In this video, we’re going to learn how to use NVIDIA’s ChatRTX, and we’ll also look at some of the newest features that just landed today. ChatRTX allows you to run your own GPT Large Language Model, or LLM, on your PC using your own content. That might not mean much, but let’s make this real. I have a folder on my PC with a bunch of different files for my business, the Kevin Cookie Company. I have some images, meeting notes, PDFs, and other documents, and I want to get information out of them. Like, what do our customers think of our cookies, or when is our next brainstorming meeting? In the past, I’d have to click into each one of these documents to try to find this information, but now I can have a conversation or a chat with the AI about my documents. When is our next brainstorming meeting? And look at that, it just spits out the response. Or show me our storefront, and it even understands what’s in an image. Basically, you can now run AI on your PC and it understands all of your content, including documents and images, and you can chat about them, hence the name ChatRTX. That is so powerful. Today, we’ll find out how you can do this too. Let’s check this out. Head to the following website to download and install NVIDIA ChatRTX. You can click on the link in the top right-hand corner, and I’ve also included a link down below in the description. Once you land on this website, click on the button that says, "Download Now." Now, just to warn you, it is a fairly large download, but that’s because you’re also getting an entire model included as part of the download. Before you proceed with the download, one thing you might want to check out is the system requirements. Click on that. To be able to use ChatRTX, you need an NVIDIA RTX 30 series or 40 series GPU. Unfortunately, if you don’t have that, you cannot run this on your computer. So have a look first and make sure you meet those requirements. Once you finish installing, you’ll see a new icon on your desktop for NVIDIA ChatRTX. To launch the AI, simply double click on that. Once you launch ChatRTX, it’ll open up two separate windows. First off, you’ll see a command prompt window, and this will show you all of the different logs, or if any errors come up, you’ll see them here. However, to be able to interact with the AI, you won’t need this window. So, in the top right hand corner, you can minimize this. Underneath that, you’ll see the main interface for ChatRTX, and this is what we’re going to use to configure the AI and then also interact with the AI. Let’s start by looking at the top left-hand corner. Right here, you can choose the model that you would like to use, and ChatRTX supports a number of different models. An AI model is a program that’s been trained on a set of data. So, imagine lots of articles on the web, or maybe lots of different images, and then that model can recognize different patterns and make decisions without needing a human to intervene. Right here, we have many different options, and each model will have its own strengths and also its own weaknesses. By default, when you install ChatRTX, you’ll get the Mistral model. Now, this has been developed by a French company, it’s an open-source model, and it’s actually one of the most highly performing models. I also have another model installed called CLIP, and this was developed by OpenAI. It’s a model that’s been trained on hundreds of millions of images, so this excels at analyzing the contents within an image. Right down below, you can also add additional models if you’d like to test those out, and when I click on this, here I see another option for Chat GLM, and this is optimized for Chinese conversation, although it also works with English as well. If you have a GPU with at least 16 GB of memory, you’ll also see some additional options here. You’ll also see Llama, which is a model developed by Meta, and you’ll also see Gemma, developed by Alphabet, or what you might know as Google, and that’s a new addition to ChatRTX, so you now have additional models that you can choose from, and again, each of the models will excel at different things, so you can install multiple models, test them out, and see which models give you the best results. Once you go through and you install additional models, in the top right-hand corner, you can click on the Settings gear, and this will show you a summary of all the different models that you currently have installed. Here, I have two installed, and you can also see the associated disk space that each one of these models consumes. Now, you’ll notice each of these models takes a fair bit of space. Here, for example, the Mistral model uses up 4 GB. To remove a model, you can simply click on the trash can icon, and that’ll remove it. However, I want to keep these two models for now, so in the top right-hand corner, let’s close out of this view by clicking on the X. Over on the right-hand side, we can now select the dataset that we would like the AI model to leverage. When we click on this drop-down, we have two different options. You can have it look at contents on your computer, so let’s say documents or images, or you could have it rely on the AI model default. And with the AI default mode, this relies on the data that the model was originally trained with. So, with the Mistral model, it’s been trained on millions and millions of different documents, and the responses that you get will be based on all of that training data, but again, it won’t be based on any of your own content that you have on your computer. Let’s start by using the AI model default. If we scroll down this page just a little bit, we’ll see a large field, and here we’ll see responses from the AI, and underneath that, we also see a text field. You probably recognize this. This is very familiar to what you would find on any other chat interface with an AI, and down here, I could type in a question. So, let’s test out the, just the general model. I’ll type in "Write a song about the Kevin Cookie Company." Over on the right-hand side, I can now send my question to the AI, and let’s see what it comes back with. Right up above, we can see the response from the AI, and here it wrote a very nice sounding song about the Kevin Cookie Company. And this is similar to what you’d find with another AI that maybe you use on the internet, but the beautiful thing here is it’s relying on the model that’s on your PC, and you don’t need any internet connectivity, so that’s pretty amazing. Now, one of the great benefits of using ChatRTX is that you can look at content on your computer. So up in the top right-hand corner, let’s switch the dataset now. Here, I’ll click on this drop-down, and instead of using the AI model default, let’s switch to a folder path. With a folder path, you can point it to a folder on your computer that contains content. So here, if you click on the edit icon, here just navigate to a location on your computer, select it, and it’ll look at all of the content within that directory. Right up on top, you’ll see the types of files that this model supports. So, your folder can contain text documents, PDFs, docs, docx files, and in the future, it’ll likely support even more file types. Now that I’ve added my folder path down below, it’s looking at all of the files that I currently have in this folder. However, if I add, let’s say, additional documents or additional images, I need to update the AI dataset. It turns out, it’s easy to do that. In the top right-hand