Name: ChatGPT 101: A Guide to Your AI Super Assistant
Uploaded: 2025-04-15T01:04:08.875Z
Duration: 48 min 4 s
Description: ChatGPT 101: A Guide to Your AI Super Assistant

Transcript for "ChatGPT 101: A Guide to Your AI Super Assistant": Hi, everyone. Welcome to today's webinar. I'm going to be getting started shortly. I'm just going to give everyone another ten seconds to join. Okay. I'm going to kick off. Hi, everyone. Welcome to today's ChatGPT for Work one zero one webinar, a guide to your AI super assistant. My name is Lois, and I am your host today. I'm a customer education specialist here at OpenAI, and so no surprises here. I use ChatGPT every single day. I've really designed this session as a foundation to get you started. And my hope is that you leave today's session with an understanding of multimodality, what it is, and how it can be applied to an infinite number of work use cases, the difference between GPT four o and OpenAI's o one, how to navigate the ChatGPT interface and customize your settings, and some tips for prompting so that you get the most valuable output from ChatGPT. Today, I'm joined by several of my OpenAI colleagues who will be answering and moderating your questions from the q and a feature, which is to the right of the screen. Also, make sure to check out the docs section, which is next to the q and a, which will give you downloadable assets and helpful links to explore. As with all of our webinars, this session is being recorded, and we will share a recording with you tomorrow. Before I start sharing my screen and jumping into ChatTBT for today's series of demos, I I want to talk a little bit more about OpenAI. OpenAI's mission is to bring safe artificial general intelligence to all of humanity and benefit all of humanity. We've created advanced AI models like GPT four o, which is our flagship multimodal model, and the model I'm going to be showing you all first. We've also created more recently reasoning models like OpenAI's o one, which has been designed for deeper, more deliberate reasoning. GPT four o can process and understand a really large amount of input. And by input, I mean things like text, audio, images, data, and that makes it incredibly versatile. So ChatGPT and GPT four o is going to be very good at generating text, analyzing images, processing files, browsing the web, and you can now even talk to ChatGPT with your voice. I'm going to jump into ChatGPT now, and I'm going to do a series of five or six demos to showcase what I mean by multimodal capabilities. Let's start with my first input, text. I'm going to ask ChatGPT a simple question, and it will generate a text based response. To do that, I simply navigate to the chat window, and I enter in my first prompt. In this first prompt, I am saying, I work at OpenAI. I am familiar with ChatGPT, but I want to understand how I can use it more effectively as a customer enablement manager. Can you provide the top five use cases for my role in a table and include the relevant prompts for each use case? If you are new to ChatGPT and you are trying to understand how to leverage AI for your role, I would strongly encourage you all to copy and paste this prompt from your prompt sheet, and then I would recommend that you edit the section and identify your own organization, and you also switch out my role for your role. I'm going to send this across now, but let's take a look at the output because I think you'll all find this really valuable as a first starting point. What we can see is very quickly, g p t four o has processed that request and said, below is a table outlining the top five use cases and how you can use it more effectively for your specific role. What I love about this output is I have been very specific that I wanted a clear table format, And ChatGVT has done a really great job of identifying the use case, how it can help me, and then an example prompt. This is the best way for you all to get started. And if we take a look at the use cases for me personally, ChatGPT is going to be great for drafting training materials. It's going to be good for designing enablement sessions. I can upload feedback from these webinars and get feedback turned into an analysis. I can personalize AI use cases, and I can also create and automate FAQs. What I find really interesting about this view is this is exactly how I use ChatGPT in my day to day to support me in my world. Grab that prompt today at some point, upload it, and see what you get back for you. This is a great way to start personalizing your experience and making AI more applicable to you. This is a really great starting point. We've uploaded text. We've asked a request of ChatGPT, and we've got a text based output. Chat g b t is actually capable of a lot more. So let's move on to my next example, and let's move from text to search. What I love about chat g b t is you can now access the web and information from the web in real time through the chat interface. So let me now show you how to do that. We simply navigate to the new chat icon, and that is going to open a new chat. I would strongly recommend that you do this every time you make a new request of chat g p t or you're switching and changing topics. Okay. I'm going to add in my second prompt of the day. And here, what I'm saying is please find suitable venues in the city of San Francisco to host a 100 person partnership event. I'm get then giving the specific requirements of the venue. I would like it to be good for presentations. It should have strong AV capabilities and also some on-site catering as well. What you would have noticed is before I ended in the prompt, I toggled on search the web, and that icon turned blue. Nine times out of 10, ChatGPT is going to be smart enough to know what capability you're looking for just based on the words you've used in the prompt. But, again, best practice, if you absolutely know that you want ChatGBT to search the