ChatGPT Memory, Nvidia's New Chatbot and More!

In this week's AI use cases, we had so many releases and they're very diverse, ranging from a new ChatGPT feature over a new way to use your voice to monetize it with AI, all the way to Nvidia introducing a personalized AI chatbot. And before we get started, let me just address the elephant in the room, and that is that this was the biggest week for AI announcements in a long time. We got Sora from OpenAI, the AI video model, and we got the new Bard 1.5 model with a ridiculous amount of context. Both of those are extremely interesting. I created a separate video on Sora, but in these videos I only try to cover things that you can use today. And that's why we won't be going more in depth on that. We'll do that in separate content. Okay, so with that being said, let's dive into this week's use cases. So first things first, let's talk about this new ChatGPT feature. Everybody has been tweeting about this, creating YouTube videos about this, and I actually have a take that is very different from everything else that I've seen. So is this a big deal? Yes and no. Let me explain. So, brief summary of what this does is it looks at some of your previous conversations and saves the info in there to create a memory bank that, that it's now going to use as context moving forward, yielding more personalized results. If you don't do any prompt engineering at all, this is fantastic. It just automatically gathers the context for you and infuses it into the future prompts. It can be enabled under personalization and at the time of this recording, I don't have it yet, but it's rolling out to everybody. But why did I say yes and no? Because that would be a clear yes, right? Well, no, because this is not really revolutionary. It's again, just like I said about GPTs, just a more user friendly way to implement this. Because up until now we had this feature, it was simply called custom instructions and you could specify your personalized context, but you had to craft them yourself. You had to write something up, or you had to use ChatGPT to generate them, or you had to get one of my presets. Either way, you could do this. You could provide it with more personalized context, but now it's gonna gather it automatically, which is a huge unlock for most users because most people don't bother with customizing the custom instructions. There's an interesting post on Twitter here by Simon Willison, essentially pointing out that all ChatGPT is doing is saving kind of a bio in the background that it now uses as context. And then Ethan Mollick nicely retweeted here that since GPT4 is a great cold reader, it can likely feel very personalized with just a few notes as hints. But that was also the case with custom instructions. So great feature for most people. But if you're more advanced and if you use features like this and GPTs not the complete game changer that some people make it out to be. Okay, so let's move on to an interesting GPT that was found by our community and this is the book summarizer and I thought this one was particularly interesting because it summarizes books really well in a way that you probably wouldn't prompt it. So let me just show you. I'll just say Summarize Sapiens. Yep, that's the book by Yuval Noah Harari. It just asks me one question. Question. I'm just going to say English and look at this. It has these wonderful instructions in the background that actually give you a 8020 analysis. This is my favorite part of the book, essentially giving you the essence of the entire content and I tried this on several books that I read and it works exceptionally well. This is my new favorite way to look at book summaries. It gives you several other points like an action plan and blind spots of the book and then other book recommendations similar to this is if you enjoyed them. Impressive. I think this is fantastic. If you have a ChatGPT plus plan, this is free. Just wanted to point it out. Now let's move on to the next one. And this one is the definition of a use case because we've talked about Suno AI before, right? It's the best music generator when it comes to including lyrics. So you just give it the type of song that you want and it's going to generate the lyrics, turn them into audio, create the music, match all that and it sounds really good. So now by the time you're watching this video, Valentine's Day is over already. But they have this use case where you could create a customized Valentine's Day song for somebody on any other occasion. I think this is one of the most interesting, unique and also beautiful presents that you can create with AI for somebody. And look, you could just package it with some other present that you plan and make it so much more personalized cuz you're adding a personalized song. So yeah, I guess gifting custom songs is a use case that I wasn't aware of before this. Give it a shot. Now on to the next one. And this is just a new feature from eleven Labs, but essentially they're allowing creator monetization, which I dare to say would be a new concept in AI. Right. GPTs wanted to do it, but haven't done it yet. With the new feature, you can sell your own voice to other people, introducing voice actor payouts. Now you can create an AI voice and earn while you sleep. I can do what? So is this a great idea? Should you be selling your voice? I don't know. My intuition is telling me that, who knows, we'll see how this plays out. This is a brand new thing. But like, you do have to consider that if you're selling the voice, they claim to have a really good moderation system that doesn't let everything through, but nevertheless, you're giving away your voice. And any words could be spoken with that as long as it gets through their moderation. But still, the reason I wanted to really feature this is because I feel like this is a new direction that we haven't really seen yet. Right. Generative