corner, you can click on this button that says refresh dataset. So again, anytime you say add a new document or a new image to this folder, simply click on this button, and this will refresh the dataset that the AI looks at. Let’s jump over to File Explorer and have a quick look at the files that we’re going to interact with with this AI. Here in my File Explorer, you can see all the different files that are contained within this folder. Here, I have a Word document. I have a variety of different images. Here, I have a PDF. I also have a text document. So, I have lots of different types of content. And there’s also a decent amount of content here. Now, one question that I have at the top of my mind is, what are we planning on paying our new safety and compliance manager? And I know I have that information in one of these files. And of course, I could try to search for it up here. Maybe I click into a few of these different files to see if I could find it, but that’ll take time. So, let’s instead see if the AI can help us with that. Let’s go back to ChatRTX. Back within ChatRTX, down at the very bottom, I can now type in my question. But instead of typing, I could actually just speak my question. Over on the right-hand side, this is a brand-new feature. I can click on this microphone icon, and this leverages something called Whisper AI to take my voice or speech, and it converts that into text. And it does an incredibly good job at that. So, let’s test this out. I’ll click on the microphone icon. What is the annual salary of our safety and compliance manager? Once I’m all done, I’ll click on that stop icon. And here you’ll see it’s now taken my voice and it’s converted it into text. And like I said, it does such a good job. Over here, let’s now send this to the AI. And just like that, we can now see the results. It looks like it’s about $350,000. I am in the wrong job. Not only does it pull out the answer from all of the content, but down below, it also shows the reference file. Here, it looks like it found that information in the health and safety compliance manager job posting. And that’s also a link. So, when I click on that, that opens up a text file with all of the details. And here, if I scroll to the very bottom and let’s zoom in a little bit, here we can see that the pay is $350,000 per year. So, it successfully understands the content of all of my different files, and then it pulls out the appropriate response. That is so impressive. Let’s now try a few more. Down below, I’ll click on the microphone icon. When is the next brainstorming session? I’ll click on stop. Here it’s processing it, got the text perfectly, and let’s send that through. And here we can see that it’s May 14th, 2024. Now it’s referencing the monthly sales review meeting text, and I know that date’s not included in that file. So, it doesn’t always get the source file 100% accurate. However, in general, it’s pulling out the correct information. So at least that’s reassuring that I’m getting the correct answer, just not the correct reference file. Let’s try another one now. Over here, let’s click into the text field and let’s see what customers think of the Kevin Cookie Company. We’ve gotten lots of different customer feedback, both positive and unfortunately also some negative feedback. And let’s see if the AI can summarize all of that feedback. Over on the right-hand side, let’s send that through. And it looks like customers are generally positive. Here it says they appreciate the eco-friendly packaging. Well, I’m glad we invested in that. But it looks like there is some negative feedback points such as concerns about the texture and taste of the vegan cookie line. That’s been a pain point for some time. So, there it looks across all the different content and pulls together a summary of what all that customer feedback is. Let’s now try one more. We recently had one of our employees recommend that we start a subscription box, but I can’t remember which employee that was and I had a few follow up questions. So down below, I’ll type in who proposed offering a subscription box to customers. I remember we had it in some of our meeting notes. So, let’s try sending that through to the AI. Of course it was Tom. He always has good ideas. He’s the one who proposed offering a subscription box to customers. And here again, I can click on the reference file. This opens up the Word document. And here I see the section on the subscription box and it says Tom suggested a monthly subscription box. So there again, it’s pulling out accurate information from my files. And again, the thing that’s so impressive is it retrieves information across a wide variety of documents and then it presents it to you in a conversational form. This makes getting back to information so easy. I also had some images in my folder that I would like to find, but the Mistral model excels at text-based documents. So over here in the dropdown, let’s try one of the different models. Let’s try the CLIP model, and this one was also recently added into ChatRTX. With the CLIP model, it understands the content within photos. This one is so impressive. Your photos don’t need any metadata and you can still find different content in your photos. Now that I have the model selected, here I’ve indicated my file path down below with the dataset and down below, just like we did with the other model, here we could type in our prompt. Now I want to see all photos with chocolate chip cookies. Let’s try sending that through and look at that. I get all the different photos back. Here I get two with chocolate chip cookies, and then I also get our factory, which is making chocolate chip cookies. So, there it’s successfully identified all of the correct images. Now I know you’ve all seen chocolate chip cookies before, but I bet you’ve never seen a blimp for a cookie company. So down below, let’s pull up the blimp. Here, I’ll type in, show me the blimp and let’s send that through. And look at that. That is beautiful. There’s the Kevin Cookie Company blimp flying over New York City. This thing is going to supercharge our sales. Down below in the text field, let’s try one more. I’ll type in, show me the storefront and then send that through. And there we could see one of our retail locations for the Kevin Cookie Company. It looks so inviting. These photos have no metadata and yet the AI model is able to understand and interpret what’s in the photo. So here, when I asked for a storefront, it knows which photo contains a storefront. Again, that is so impressive. We’ve now looked at all of the core functionality of ChatRTX. And once you’re all done and you want to close it down, in the top right-hand corner, there’s a shutdown icon. And when you click on this, this will release all of the different video memory. Once you shut it down, you could close out of this window and you can also close your command prompt. And then to launch again in the future, simply click on the desktop icon. And look at that, you can now run AI directly on your PC. Once installed, you don’t even need an internet connection. You don’t have to worry about privacy. Your data stays on your machine. I’m sure coming soon, you’ll be able to run ChatRTX on broader sets of data. Imagine being able to chat about your email. Now that will be a game changer. To watch more videos like this one, please consider subscribing because the next AI update is always just around the corner. I’ll see you in the next one.