web, you should also toggle on the capability at the bottom there. As with the last prompt, we send that across. And notice that straightaway, ChatGBT will tell you that it is searching the web. We can see at the top, we get a summary, and it's identified all of the relevant locations in San Francisco that meets my specific requirement. What's great is is we have an overview of capacity, features, and accessibility. And we've also got some hyperlinks included in the overview, which will allow us to go and explore that website. We've got a location here at the Ferry Building. We have one in SoMa, 1 in the Financial District. And, again, very easy for me to go and explore the sources where this information was taken from. What I love about search within ChatGBT is I no longer have to go to Google search and manually review web pages myself. Chat g b t is doing all of the hard work for you, and you would have noticed that that search took less than five seconds. So previously, what would have taken me an hour or two to try and get that information from the web, has been reduced down to a task that takes a matter of seconds. If we scroll to the bottom, what I also love about this view is we have a sources section. When I click on that, an interactive view will appear. And, again, every source that was searched to generate this answer is going to be aggregated in the right hand panel, and you can go and explore this in more detail. I want to talk a little bit about something called human in the loop. Human in the loop is where a human like myself is driving AI to get an output, and then I'm also verifying the output. You can see at the bottom of the screen, there is a disclaimer that says ChatGPT can make mistakes. And I will strongly encourage you all to adopt a mindset where you take the response from chat g b t, but then you also fact check it as part of the task. For example, this is really, really helpful. But if I was doing a more in-depth or rigorous piece of research, then I absolutely should be exploring and verifying the sources before I take the information out of ChatGPT and pass it on. And this really applies to any work that you're doing with ChatGPT, whether it's writing, content generation. Really important to make sure that you're happy with the output and you fact check it before you do anything with it. Great. So moving on from searching the web and staying within ChatGPT to do a piece of quick research and talk a little bit more about vision. ChatGPT excels at analyzing images and providing clear text based summaries of its analysis. In this example, what I'm going to do is I'm going to upload a hand drawn proof of concept outlining how I want a new webinar landing page to look. To do this, I use the plus icon. I upload from my computer. I add in the image, and I'm going to just quickly show you what the image is. So this is a hand drawn mock of a webinar registration landing page. This was just in my brain. I was just doing a very quick draft. But what I love about ChatGPT is I can now ask ChatGPT to analyze this image and explain what is in this image. Let's add in the prompt. And let's see how good ChatGPT is with its vision capabilities. Straight away, ChatGPT has identified that the image is a hand drawn flyer for a webinar event. It's got a breakdown of the content and structure, the event title, the description, and the agenda. So what you're seeing here is that ChatGPT is incredible at analyzing images and then summarizing the analysis with text. You can, in your world, use this capability to do things like scan diagrams, detect visual patterns like defects, or interpret images like whiteboards or product samples all in seconds. Vision also pairs really well with GPT-4 o's upgraded image generation capabilities. I'm going to now show you how to do that. So in this example, we uploaded an image, and we leveraged vision, and we got a text based analysis. Now what I'm going to do is I'm going to ask ChatGPT to take what it has learned from the image we uploaded and now create something entirely new. So what we're going to do is we are going to add in our prompt, and let me break down the prompt for you. Here, I'm saying, please generate a proof of concept image for a registration landing page. Maintain the same layout as the original reference, which is the file we've just uploaded, but incorporate the provided images to enhance the overall look and feel. Keep the core structure and placeholders, and the goal is to present this design concept to our team so that they can iterate and build a final version. What we're going to do is we are going to add in two images that we want ChatGPT to incorporate into this design. I really love that the updated image capabilities within ChatGPT can do this. It can take the original diagram that I've uploaded. It can reference more images and a text based prompt in order to provide me with a really nice design. So let's show how this works. We are going to add in my headshot here. We want this image to appear on the speaker section of the design, and we're also going to add in some OpenAI branding. This is an image that we use, in some of our designs, and so I'm really asking ChatGPT to incorporate these images and create this new design mark. Let's send this across. Now please note that we are asking ChatGPT to create a design, and we're really leaning on four o's image capabilities. This can take some time. So you will see here that it will be blurred out for now, and then it will start to render. This can take anywhere from ten, twenty seconds to even a couple of minutes. So while that is loading, let's come back to ImageGen, and let me show you a project that I have created which has some really awesome images. Let's come in here and you will see that any chat recently that I've had where I've created an image, I've stored it in this project. We're going to talk a little bit more about projects, what