Speaker 1: This ChatGPT tutorial is a small crash course for any beginner, so we'll be doing a number of things, from creating a brand new account on OpenAI, to logging and using ChatGPT, to some of the more advanced ways to use it. If there's anything you want to know, this video should cover it all. Let's begin. I'm going to start on Google here, and I'm going to browse to find OpenAI. You can also head over to OpenAI.com if you want to skip this step. Here is the company that basically created ChatGPT. They've created a number of different types of APIs, and if you want to have a look at them all, you can head over to the API page. We're going to be looking at the ChatGPT version. If you want to head to that, you should be able to select the top nav bar here. This will take you to a page called chat.openai.com. If you don't already have a login with OpenAI, you can select to sign up and create a brand new one. For the sign up, it's quite easy. You can simply log in with an email address, or with a Google or Microsoft account. While I already have an account, I'm going to create a brand new one, so that you guys can see this process. I'm going to sign up with a new email address, and go through the prompts to create an account this way. This will require the email address, as well as a password, and a two-step verification process, which we'll do. The first thing that'll happen is an email will be sent to the mailbox that you're signed up with. In this case, it'll be sent to my Gmail address. I'm going to head over to that inbox now, and we're going to select the verification for email address button. This will basically activate the account, so we can log straight into the OpenAI systems. Now, the OpenAI dashboard isn't ChatGPT. It's a separate dashboard to start using some of their normal models here for their OpenAI systems. But to go back to ChatGPT, we need to select the link up here that takes us to chat.openai.com. Here, we'll need to log in one more time. And this should take us to the ChatGPT system. So again, I'm going to put in my email address, as well as my password in here. And this will now finally take me to ChatGPT. As this is the very first time I'm using it, I need to fill out just a few more details. So here, I'm going to fill out my name, as well as my phone number, which will verify my account. This will send me that two-step verification. So I'll put in the code I received on my phone number. And finally, we're now in ChatGPT and its systems. And we can start to use it. The first thing to know is that ChatGPT is free, and it's mainly used right now for research purposes. There are a few prompts worth reading here when you get started. Things like their goal here is to get some external feedback in order to continue to improve the system, as well as making it safer. They also have a few safeguards in place, because there are occasions where the generations are incorrect or even misleading. So be aware of that, because if offensive or biased content is produced, then you might not want to use the given advice. Let's have a look at what else is being used here. So we also have the collection of data. So the prompts and items that you are entered in are actually reviewed by their AI trainers in order to improve the systems. This means just take notice of what kind of prompts you're putting in, and also don't share sensitive information that might be happening inside your conversations, which means that if you're putting in passwords, which really you shouldn't be, try not to post screenshots of that. There's also an importance of feedback here, because as you do discover and utilize the system more, giving feedback can help them improve the system. There is a Discord server that you can join where you can share your feedback, and this one over here is actually accessible. So I'm just going to log into this OpenAI one, and it already has half a million members, which, wow, that is quite a few. You'll need to go through the prompts here. So visit the website, make sure that you read this, and then complete the little survey here, making sure that you're not going to be doing anything such as spamming or anything like that. So I've just applied here, and here we've got all our discussion threads and whatnot. So you can join as part of the community. Now, let's begin. This is chat-gpt. It actually has a couple of elements here on the page worth knowing. We've got the sidebar here, which essentially doesn't really do much. The main item here is the reset thread, which I'll get into shortly. Here we have a couple of example inputs that we can place, and the input is just over here. This is where you can type your text. We also have some details here about the capabilities and limitations. Let's take a look at all of this. So in terms of the capabilities, the chat-gpt system actually remembers what you said in previous conversations. Now, be aware that this is the conversation thread that you're currently having. If you strike up a brand new conversation thread, that won't be remembered, just the current one that you're in. It also allows you to do follow-ups and corrections. So for example, if you didn't get quite the answer that you're looking for, you can always ask for another one, or really just re-prompt for a different option. The AI is also trained to decline providing negative or inappropriate requests. So if, for example, you're asking for something malicious, most times out of 10, it will not provide that. However, there are odd situations where it might accidentally do this. And let's go to the limitations this time, to have a look at some of the problems that you might encounter