Why Is Nobody Talking About These AI Use Cases?

Another week full of brand new artificial intelligence use cases. So let me show them all to you. Because we didn't just get a lot of upgrades to apps that already worked super, super well, but there's some brand new ones that I haven't even considered possible. Like Magnific 16x upscaler or Google Maps. Receiving some much needed upgrades, but there's some that popped up that haven't even occurred to me. Like just taking an audio file and turning it into a complete 3D scene that looks photorealistic with the characters mimicking the voice. A lot to talk about. So let's get right into it with the first one which is Fix Key Key AI. And this is another one in the category of I didn't even think about this. But this makes so much sense. The concept is really simple. It's a better autocorrect. And as somebody that struggles with typing regularly, I don't know if it's my large hands or it's the fact that I'm always in a rush. I make a lot of mistakes while typing on my phone. And let me tell you, the autocorrect on there is quite annoying. I speak English most of the time, but I have a lot of German speaking friends and my family Slovak speaking. So I have three different languages. I always have to switch back and forth and then when I'm typing in Slovak it's suggestion corrections. In English it's a Mess. So FixKey AI promises to do this better with the power of AI. So you can just download the demo up here and once you install it, you can just get started. Give it access to the keyboard and the first 15 fixes are free. To be honest, while I'm at my laptop, I don't make as many mistakes. But on my phone this would be great. Let's try this anyway. Hello, what are you doing? So this is supposed to say hello, what are you doing? And I'm going to hit the keyboard shortcut. Hello, what are you doing? It Got it. Okay, one more test. Okay, so it selected really well. I'm in a rush A might be making mistakes along the way. Okay, second try didn't work so well. One more time. How well does this work? If I just include minor mistakes here and there, that's pretty good, but not flawless. So as you can see, it got one out of three perfectly. Not great, not terrible. I think the real use case here is the phone. And over time I expect players like Apple to integrate this. But I will start experimenting with this more as it's Something I actually need. All right, moving on to the next one. Magnific AI got a fantastic update this week. A 16 times upscale. Check this out. Amazing. So I just brought in some old Pokemon pictures here, and yes, this one is paid, but in a second I'll show you a free alternative. But here's the before of the original Pokemon Yellow. And here's the after. Okay, the difference here is the resolution is way higher. I can zoom in like this and it's perfectly sharp. All right, so let's give the tool a fair shot here. Let's take one of these old Lara Croft images that people have been upscaling. There you go, 16x. I leave all the sliders as they are and actually it can't go beyond a certain resolution. So here I have to go to 8X and we're going to get a 40 megapixel image of Lara Croft from, I don't know, 25 years ago. So almost 30. Okay, so here's the first result. Okay, this is actually decent, but I re ran it with the video game asset preset as that is only fair. Right? So that's this result right here. Oh, my God. So this is interesting. By no means is this great. So I think the more you upscale, the more variety you will get. So be careful with this. If you stick to 2 or 4x, it will be consistent. Some of these 8 or 16x results have been all over the place. Nevertheless, you can get some really good results with them. All right, but Magnific is. But I told you, there's a free alternative. And here it is upscale with an Y. As per usual, you can find all the links in the description below. But this is an open source alternative that upscales images locally. So you can go ahead and download this. I'm going to download this for Mac. And once I open this, I have a local version. This doesn't require Internet. It's open source. And if I run this on the same Lara Croft. Let's see what we get. Look at that. That didn't actually improve the quality. It upscaled it. It has more megapixels, more sharpness. But it didn't actually do the AI magic that Magnific does where it actually generates a new image that it overlays. So I'll run it again, but this time I'll change the mode. It doesn't have as many presets as Magnific. So I'll just take Digital Art here and we'll stay away from the double upscale. And by the way, while this upscales, the way these double upscales, these 16x giga upscales work is they just do it multiple times so you lose even more of the original image. But therefore you get these massive files that you can zoom into a lot and they have a lot of rich detail. But here we have the result and this is actually better. But it didn't do the magic that Magnific did. So it's high resolution, but it really preserves the original here. So just a different tool. This one is completely free. Feel free to experiment yourself. A lot of these AI tools