they are, and how to build one in next week's one zero two. But for now, let's take a look at this chat here, and let's take a look at the original prompt. So in the prompt, we're saying create, an image of photorealistic astronauts, and they're in this galaxy setting. We want shimmering trails of cosmic dust swirling around, and, we want it to be some kind of ballet scene, humankind versus the cosmos. And then we get an image like this. I personally think the images that ImageGen creates are really quite breathtaking. I think what this has highlighted to me is that OpenAI can be used to inspire a range of creative activities, not just day to day research, writing, academic type of work. If we come back into the chat, you will notice that I leveraged the edit feature, and I did this by clicking into the image, navigating to the lasso, and then here at the bottom, I entered in the prompt explaining what I wanted changed in the image. Coming back out, you can see that I added the prompt add faces. So really what I was asking OpenAI to OpenAI to do, sorry, is to enter faces into the astronaut helmets. And if we take a look, this is the result that we got. So you can see there that the the background image is actually very similar. There are some slight differences, but we have been able to add in those faces. What I'm highlighting here is that you can not only generate very impactful images, but you can also edit them as well. Impactful images, but you can also edit them as well. This is not to replicate or even be on the same level as something like Photoshop, but what it does do is it helps you with, creating design material as a mock, as a first pass. And that's really how we're thinking about ImageGen. Just to show you, another one as well, if we jump back into this project, we also have an eye reflection concept here. So similar, I was trying to get this universe to render within the iris of an eye. Again, really beautiful, almost photorealistic type image. But what was great is I now then went back to chat g b t and said, now change this to be in the style of a specific studio. And if we scroll down, you can see that chat g b t has been taken that image and changed it. So these are just really for inspiration. Again, I would encourage you all to test and play around with this. Think of a prompt or think of a scene, think of an image that you want to create, and then ask ChatGPT to do that. I noticed that the chat that we were waiting on is now complete. And if we come back down, Let me refresh. Ah, here it is. Okay. So let's take a look at what we got back. So remember, I was asking for this proof of concept. I've included an image of my face and then some OpenAI branding, and let's take a look at this webinar page. Okay. So we have the exact same structure and text as we did from the original image up here. So if we take a look at this, remember I wanted join our webinar, a registration link, and then some text to describe what was in the webinar. And if we come back down to this image, we have exactly that. Now it hasn't quite got, the speaker section right, and I do enjoy that this is a, an AI generated image of my face. So it's never going to be able to take that photo of my face and then paste it into the image. But, again, this is really just a proof of concept. What I'm trying to do here is have a design, create a design, and then send it off to my team. What I really love about the updated image gen capabilities is the text rendering is incredible. Previously, the old model struggled to do this, but you'll notice that almost nine times out of 10, the words are spelled correctly. They looked correct, and it followed that original image. Okay. So we're moving away from vision and image generation. And now let's move away from multimodality and move on to talk about a reasoning model. Coming back into ChatTBT, what we've seen is that four o is incredibly flexible. It can take a lot of input. It can generate a lot of output. But our o one series, OpenAI's o one, has been built for advanced reasoning. Now what I mean by that is this specific model that I'm about to show you has been designed for solving complex problems with clarity and precision. So unlike g p t four o, o one takes more time to deliver thoughtful, detailed responses. This makes it really great for math, strategy, logistics, coding, and science. And for quicker tasks, o three minutei is the lighter weight version. So o three mini has actually been optimized for coding and math. So if you come to the model picker and you're looking at the different types of models, really, that's what you're going to notice. I can't go through all of the models today, but I will do a piece of deep research next week and show you how you can actually leverage our tools to understand more about them. What I'm going to do now is I'm going to walk you through a demo of o one. Let's imagine you're in finance at a pet product company. You have 12,000,000 in revenue and 500,000 to invest. The goal is a 15% ROI in twelve months, and you're choosing between either, a, expanding into Canada or another market, or, b, investing in a US specific marketer or influencer type marketer with the possible of subscription add on. O one is actually going to weigh the options. It's going to calculate the ROI, identify risks, and also offer a fallback plan. You'll notice when I load the prompt, it's even going to consider things like shipping costs, exchange rates, and market trends to help us make the best decision. So let's see it in action. I'm going to grab the prompt, and I'm not going to read the prompt for you word for word. But, again, it is available in your prompt pack. And what I'm saying here is pretty much what I've just explained to you. I'm really outlining the scenario, what I'm working on, and giving a detailed overview of the two options. I'm then being very specific about what I want. I want timeline