when using chat-gpt. Firstly, some of the generations might actually be inaccurate. It is only trained until the last couple of years of information, so it's not able to, say, access the internet right now. It is in an offline state as of a year or two ago. It also doesn't do mathematics very well. So you might have some issues if you're doing things like 10 plus 9, it might not give you the correct answer. There is also some bias to the content that is produced, based on what you've produced earlier in the conversation. And this could essentially mean that it occasionally produces some harmful instructions, and those have to just sort of be ignored. And finally, as I mentioned earlier, there is limited knowledge, because the actual AI is trained for events just towards 2021. This means that it won't have up-to-date information of relevant and current facts that are happening in the world. With all of this said and done now, we can now actually start taking a look at using chat-gpt. The first thing to take note of is that using it is very simple. All we have to do is enter in a message, or a query, or a prompt of what we want the AI, chat-gpt, to do for us. And then we get an output. There are a few examples here with buttons, basically asking us maybe explain what quantum computing is, or how to make a HTTP request in JavaScript. But I'm going to minus these out, and I'm going to use my own, which is what is 10 plus 9? As a question mark. This is a pretty simple prompt. Then we simply select enter, or set the button here, and press and click on it. And we start with our very first prompt. Now here, we have two parts to this prompt. We've got what we've prompted, and then we've got the output here from OpenAI. Now this is pretty damn simple. But there are a few things to take note of here. The first thing is the question, what is 10 plus 9? Now we've got the output here that says 19, but we also have this option here to try again. So if I was to select that, what we end up getting is essentially the same answer, but just written in a different way. In this case, it's been written more mathematically, which is 10 plus 9 equals 19, but it's not just 19 like it was before. And we can continue to click on this, but if, for example, we've given as many results as we possibly can, then it might error out like you just saw earlier, or it might continue to give us better answers. Like this is a half mathematical, half verbal one, where it says 10 plus 9 is 19. We can try again, and this time it says 10 plus 9 is 19. So it's almost repeating now. You'll get to a point where there's only so many ways it can be written. We have a number of other things to take note of. We've got these thumbs up and thumbs down icons. And this is to give OpenAI some feedback as to the prompts we have here. We also have these arrows here when you hover over to show the different types of outputs that you've gotten based on the input that you've provided. Now, I've just done five of them, so I can toggle through them. And I like number five the most. So this one, I can give a thumbs up. Now, if, for example, it gave a poor answer, we could do the same by going back here. It says an error occurred, and I say no answer provided. So this, for example, is how you would provide feedback to the system if you needed to. Now, this is our very first prompt in this conversation. We can continue to make more prompts, such as saying, how good is OpenAI's chat GPT at maths? And ask this question. So this question is still part of this conversation, and we're getting an answer now being streamlined and generated from OpenAI's chat GPT. It says, as a large language model trained by OpenAI, it doesn't really have the ability to perform mathematical calculations. Now, here's where you might have to be careful. There are situations where the outputs might not actually generate the kind of consistent answers you might be looking for. Such in this example, I'm going to trick the AI, saying that it is actually 20. 10 plus 9 is 20. Now, the prompt will say that it is a language model. So what's going to happen now is I'm going to ask one more prompt, saying, is 10 plus 9 equal 20? And since I previously said that it was, now it's been tricked. It actually does think that 10 plus 9 is 20, and it's going to confirm that answer. This is where a confirmation bias might come into play, and you have to be careful what kind of prompts you're putting in, because they will influence the outputs that chat GPT provides. Now, this entire thing here is a conversation, and we could continue this conversation indefinitely, but we're not going to do that. We're going to start a new conversation. This is because all of this conversation is essentially structured in such a way that it's in history. It's essentially taking reference of what we've said earlier in the conversation and pulling that context into new responses. This is why we got 10 plus 9 is now correct as 20, which doesn't make much sense. In order to go back, we're going to reset the thread, and we're going to select the top left button over here in order to do that. It'll take us back to this main dashboard here, and now we can start up a brand new conversation. I'm going to ask the same question here. Is 10 plus 9 20? Which we know is incorrect, but we tricked it earlier that it was. This time, however, since the chat GPT system is not referencing that previous conversation, it's going to answer the correct manner. Even though math isn't the best example, this is just a good example to showcase that, no, 10 plus 9 is not equal to 20. It's actually 19, which is correct. And with that said and done, we now