we're seeing in 2024 seem like magic. Magic. But once you understand how they work, you realize it's just technology. You can even learn how to build them yourself. And the great thing is that it doesn't require you to pay an insane amount for college tuition or spend years of your life stuck in a classroom. And that's why today's sponsor, Brilliant.org comes in. Brilliant is an interactive learning platform where you can deepen your understanding of AI, data analysis, program programming and more. They already have thousands of bite sized interactive lessons available and they're adding new lessons every month. A quick quiz when you sign up assesses your interests and current skill level, customizing your learning experience. So it's great for both beginners and advanced learners. The truly great thing about Brilliant is how interactive each lesson is. You're not just passively absorbing information, you're actively solving problems that challenge and expand your understanding. One course I would recommend taking is called How LLMs work. It's particularly great because it breaks down important topics like temperature and tokenization into easy to consume lessons. With Brilliant, you're not just learning the what, you're mastering the how and the why interactively. So here's the deal. Head on over to brilliant.org daidvantage and you can start this incredible journey for free for a full 30 days. The first 200 people to sign up using our link in the Description will get 20% of an annual premium subscription. Start deepening your knowledge today with Brilliant. And now let's get back to our next AI use case. Okay, now let's talk about a funky one. This one is called Dream Talk and it's released by Alibaba. And if you've been paying close attention, they have been releasing more and more research on these different use cases. And this one in particular is probably the best face animator that is out there. You can look at a comparison here. So the middle one is theirs. But the thing is, these Demos are always so cherry picked, right? So in the video it looks really good. It's like, wow, that's excellent. But if you actually go ahead and use it, which you can do under this link. Well, even if we use some of these preset images, like let's say Angelina Jolie over here, and then we use the audio input that they picked as a preset, right. So we're not even giving it new variables and not even. That is great. So I'm sure you could make it work somehow if you play a lot. But ah, this just isn't it. I guess it's better than the other terrible ones we've seen. But I just say these aren't in the very usable category yet. So let's move on to the next one. And this is the exact opposite. So this one is called Jan AI and it's a better alternative to Ollama that I showed you last week because this is even more user friendly. No terminal, it just works. So this is a good alternative to app that is out there called LM Studio. And both these allow you to run some of these large language models like locally, the ones that can be run locally. And the cool thing about Jan is it combines GPT4 and 3.5 via API and some of these large language models that you can run locally, like Llama or Mixtral, into one app. So by downloading this, you could actually go ahead and start using this yourself. That's how simple it was. This took under a minute. So if I just explore the hub, you'll see all these different models in here. And the cool thing is they're actually super swift with updating this because the new Mixtral 8x7B model that just came out is already here. So you could download this and use it locally. But the thing is, these models do have a minimum requirement. So I'll do a separate video where we talk about this more in depth and I'll give you an overview. But for example, this new model actually requires a minimum of 100 gigabytes of GPU RAM. No normal human being will have this on their machine. So you have to go into some of these compromises, like using the Llama model or like using the Tiny Fee model. Again, I'll do a separate video on this, but if you just want to play around, I just recommend getting this Tiny Fee model. Most people can run this, it's only 1.8 gigabytes. Or you can alternatively use GPT4 through the API through this wonderful interface where you can just have multiple Chats with different chatbots with different large language models all in one place. This is new and it's completely free as long as you use one of the local models. Okay, so next up I want to point your attention towards this meta research project and they're coming up with increasingly weird stuff I actually really enjoyed. But this one we're not actually going to be using here today because there's a GitHub link and you can install the whole thing. I'm not going to go through that now. But I just to want want you to be aware of this concept. What they do is they take a audio waveform and they turn it into a 3D scene. Okay. So it's called audio to Photoreal and if you actually install it on your local machine, you could just upload a waveform and it's going to generate the scene of a photorealistic human being that is speaking and