and key assumptions, resource allocation, major risks, and a fallback plan if revenue or ROI targets fall short. Okay. So a couple of things about working with o one, and this is kind of the same when we're thinking about prompting. You really need to outline the scenario. You need to be very clear about what you want o one to work on, what's the problem at hand. You need to provide the figures and the numbers, especially if it's scientific or math related. And you also need to be explicit about what you want o one to do. This is going to guide the model to provide you with the most valuable output. Now the most important thing to remember is if you want to engage with a new model, you need to come over to the model picker and you need to select a one. You can see here that that has been selected. And now when we send that message across, we will be engaging with a reasoning model. You will then notice that ChatGPT will tell you that it is reasoning. It is working through the problem you provided, and then it will start to flash up and tell you what it is working on. I find this fascinating because you can actually see how the model is crafting and designing the answer that you're looking for. You can follow along, but what you'll notice is unlike four o, o one is not rushing to a conclusion right now. It's weighing the potentials. It's modeling the scenarios. It's doing deliberate reasoning before providing you with a response. And, again, sometimes we have to wait a little bit of time to get this output. Great. And what we have is the overview or the answer that we were looking for. You can see that this example took one minute four zero one to reason. Below is a structured comparison of the two major investment options. We've got the option of expanding into Canada, and we also have the focus of staying in The US market and looking at the influencer and social media push. We have projected ROI for each option. We've got the investment, projected revenue increase, assumed net margin. We've got option b in the same format as well. You can see there that there are some detailed calculations, but also assumptions, so we should fact check this. We have a breakeven timeline and key assumptions. So Canada expansion again and remaining in The US market, but looking at influences and subscriptions. We have resource allocation for each option. We have major risks and mitigation strategies. The full black plan that we requested, and a recommendation summary. You can see there, overall, if the primary goal is maximizing ROI and you can handle the operational challenges, the Canadian expansion is the stronger single option. If brand building and lower operational risk are the priority, the influencer and subscription box plan within The US market is more straightforward, though it yields a slightly lower return. I love this. This is incredible. What took one minute for o one to do and generate could take a single individual or multiple individuals, series of days to be able to come up with a strategy like that and do do the financial modeling. O one is incredible for things like strategy and logistics and planning. If you are in a role right now where you have to really think about scenarios, model scenarios, and you have to strategically create plans. O one is going to be amazing for this. Again, I would really encourage you to think about these scenarios, the types of questions that you could ask o one, and please just test it out. Grab a scientific problem or a math problem from online and paste it into o one and see how well it does. Moving away from o one, I want to help you navigate around the interface. This is important. So so far, we've remained focused on the chat window and also the new chat icon. You would also notice that I navigate to projects, and I also have my chat history over here on the left hand side. The interface is relatively flat, but there are some key things that you will want to know. Let's come up to the search chats feature. That allows me to do a keyword search. For example, if I type in venue, you will see that anytime I search for a venue or the word venue was used, I can find that chat and navigate back to that chat. Search is going to be really helpful if you can remember something within a chat but can't remember what the chat was called. You can see that all of my chats for the previous seven days have been captured. And at any point I want to go back in and review a chat, I simply click on it and return to that chat. Anywhere you see three dots in chat g b t indicates that you can do something additional. Here, we have the option to add the chat to a project, to rename it, to archive, delete, or share. You also have the ability to close this panel and continue to engage with chat GPT in the chat. Sometimes the left hand side can be distracting, so feel free to close that off. That's simply a user preference. We have the model picker in the top left, and you always need to make sure that you're leveraging the right model and sending the prompt to the right model to get the output that you're looking for. Coming back over into the center of the screen and navigating to the right, you will notice a circular avatar. This avatar identifies the account that you're working in and is how you navigate to your settings. When I click on this, you can see that I am in my demo account. And if I scroll down here, you'll notice I'll be able to customize my chat GPT. Top tip from me today, if you haven't done this already, please jump in and customize your chat GPT. Customizing chat GPT is going to ensure that you get better personalized responses. So same as what I showed you earlier with that prompt and making sure that you understand the use cases that are relevant to you, please also personalize your ChatGBT account. You can see that I've told ChatGBT I would like it to call me by my name. I've identified the role that I do at OpenAI. I've