know all the basics to more or less utilize chat GPT. There are a few other things I want to show on the website before we move on to some more advanced examples and how to utilize chat GPT to its full effect. Let's take a look at the rest of the website here. We have this toggle here for dark and light mode. So the default, I'm pretty sure, is the dark mode or whatever your operating system is on. I'm going to swap to dark mode, which, well, doesn't do anything. But then I can also swap to light mode, which is a little bit harsh on the eyes. Personally, I prefer the dark mode, which is what I'll keep it connected to. I also have the option here to connect to the Discord server that we had a look at earlier, as well as the option to have a look at the FAQs and updates, which takes you to a forum-like website here, where you get to learn a little bit more about chat GPT, how it works and its background. It is essentially based on the GPT 3.5 model, which is a newer one that provides human-like feedback. Let's begin with our first advanced prompt. This prompt will be on how to create some food, specifically some seafood. First of all, I don't particularly know yet what type of food I want, so I'm going to ask chat GPT to give me 10 dishes with seafood. And here is an example of a list prompt. This is where you ask chat GPT to give you a specific number of items of a specific number of things. In this case, it's 10 items of seafood. The output here should be based on the context of that query. So in this example, it's a human prompt that is being outputted with a list of 10 items, but it also gives you a little bit more context here, saying that there are different levels of difficulty for each one of these dishes. Now, this is where things can get interesting, because we can now start having a look at this list and selecting options we like, such as the first two for shrimp and salmon burgers, and then following up this conversation with more queries. I'm going to ask in the next part of this conversation if chat GPT can create a shopping list for me based on the very first two items. Those two being shrimp and salmon. I don't have to particularly say shrimp and salmon. It should just know based on the context of the conversation. And the result here is essentially creating another list. But this list does not have a specific length. It's based on the options we added in earlier. So here it provides an output that is indented and bullet pointed for each one of the dishes for the salmon burgers and the shrimps to let me know exactly what I would need to buy from the shops to be able to put these together. That is pretty damn cool. And this is just also still the beginning, because we can continue this conversation based on all this context. Now that chat GPT has provided us a list of 10 dishes, as well as a shopping list for all the ingredients we would need to make one or two of those dishes, the next practical question is quite intuitive, which is, can you tell me exactly how to cook, for example, salmon burgers, which is something of a thing that we would normally have to Google a lot for. But based on this and the ingredients, chat GPT will provide us with one final list of exactly what steps we would need to put them together. This is just a quick shopping list example of OpenAI's chat GPT. Now that you have an idea of how you can continue a conversation, let's have a look at some more examples. This time, I'm going to use chat GPT to do some coding for me. I'm going to use one of the pre-made prompts here on how to do a HTTP request in JavaScript. This is a pretty ordinary prompt, but here we'll get to start to see some of the more coding aspects of chat GPT and what it can perform. Here, it's going to answer this question, but it's going to do this in a number of ways. First, it'll give me some context as to what it's doing. Then it's going to give me the code snippet of exactly how to perform the function based on the prompt I've put in. And then finally, it'll give me a description of what the code is specifically doing. All of this is quite useful, and this is just the beginning. I'm going to request chat GPT to update this request. And instead of it just being a generic API call for data, instead, we're going to grab some weather data. Now, I'm going to ask if it can get me the weather in New York City. And I wanted to output how this query might look like in terms of JavaScript. The response back is that OpenAI will generate the request for you. It will create a generic version of that request using a weather API that doesn't really exist. You'll need to pick your own with an API key that you'll need to enter. But the request should look something like this, where it's getting an example.com, passing in the weather city, being New York City, as well as the units, and passing in the API key. It'll then explain exactly how this GET request would work, as well as how it would come back in the form of a JSON that you could then utilize inside your application. This is all pretty useful, but not quite everything that we're looking for just yet. The next step here is to ask it to update this query in order for it to loop through the data that we would get back from this API, and then place this into a website. Now, something to take note of here with ChatGPT is that you have to be quite accurate in how you're requesting different things. This is why if, for example, you have a background in programming, writing these prompts might be a lot easier, because you know what you're after. Whereas if you're not quite sure, it might be a little bit more difficult. And this is actually kind of true for all