presenting the audio that it heard. That's kind of insane and I just wanted to bring your attention to it. Here's a quick demo. I mean not late but you're supporting than me. That was. Yeah, yeah, yeah, yeah, that was, that was me. So why are you so again this has been generated just on an audio file. This is pretty trippy. So let's just move on. Most of you probably have used Google Maps in one form or another. For me personally, it's actually my navigation app of choice. I found it to work more reliably than Apple Maps over the past years might have changed. I just stick to Google Maps and they got a significant AI update with two things that I personally really enjoy here. So amongst many new features in there, like this advanced navigation that actually shows you the lane that you need to use to exit a highway properly. This is not an AI feature. Let's talk about the AI features here. And the one I'm most excited about is this one called Google Lens in Maps and they're using AI to enhance your understanding of the surroundings. So you can literally just whip out your phone, use the camera inside of the app and when you point it at a subject, all of these little pop ups are going to show up. That's actually incredible, especially when you're traveling. And the second AI feature here is you can actually look for specific words that have nothing to do with the city. And it's going to use AI image recognition features to actually find you shops with that. So in the demo here they look for animal latte art and it shows all the cafes that actually have that as a part of Their reviews. Right? So now you're going to be able to find all the cafes that do animal latte art. Now, this might be a silly example, but you get the point. You could be searching for a specific object and it might have nothing to do with the name of the shop, but they offer it. So maybe they have a bunch of pictures of the shop and a bunch of reviews include pictures. You're now able to search through all of them through the power of AI. This is incredible. And I'll have to test this more. But now you know that it exists. But we're not done for this week. Another one. Because as you might know, Midjourney V6 now officially released. Everybody has access to the full version if they're a paying subscriber. And there's this one use case, which I thought was exceptionally practical. A lot of the AI art stuff is very fun. Fun and interesting and artistic. But when we're talking marketing use cases, actually generating images like this, where you have devices with a green screen that is showing up, is incredible because this allows you to really easily replace it. So shout out to Hugo Ventura to actually featuring this on X and actually like this so much that I brought this up in our course community. And one of our team members, Dominique, that is super into midjourney, actually hopped on this and he created an entire little guide of different ways how you could be using this. You can use this as billboards or computer screens. You could use this as a mock up of a bedroom for interior photography and a bunch of more use cases that he came up with here. Just a quick side note, the course community is going to be closing down over the next two months because we're going to be building it out and I don't want new people streaming in while we build it out. So if you want to be an early part of the AI Advantage community, enrollment to the course is open until January 19th. Then we close things down and build more things like this guide that you can find in there. Okay. And for this week, I have one more use case for you to review here. And that is the Perplexity update that you just came out. And I think this is actually a big one because as you know, I'm a big proponent and fan of vision features inside of large language models. I think it's the easiest way for everyday people to actually prompt them and get useful results because fleshing out all the aspects of the context properly is not an easy task and it's just not realistic for most people. You need some training, practice and patience to do it. But Perplexity added image uploading features. And if you didn't know, Perplexity is an alternative to ChatGPT, which specializes in research use cases. That's at least the way I would put it. They're really good if you're researching new topics because they crawl the web widely, give you references to everything, and it's just a better experience if you're researching something. And by the way, if there was any sponsorship, I would clearly disclaim that. I just enjoy Perplexity here, so I recommend it. And now with the image upload features, I think it's even more powerful than before. So if you're researching stuff, this is definitely worth a try. So if you found something in this video useful, you should probably subscribe to the channel, because I do this once a week and I only focus on on things you can actually use. No drama, no announcements that might be available sometimes in 2025. No practical AI use cases presented to you in a concise manner. You can check out last week's video here, the full playlist over here, and I'll see you next week.