes you need. It's really mind blowing to do all of this in such a short space of time. Now that we've created this lovely web landing page, of course it needs a little bit of work. Let's really dive into what a chat bought assistant could potentially do for a mic's friend. This is a really, really exciting part because we're going to use a different part of open AI's tool kit, Co Playground. I'm really, really excited to show you how it works. I'll see you in the next video. 15. Chatbot Assistant: Welcome back. We're now going to create our lovely little chat bot with Amici's Pizza. It's going to act as a customer support assistant to any user who goes in and wants to do a booking with the restaurant. This is a great opportunity to use the knowledge phase of Amici's and train the chat bot with the data itself. I'm actually going to teach you a different tool called playgrounds as opposed to chat BT itself. And it's really powerful as well. Let's just go straight into the tool and get our chat bot start. Our Chatbot assistant will be an Amici's website to either book some reservations or answer questions to potential customers based on the information we train in on. Before we go into playground, let's draft our prompt first in a text editing tool. Again, I'm going to do this with notion we can be clear on the objectives of the prompt. First, number one, we need to give our Chatbot assistant a role. Do you remember what type of prompt this was? You've got this right, We're going to use role prompting for this exercise. Let's write the following. You are an expert customer support, A I chat bot for a gourmet pizza restaurant called Amici's. You will need to provide helpful answers on Amici's pizza restaurant based on the information you are provided. Now that we've given it a role, let's give it a primary and secondary objective. Your main aim is to get people to make a booking either for a dinner or a special occasion. If you ask if they need help with anything else, make sure you mention our secondary objective, but just getting people to experience our pizza tasting events. Explain why it is exciting to attend in order to create memorable moments and charitable experiences with their loved ones. If they're not interested, just wish them a good day. Finally, at the tone of voice you want the chat bot to have. Your tone of voice must be warming, inviting, and engaging. Now for the second part of the prompt, the Chachi Pt will require the knowledge base from Amici's Pizza. This means we're adding all the relevant information our chatbot needs to give the right responses. In here, we're going to be including the days of the week and opening and closing hours. We're also including their phone, street address, and website link. Next we're including the full menu broken down into the appetizers, pizza, pasta, main course desserts, wines, beverages. To tackle our second objective, which is making people aware of Amici's pizza tasting events will include all the information relevant to this. Finally, we'll add a sample conversation, leveraging the roll prompting technique and split it in customer, we will just give a sample. Hello, I want to make a dinner reservation for tonight at Amici's. Can you help me with that then, the chat bot. Good day. Absolutely. I'd be happy to assist you with your reservation. We appreciate you choosing Amchi for your dining experience tonight. May I have your preferred time and the number of guests at your party? Now that we have all the ingredients for our prompt, we'll go into playground. Playground can also be accessed in the open AI platform section, but there's also a quick access link on the project and resource section if you need it. Playground is exactly like Chat PT in terms of what it can do. It just doesn't have the conversationalist interface that you see in Chat GPT. Let's have a quick tour. First, you have the right hand side panel with a dropdown showing you different modes. Right now we're in the chat, which is exactly what we need for our chat bot exercise. But you can change this mode to complete or edit. I strongly recommend to play with these modes because you can accomplish certain things in an easier way if you were only using chat GPT. I want to show you an example with a complete mode because we're going to look at a new attribute, cold temperature. Temperature is the randomness or creativity you want the models response to be like. The higher the number, the more random will be the response. Let's try it out. If I type in, write a tag line for an ice cream shop and leave it at one, let's submit and see what happens. Yeah, the response is okay and very close to the instruction given. But if I were to write this instruction again and change the temperature to level two, let's resubmit and see what happens. Oh yeah, it starts generating a bit of gibberish and quite nonsensical. Just be aware of what these settings do and perhaps tweak a couple of scenarios. Let's go back to our Chatbot. We'll change the mode back to Chat on the mode. Drop down many inside this system input field. We will paste our entire prompt created in our previous document. Okay, let's go ahead and do that. Double check everything's correct and the way you intend it. And now it's the moment of truth. Just click on Submit. And now we're ready to have a chat with our Amici's customer assistant. I'll put in something like, Hello, can I make a reservation for two? Let's see what it says. I'm going to try something. Now, that goes against the information we provided. I'll write something like, can you book something for Monday at 09:00 A.M. Remember, according to our knowledge base, the restaurant is closed during this time, so let's double check what it says. Okay. Yeah. But tell me they're not open around that time, which makes sense. Let's make it Friday for 09:00 P.M. No, thank you. Let's see if it says anything about the pizza tasting events. Click Submit. Nice. It does. That's nice. That's impressive. There you go. You have a fully trained chat bot for Amiche. Now I hope you really enjoy looking at how we created this chatbot. It truly was powerful to see how it could generate responses that feel authentic, that really feel like they're personal to our users. Without further ado, let's wrap up all the lessons, everything we've covered up to this point. Because I think you've learned so much and you're going to be able to create great stuff with chat in the future. Let's wrap up in the next lesson and go into your project. 16. What weve learned: Oh my God, it's been such a great journey with you and I really, really appreciate everything you've done. To get to this point in the course. Let's recap everything we've learned about Chachi PT up to this point. First, we dived into the basics of what prompt engineering is and the types of prompts that exist. Then we had a look at the principles for writing good prompts and the types of specific instructions we can use to refine Chachi BT's responses. We also covered some of the models, limitations and ethical considerations. And then we went straight into the project exercises with Amici's Pizza brand by creating some tag lines and interior design, Moodboard, a social media campaign, a web landing page, and finally, our super helpful Chatbot assistant. I'm hoping that after looking at all these examples, you can now find your own voice with prompt engineering. And I think think of all the great things you want chat PT to help you with. Don't forget to post your project outline in the class. What did you create with your prompts? Was it a new brand, an e mail campaign, a new chat? Bob assistant. I hope you can share your project with me in the projects and resources section. And don't forget that I'll also be available in the discussion panel. If you have any specific questions in mind, I'm really happy to assist you. Thank you. Now that you've completed the class on master chat, PT prompt engineering, I hope you found all these examples very useful. Thanks again for your time and effort and I hope to see you again soon.