types of industries, where ChatGPT might be useful. In this example, I'm getting it to loop through temperature, wind, and precipitation. And this should create a brand new output for me. And this is where the complexity and benefits of what ChatGPT start to shine through. Because what's going to happen here is we're going to get a piece of code that is not just a fetch request, but it's also going to be a loop function, which is going to pull out the values from the data response, and then it's going to put this into a website for us. After it pulls out the values of temperature, wind speed, and precipitation, it's then going to create the JavaScript code to place them into a website using P element tags, and then append them to the website body. Finally, it's going to go through the code here and explain exactly what's happening, like I just have been, where it's using an example API with example code, so you really need to update this to make it work for your own project. There are going to be occasions where the ChatGPT systems are going to be offline, or for example, unable to answer a prompt. In this next example, I wanted to do a story about a rabbit and a mouse. And I put in an option here, and I got a response back that an error has occurred. I've selected try again a couple of times, and that didn't work either. Now, this can sometimes occur when the output of a prompt might, for example, have some incorrect information, maybe something that might be harmful, or maybe even inaccurate. And in those situations, ChatGPT is automatically set in such a way as to not to provide a response. This is a safety mechanism. Sometimes a different way to approach this is to simply write a different prompt so that you might have a different output from ChatGPT. But in some situations, if the servers are entirely down, you might find that you just might have to wait a little bit. Another thing you can do is always refresh the page, or alternatively, just log off completely, restart the browser, and then log back into your session, because that also sometimes fixes the problem. Now, let's do a short story. I'm going to do a story about a rabbit and a turtle that went on an adventure. This is a simple request for a story from OpenAI, and it's actually quite good at providing these. It's going to give us a story about these two that go on an adventure where they swim across a river. There is a small moral to this story, which is useful, but it's really kind of generic here. If I want to have a more specific story based on the prompt I'm putting in, I'm going to have to be more specific with the prompt. And this is where follow-up conversations can be useful. For this example, I want to change the moral of the story to be one more focused about patience, especially when doing a story like this, it makes sense between a rabbit and a turtle. ChatGPT should be able to now rewrite the entire story with more emphasis on this specific moral. If we look closely, it has done just that. It's got a moral here about how it was important to be more patient, and the turtle was able to do this. This time along, I want it to also give me the same story, but throw in a little bit of humor. So I'm going to ask it to keep the moral and add in a joke, maybe somewhere at the end. Let's see what happens this time. I'm going to fast forward a little bit, but we can see that it's more or less the same story. But at the very end of the story, we have a nice little joke here that I guess it goes to show that slow and steady really does win the race. Another type of prompt that you can work with ChatGPT is one of a sort of storytelling video game, which is a really cool one that I've always enjoyed because of the creative answers that ChatGPT provides. In this case, I'm going to give the prompt that ChatGPT is going to pretend to be a text video game with options of A, B, C, and D. And then from that, we're going to pick a scene from the matrix, and we're going to see what we get presented with. The output is based on some storytelling. It begins nice and simple. You find yourself in a dark alley, and we've got the options to select from. Here, I can now continue this conversation by selecting an option, such as the option B here, and the story continues. What's really interesting about this type of prompt is that you can pick any type of story from any type of scenario. You can pick ancient history, you can pick literature, you can pick movies, and then you can enjoy the creativity that ChatGPT provides while also going through and being a little bit part of that experience. Another popular way to use ChatGPT is to generate content for things like websites or even documents that you might need to prepare. While you might not want to use the entire document as is, you'll need to modify it. It can give you a helping hand in creating these. In this example, I'm going to create a copyright and privacy policy for a website. This is something common I have to do as a developer, but here I have an outline of a bit of a template that I can start using, and I can be a bit more specific, saying that it's for a counting website, which allows me to maybe have something more, well, outlined for that type of a website as part of the privacy policy as well as the copyright statement. Of course, if you want to take this a step further, we can do that too. Since I'm located in Australia, I want these templates to be more specific to this country. So I'm going to ask it to update these. In the prompt, I'm just going to simply write that as someone who lives in Australia, please update these to be relevant for the country that I live in. The output from