AI tools that, that are monetized by creators, by the people providing their voice. In the case of GPTs, once we get monetization in the store, it's people creating custom instructions and linking actions to them and uploading knowledge bases and then getting a cut of the value that it creates. Same with the voice here. I reckon this will be a thing across the ecosystem. Just thinking about it, it does make sense, right? If people are able to create something unique and something valuable, yeah, they should totally get a part of the upside. Just like YouTube does it. Now this is very different because it is generative AI, but I'll be watching this closely. I'm very interested in how this plays out. And yeah, 11 Labs is one of the first players to enter this. If you know of any other AI tools that do revenue splits, I'd be very interested in seeing those. Just leave a comment below. So all of the AI tools we're looking at here every week are great at enhancing your skills, but you do need a strong foundation, a so called core skill to build upon to use these technologies to enhance that. And one of the best ways I found to build my own skills is skillshare, today's sponsor. So, skillshare is a massive learning community with classes taught by industry pros across tech, entrepreneurship and content creation. One of my favorite classes on there is called notion for YouTube creators. I personally use Notion to manage everything in my business from the video production to the community content and everything around it. And some of the tips and tricks in that course really helped me to improve all that. And the course notion for YouTube creators is actually just one step in the learning path that teaches you everything you need to know to use notion. Well, actually, you can get started super fast, but it's very rich in terms of its capabilities. And what I also like is that the classes in there are taught by different creators, so you're constantly exploring various perspectives. And of course, that's just one of the learning paths available to you. Whether you're creating content, building a business, or you just want to expand your skills, skillshare has the right path and courses for you. So invest in yourself by signing up for Skillshare today. And the first 500 people to click the link in the description will receive their first month completely for free. Thanks again to skillshare for sponsoring this video. And now let's get back to this week's AI use cases. Okay, so for the next use case, it's just a tiny little feature inside of Microsoft Copilot that we already talked about. It's Microsoft's personal assistant, where you can get access to GPT4 for free. Still, all you need to do is log in with Microsoft Account, but they added a new image editing feature. And I think this one is important to point out because it really shows the direction that this space is evolving into. A lot of these tools are being bundled and they're being put into one app rather than being separate tools that you need to link together yourself. So, so what we had up until now is Dall e free Inside of Copilot. You could generate images, but now there's also editing features right in the image. You don't have to leave the app. You don't have to go to Microsoft Designer, their canva equivalent, where you could do things like click the image and then say, customize over here. And then it takes you to Microsoft Designer. And then this is their canva competitor. So you can do a lot of things here like remove the background and then I can add a new background, like. So it's really simple, but as you can see, it's a separate app. It's a separate tab. Now all of these things are converging into one. As of now, I don't have access to this yet, even with a vpn. But luckily Daniel from the team had access and he just tried it out briefly. Works super well. You just click the cat and then you can independently blur the background. You can use different filters you can turn it into watercolor painting like so, right in the app. So again, this is important because that's the direction that we're evolving into. Right? They're going to bundle all this together and these AI features are just going to be the next phase of software. It's not like a separate thing that is happening. It's just going to make all the software you use better and in some cases do stuff that was impossible before. Okay, moving on to a quite niche one, but nevertheless interesting. So we're staying on the topic of AI art and as you know, there's a whole sub community of people who are really into stable diffusion and more advanced tools like Automatic 1111 and ComfyUI that you need to install locally. And then you can use all the models and build advanced workflows or download them from people who built them. It's really great, but it can be quite demanding on your computer. Now, this new interface came out called Forge Web ui. And what it claims, and from what I've seen, seen and heard so far, is that especially on the less powerful GPUs, like a GPU with six gigabytes of VRAM, you can expect a 60 to 75% speed up. Okay, so this makes some of the models that were a real pain to use locally very usable. Now, if you want a step by step tutorial on how to use this, olivio Sarikas on YouTube actually made a great tutorial. He leads you through it step by step. It's actually totally doable for beginners. Just set aside an hour and open up ChatGPT that you can ask along the way if anything is not clear. And I'm sure you'll figure it out. But this is really great because it makes some of these intermediate to advanced tools accessible to everybody, independent of the computer you have. I love that. And while we're on the topic of running things for local interfaces like ComfyUI or Automatic 11.11, there's actually two new ComfyUI workflows that I just briefly want to highlight. And look, I'm no expert on