been specific about the traits that I would like ChatGBT to have. So for me, I'm doing a lot of, demo creation, script writing, content strategy type work, but I need ChatGVT to be factual, friendly, and professional. Scrolling down, I've given ChatGVT a little bit more about me, so feel free to tell ChatGVT as much about you and the work that you do, the organization that you work for, and the industry that you work for. Again, these small tweaks in chat GBT are going to ensure that your experience is more personal to you and your role. You're going to want to toggle that on for all chats and then click save. And I don't know if you've noticed, but previously, the tagline there said, hi, Lois. What do you want to work on today? Coming back over into settings, you also have some other options as well. You're going to be able to change your theme from light to dark. You can change, your languages or your language preferences. You have the ability to archive all of your chats if you so wish. And, again, you also have the option to add in additional personalization features. Memory, for example, is the ability for chat gbt to capture and understand specific things about you, and it will reference that information in future chats again to offer a more personalized experience. Please note, and this is really important, not all of you are going to see exactly the same features and capabilities like I have in my screen. If you can't see all of these options, there is a high chance that your chat gbt admin or your IT teams have turned off certain things. So don't panic if you can't see exactly the same options. It's a high chance that your organization doesn't allow for all of these features. Just quickly, and unfortunately, I wasn't able to engage with my voice earlier, but I have selected a specific accent that I like. I've chosen veil. Hello. It's lovely to meet you. If there's anything you need, please don't hesitate to let me So, again, you can really personalize that experience. Great. Okay. So I'm going to move away from, customizing your ChatGPT and personalizing it, and I now want to talk about Canvas. Canvas is the ability to partner with chat g b t on writing and coding tasks. It's a new interface for real time collaboration. What it is going to allow you to do is reduce the back and and forth by offering inline feedback and proactive suggestions. You can directly edit, restore, or refine outputs, giving you a lot more control. I think of Canvas as the new AI partner that gives me feedback on my writing similar to how I lean on my colleagues and I share word docs and I ask for their feedback and their comments, you can now use OpenAI's ChatGPT in this way. Again, this is gonna save everyone a lot more time. And once you're really happy with the draft that you have or even a final version, that's when you can ask your colleagues for their time and to give you feedback. In this next example, I'm going to upload a blog post, and I'm going to ask Canvas to give me feedback. I'm also going to show you how you can adjust the settings to make reviews and edit suggestions as well. What I'm going to do is I am going to paste in the first part of my prompt, and I'm going to ask OpenAI's ChatGPT, can you provide feedback on the following blog post I have created for the launch of our new travel app? The blog post should be clear, concise, and in plain English. The tone should be friendly but professional. Then what I'm going to do is I'm going to add in said blog post. Just going to add this below the prompt. And really important, if we want support with writing and we want to turn on Canvas, we need to navigate to the three dots and we need to click on Canvas. When I do that, you'll notice that canvas gets inserted into the beginning of the prompt. We are going to send this across, and what you would have noticed there is instantly this new interactive left hand panel. So you can see that ChatGPT has said, here's some feedback to help polish your blog for clarity and engagement. We've got some strengths, what ChatGPT likes about the blog post, and then we have some suggestions for improvement. What's really great is we can read through that, and potentially, we can ask ChatGPT to to do something. So we can engage in a conversation back and forth, or we can navigate to the suggest edit section. And when I do this, watch what happens. You'll notice there that it goes through the blog post line by line, and it starts making suggestions. Very similar to an assistant or a colleague who is going to give you feedback on individual line items. When we click on the comment section, you'll see that chat g b t is telling us we can consider breaking this into, a sentence or two for better readability. Here, it's saying this sentence is engaging, but maybe move it earlier. And we can go through these line by line. And if we like ChatGPT suggestion, we can simply click apply, and ChatGPT is going to make that change directly. Again, we can engage in the left hand side and go back and forth and ask for changes, or we can do the suggest edit section. So let's go ahead and apply some of these changes. Perfect. And I want to introduce you to a couple more edit features. You have the ability to just adjust the length. Let's say I want to make this a shorter blog post. Let's see what Chatubit can do. It will go through and update this, again, making it much shorter. We have the ability to adjust the reading level. So depending on your audience and the type of writing task you are doing, you might want to adjust this. You can see you can go down to kindergarten or all the way up to graduate school. Let's keep this at high school. And at any time, you can cancel those changes if you've made a mistake. Coming back down, you can add some final polish, and you can also add in emojis as well. Let's wrap this up by adding in some final polish, And you can see that we now have, a really