ChatGPT essentially is the same, but with new updated information, such as the Privacy Act, which is specific to Australia, as well as wording, such as including Australia and international copyright law in its template. ChatGPT is also quite creative as well as being able to produce humor. In this example, I'm going to ask it to write a nerdy pun. Why was the math book sad? Because it had too many problems. Yeah, good one. If we do a refresh, we get more or less the same output, but we can continue to ask it for more jokes or more puns as we need, and we'll get quite a few many. Of course, depending on how you write this prompt is how you'll also get the output. So for example, you might want a longer joke. So in this case, I'm going to say, write me a nerdy, maybe joke or a short story, and they would produce a longer content versions of the same output. You can also be deliberate about your jokes and content, such as combining two elements together to see what is produced. In this example, I'm going to combine a hat with a rabbit and see what ChatGPT can come up with. I'm going to ask it to tell me a joke, combining these two items. And here I get, why did the hat keep falling off the rabbit's head? Because it had a hairline. Nice. You can also put ChatGPT into situations to produce humor and jokes too. So for example, I'm going to get it to pretend to be, say, Elon Musk, and I'm going to pretend it had just bought Twitter, which, you know, Elon sort of has done. Then to create a joke, Elon would say, after he's bought Twitter. Now, since the API doesn't know that Elon has bought Twitter, it's going to have to guess here. But its guess is actually kind of accurate, saying that Elon would increase the character limit. There is another type of prompt, which is useful. And this is the one of writing things, such as documents or cover letters, or even job applications altogether. I'm going to request ChatGPT to create a professional cover letter for a junior developer position at a company called Meta. And it's going to be for myself, Adrian Tuarog. And I want to essentially use it as something that I can do to apply for this job, who knows also the basics for programming in general. So with this kind of a prompt, it's quite specific, but it will also give me a specific output. So here I've generated one to a hiring manager. It's giving me a good introduction of who I am and what I've done. Apparently, I've done a bachelor's degree of computer science, which is impressive, as well as working at an internship company called XYZ. So all of these are things that I would maybe need to update a little bit before, say, posting it off. But this also might be inaccurate. I might not even have a bachelor's degree. This is where we can go through and update this prompt of this cover letter, say, removing the fact that I have maybe no prior job experience, but maybe listing out that I worked on quite a few different types of projects when I actually was doing a boot camp in programming rather than going through university. OpenAI will go through this prompt and update it based on the context of what is now required. And this is what makes it pretty cool, because you can essentially version out different types of cover letters here, depending on your needs. The real trick here is to be specific whenever you're prompting ChatGPT to create an output for you. But in terms of that, it usually grabs the context of what you're saying quite effectively, as long as you write it in a clear and understood way. Let me continue this example. This time, based on this cover letter, I want ChatGPT to create a resume for me that I could maybe combine as part of this application. So I can simply write, based on this cover letter, write a resume, which includes things like skills, maybe work experience, maybe even other things such as languages or etc. Based on the context of this, I'm applying for the same position. And let's have a look at what GPT produces. So here it starts off with some essentials, which is like a phone number, as well as a spot there for an email address. It has a small summary. If, for example, the person skipped reading the cover letter, it then goes through and lists out my skills, which is quite cool, as well as the education. It actually has combined both a bachelor's degree and a bootcamp, which are two separate things from that context. But we could update that. It's added in the experience, as well as some example projects. And all of this is content that could now be used in a resume, as long as it was properly updated with real world examples of what I've actually done. ChatGPT is also very effective at doing summaries, as well as pulling out content from documents. Let's have a look at the Privacy Act of 1988 here from Australia. I'm going to scan down to this document and I'm just going to copy paste a random section of it into ChatGPT. Then I'm going to give ChatGPT a query of what to do. In this example, I wanted to maybe summarize or explain to me this Privacy Act in simpler terms. And then I'm going to copy paste the text here. There are a few other variations that we can do here, and I'll get into those shortly. But for the time being, let's just start off with this simple example. The output is just a sentence or a paragraph here explaining what this section of the Privacy Act intends to explain. And while I won't get into the specifics of that, this is just an example of how you can do this. This is basic summarization. Of course, you can summarize it even further, such as requesting ChatGPT to simply pull out the five most important keywords from the above statement. Summarization with ChatGPT