these. I spent a few hours playing with ComfyUI and figuring out the basics. But if there's interest, and this is hard for me to gauge, I'd be happy to spend a weekend really diving into it and then presenting you with a basic tutorial that shows you how to get those first steps done very efficiently. From what I see on this channel, those videos seem to be useful to you guys. So if there's a comment with 10 or more likes that says yes, please. A tutorial on ComfyUI. I'll get on it right away. So why would you want to use that? Well, you can do things like this. Okay, so I'm going to show you two quick Reddit posts of things that you could generate in there. Right? So one of them would be these 360 images generated with stable diffusion. As you can see, they seamlessly connect. And obviously this would be very powerful if you're building some sort of environment on something like the Apple Vision Pro. I mean, it's a great starting point. And as we talked about in the last use case, with something like Forge ui, it's getting increasingly simple to run some of these models. So at a certain point, you're going to have environments like this that are generated on the fly while you're in the VR or AR experience. This is getting closer and closer. And another example of this would be this 3D workflow inside of ComfyUI, where you go from a simple image to a full 3D model in about 30 seconds. Now, this one might not be for beginners, and it's probably not useful to most people. Nevertheless, here's the GitHub page with a guide on how to install this, how to run this locally. Luckily, it's still severely limited. For example, it only takes square images as an input. But this is something that wasn't possible at this quality and speed a week ago. So now you know what's happening on this front. Interesting. Okay, so let's get to our last use case. And this one is actually big because it doesn't just paint the trend that is happening, but apparently it's also a very useful tool, especially if you're interacting with YouTube playlists. Now, unfortunately, this is Windows only, and it's not just that you need Windows, you also need a concrete graphics card. So. So, concretely, you need the GeForce RTX 30 and 40 series GPUs. As of now, I don't have those, but I'm gonna keep an eye on this and it might just be worth getting one. So I didn't get to test this, I didn't get to use it, but basically Nvidia brought out a chat interface. So this is big because they're starting to move into the direction of full vertical integration. Right. As of now, they dominate the hardware space and the software front. They have a lot of research, but so far they didn't really have a competitor to something like ChatGPT. So far, this is their first step in that direction. Your personalized AI chatbot. Because if they're in control on both hard and software, I expect this to be more efficient than most other things over time. So it's very interesting to me that they entered the arena with a software product like this because now you can use all the open source models in there. And as those get better and their hardware gets better and the software gets better, well, you're going to be able to get everything you need from Nvidia themselves. I mean, you can upload local files and via rag chat with those. You can chat with YouTube videos. Those are some of the most important features for large language models. So if you have Windows and a graphics card, you can test it out today. So as you can see, today was really a mixed bag with some new features and powerful apps that we love like GPT4 or Copilot, but also all these efficiency releases and new capabilities of these image generation models. And in my mind a lot of these connect because if you think about the fact that it's easier than ever to run some of these image generation and 360 degree environment and 3D model generation models locally, well, I don't think it's a coincidence that these are coming out shortly after Apple Vision Pro release because these work together really nicely. And down the line you could totally expect some of these features built into the apps on devices like the Apple Vision Pro or the meta quest 3. It's going to be taking your voice input and using something like OpenAI Whisper, which is freely available and transcribes your voice. And then a large language model, either open source or one of the GPT models, it's going to interpret your language. Maybe there's going to be a smart system prompt in the background like the book Summarizer has, that is going to interpret what you're saying and give you the right type of result. Maybe the prompt is going to be set up to kind of figure out what you would prefer in that situation. What is your emotional state? What is your psychological profile? Right? Because just like the memory feature, it listened to you previously, so it knows a bit of stuff around you and it can make educated guesses. And then it's going to take those educated guesses and on the fly it's going to be generating environments that suit your mood and 3D objects that fit into those environments. Now this all sounds very sci fi, but I mean, look at these releases. This all kind of aligns this week, right? I found this week to be particularly interesting. And if you add the fact that you're going to have a creator economy in the gen AI space too, with releases like 11 Labs or monetized GPTs. It really is going to be a brand new world inside of these headsets or inside of existing software that we already have. It's going to become better and we're going to get features that seem unreal now. Okay, So I hope something of this is going to be useful to you. You could go ahead and create a song for a loved one, and if you care about more brand new releases and use cases, then check out the playlist because we do a video like this every single Friday. All right, I'll see you in the next.