great blog post. Again, would strongly encourage you always to check this. And with any writing tasks, make sure that it sounds like you and your brand tone of voice. A couple of options, you have the ability to show and track changes and go back to previous versions. You can undo something and return to the previous version. You could copy the text directly from ChatGPT and then paste it into your preferred, document software, and you can also share the chat as well and engage with colleagues and partner with colleagues. You can also use Canvas for code. You can insert your code directly into chat gbt, toggle on Canvas, and ask Canvas to help you debug your code, and you'll be able to work on your code in this interactive view. Perfect. I'm moving on into the last two to three minutes of the webinar today. What I want to talk about is prompting and how we refine and design our language to get the best output from chat g p t. To clarify, a prompt is simply the question or instruction that you give chat g p t. You will have noticed throughout this session today, I've been adding in prompts. Some have been short. Some have been long. But, ultimately, prompting is really about making sure that you're giving ChatGPT the most refined, the best communication you can to direct ChatGPT to giving you a valuable output. Prompting, also known as prompt engineering, but you do not need a technical background to be good at communicating with ChatGPT. Quite the opposite. Really, this is all about good communication, and my advice is that you think of ChatGPT as your assistant, as your colleague, and you start to communicate it in the same way you would a human. Humans respond best to clear, concise, specific instructions if you want them to perform a task, and that is exactly the same for chat GPT. In short, and I want to make this really simple, a good prompt requires two things. It requires context and expectation. So what I want to show you now is a basic example of a prompt. We're going to look at the output, and then I'm going to give you the best version of the prompt and we're going to examine the output so that you compare the two. A very basic prompt would look something like this. Summarise this research paper. I'm then going to upload the research paper. This is a paper that was created by OpenAI outlining how LLMs can reason. And, honestly, sometimes a prompt like this is okay. You might not know exactly what you want or what you're looking for, and you can be vague. But when we use a vague basic prompt like this, we need to expect that the output is likely to be generic. When we send this across, let's see what we get back from ChatGPT. Here, we get an overview outlining what is in the research paper. We get some key advances, performance highlights, human evaluation, safety, and alignment. We also get a conclusion. So listen. This response is helpful. If we really did just want a high level overview of what's in the paper, then this can be really valuable. But what I want to show you now is when we spend the time thinking about our request, thinking about what we're working on and the task in hand, and we are very clear and concise with ChatGPT, we get a very different output. To do this, I am going to navigate to a new chat, and I'm going to give you, an updated prompt. So this prompt is really the best version of the prompt. So here I am saying, I am an education specialist at OpenAI responsible for training users on how to use ChatGPT. I need to understand the latest research paper in detail to enhance my training content. Can you summarize the paper with a focus on fight with a focus on how the findings impact practical applications of chat GPT, highlighting key concepts, and clarifying technical terms relevant to training users effectively. Can you see there that I have explained who I am, what I'm doing, so really detailed context, which is the first part of the prompt, And then I'm being very clear about my expectation. I really want to understand the impact of reasoning models and how that can be woven into my training, and I want ChatGPT to really pull out key concepts and clarify technical terms. Again, I'm directing ChatGPT to the output that I know I'm looking for. Again, we upload that research paper, and we send that across. So straight away, we have a similar overview. Then what we have, which is really helpful, is the key concepts. So the concept is on the left. The explanation is in the middle, and then the practical application for how I start to teach this in webinars is then on the right hand side. We've got performance highlights, human preference data, safety and alignment, so very similar to the response we got before. But here is the gold. We've got some example use cases for training. You can see there that it's identifying the use case, and then it's even giving me a prompt to help me in the future actually train users on how to use something like o one. We then have a really great summary. So what I'm trying to highlight here is that just by spending time thinking about the request and designing the language, I have got a really valuable output here. This is what is relevant to me and my role. So in short, when we think about how we communicate with ChatGPT and we communicate with ChatGPT more like it is a human completing a task, the higher the chance that we find the output more valuable. That brings us to the end of today's session. I really hope that that was helpful and you gained some tips and tricks. We've covered a lot today from multimodality, reasoning, uploading files, generating images, navigating the interface, and also prompting. It would be great if you could take a second to answer a quick survey, but thanks again for spending the time with us. And, of course, this week, I really want you to take some of these tips and tricks, jump into ChatTBT, and start testing it. Thank you so much. I hope you have a great day.