can work in a number of ways. You don't always just have to explain a summary. You can give it specifics of how you want to present that information, such as explaining it to me like I'm five years old. This can give quite a few different types of answers. You can do it for someone who's older, younger, or you could even simplify that even further by asking to explain it for a three-year-old. And I like this answer because it gave a prompt here where I could understand that if someone does something bad with your private information, the government will help make it better. That is definitely for a three-year-old. Another way we can request summarization is by being specific with the word count, such as explaining it in 20 words or less. Here, it looks like OpenAI had to think a little bit on how to explain this one, but the resulting answer was within the context of what I requested, which was for it to be in 20 words or less. And I actually really like this one. It said, the Privacy Act is a law that protects individuals' information and provides a mean for privacy concerns. I'm now going to explore some of the strengths of ChatGPT in terms of blogging, writing articles, rephrasing, and much more. We want to create a blog topic, and we're going to use ChatGPT to assist in that process. We're going to request that it provides 10 blog topic title ideas for an article that might be something to do with, say, graphics design or website graphics design. And that is it. From that, it's going to give us a list of 10 different types of options. We've got things like the essential principles for graphics design, role in color, common mistakes, etc. From this list, I'm going to select option three, which I like. Seven common mistakes to avoid in website graphics design. But I don't particularly like that title just yet. So I'm going to ask ChatGPT to give me five variations on that topic title. Then I'm going to paste in the topic title here. It gives me five different types of options here, but all of them aren't that interesting. And I'm actually quite happy with that very first one. Now we can move into writing the outline for this blog post using ChatGPT. I'm going to ask it to provide me the outline for the article in clear writing and the title, which I picked, which is the seven pitfalls to avoid for website graphics design. Now I will get a pretty well written outline. It'll include an introduction with a couple of points, the seven pitfalls themselves. So it'll pick out those pitfalls, give me a short summary and dot points of what they are and how they work. And then it'll also fall in a conclusion. This is pretty much everything that I would need if I was writing a blog article. And this is, again, still just the beginning, because now I'm going to ask it to even write the introduction for this article. And I want it to be at least three sentences long. The output here will give me that introduction, which is graphics design plays a crucial role in the success of a website, etc. Here's where we can start playing around with writing blogs and content in general. We could ask ChatGPT to rephrase the following sentence and then paste in the sentence we want. And here it'll give us an alternative sentence, which is a little bit more nuanced, I suppose. It says the success of a website is heavily influenced by its graphics design. It's essentially saying the same thing. But we can also ask ChatGPT to do a number of other things, such as lengthen the sentence. So here it's saying the same sentence, but it's a little bit longer. The reverse is also true. If we've got sentences that are slightly too large, we can also ask ChatGPT to shorten them. And this provides a shorter version, which is very similar to what we originally had. And we can even ask it to fix the grammar. So say I was just writing my own blog and I wanted to spell check and grammar check, then it can work as a bit of a co-writer there for you, essentially making sure that your sentence, I good graphics design impact good website, comes out essentially saying a good graphics design can impact a good website. Much better written. If you're writing for blogs in different languages, ChatGPT can even translate that to pretty much any language. I'm just going to take a guess here and see if it can translate the above statement or sentence into, say, German in this case. I could probably plug this into Google to make sure it's correct, but honestly, I think I almost trust ChatGPT here more with a translation than even Google. The reverse is also true. If you've got something in a different language, such as German, and you want to translate it back to English, you can do that. You can even add in things like adding it in a friendly, joking way. And we get back this example here. A good graphics design can totally rock a website's socks off, which is pretty funny. Now that we have an outline for a blog, an introduction, and say we've written it as well, we can even prepare a newsletter for it to be sent out by email and even have ChatGPT create the subject line for that email. Here, I'm just going to ask it to create the subject line for an email newsletter based on the above, and it's going to give me an output here of maximize your website's potential with these graphics design tips. Perfect. I hope you guys enjoyed this ChatGPT tutorial. This crash course is just an introduction so that you can understand a better way of how to use it to your best advantage. Of course, if you have questions or if you have suggestions, let me know in the comments below. Don't forget to like the video and subscribe if you want to see content like this. Thank you.