This Uncensored Chatbot is WILD & More AI Use Cases

All right, on this week's music and news, we have a bunch of applications that you actually might want to start using regularly. Look, I cover all the new AI releases that you can put to work today, but sometimes those releases are not necessarily practical or applicable. I think this week it might be a little different because we have some advanced prompt improvement tools, shareable artifacts that's absolutely wild, and even some fun apps that you might want to throw on if you're just hanging out with friends or family. And I also got to say, it's so good to be back at the studio. I've been visiting family back in Slovakia for the past two weeks, still maintaining two uploads a week, but it just feels great to be at home. So here's all of this week's AI news that you can actually use. Starting out with anthropic announcements. They have been absolutely killing it. They didn't just release Sonnet 3.5, which now is the go to model for most AI power users, but they expanded on their artifacts feature. Just a quick reminder, artifacts are this feature where it creates an application and you can actually use it in there. And look at that. This is a one shot prompt that we'll look at in a second and create a playable game. But the new feature here is this publish button in the bottom right. And as you might already imagined, that's exactly what this is. Now I can publish my little game and share it with anyone with you. For example, there's a link in the description. If you open up a brand new browser and you go to this link, you can play my little space shooter like so. An absolutely incredible development because now players like OpenAI don't just have to implement artifacts, they also have to make them shareable to compete here. And while OpenAI hasn't made that move yet, other platforms like PO here for example did make the move and they integrated something similar on already. The difference is that in post interface you can use various models like GPT4O and then if you generate code, you can visualize and share the results right away. So it's kind of funny OpenAI's own interface doesn't have this yet. You need to use PO to get an artifact like feature to get it. Either way. This is very powerful and I created a standalone video talking about various Claude use cases. Matter of fact, I collect 21 of them. Half of them come with prompts that you could just use right away. The simple Space shooter being an additional example from within our community, we did a little challenge where people had to create games. And actually DOM went in and created this game. And then with further prompts, he customized it with his custom game assets that he created with midjourney. Matter of fact, in four minutes he's starting his very own lecture where he's going to be showing people how to create these game assets that you can then integrate in here to upgrade this exact game to something more like this. And with this new publishing feature, you could share the little artifacts that you built with anyone now. But that's not all out of Anthropic, they're shipping features left and right. I'm talking about the extension to the prompt generator. So we talked about this before, before. If you go to console.anthropic.com dashboard, as per usual, all the links are below. You can generate prompts here if you link your API key, meaning every request will cost something. So let's say I want to improve my classic test prompt here. Write me an essay about penguins. I will simply go in, say generate prompt, and this will take a minute and really flesh it out. It will add various building blocks, flesh out some of the context that is necessary for this. It's very similar to Sam, the prompt creator that I shared with you recently. Here a free GPT to improve your prompts without needing to spend tokens with Anthropic. But as you can see, it fleshes out the prompt. This is wonderful. We'll just go with the version as is. And here's the new feature. Click on continue and it goes into this next step where you get to activate some additional features. Which ones? Well, let's just go for the entire workflow because you start by defining a variable. As you can see, we have a topic variable here because this is a writing prompt on a specific topic, not about Penguins, as this prompt generator will create prompt formulas for you. As you can see, some of the prompt is customized to what you typed in originally, but it does keep the topic open. So we just say the topic of this is penguins and we're going to run this and look at what it presents you with here. It is a standard response. All right, but here's the new part. If you go to evaluate, you'll find this new interface where you see that as one variable we defined penguins and this is really a fantastic interface where you can battle test your prompts with various inputs. Because if you're building something like an application or integrating the prompt into an automation, you really want to make sure this works to according across a variety of use cases, not just in One case. So I could add more rows here and for example, say elephants. And then I run the prompt and here's my favorite part. A prompt like this is really easy to test because you just have to switch out one word and then you get the results. All right, but what if there was four variables? And what if it wasn't just a topic, but maybe a customer support email? Well, you would have to find these. And often just testing the prompt would take more time than crafting it. Well, here we have another new feature called Generate test cases and it will just go ahead and come up with a word that fits again, in this case it's just a topic, so I chose bees. But in other cases, it can do everything that Claude can do for you. So here we have free test results and I could go ahead and rate these results. Let's say I really like this one, this one is good. And there you go, the beast generated. And I really like the beast result, let's say. So now I can go ahead and add a comparison prompt. Okay, so for that I need to go back to the prompt interface and instead of engaging, we'll make it boring. And now when I go over here and just say rerun, it will save this as version 2. And then when I switch over to the evaluate tab, I can add a comparison and then I can compare these because as you can see, here we have V1. Here I have V2. This has only been ran on Penguins. I'll run this on the other variables too, and then I could rate it, have it side by side. It is really the perfect interface to craft and improve your prompt. Just always keep in mind that running this will always cost you a few cents and you will have to top up your balance here and there. As opposed to the subscription model where you just pay $20 a month and you have access to Claude and with the rate limits that come with it. Fantastic tool. One final note, if you're curious about this prompt improvement workflow, well, they actually open sourced the meta prompt behind it. I will also include a link in the description for you to check out this full thing. But again, let me point out that we also have a free GPT that does the same thing for you equally as well or as many people report even better. And it's completely free. You just pick a button and build a prompt. All right, enough prompt engineering for today. Let's get into some of the exciting stuff. Prompt engineering can be exciting sometimes, but not most of the time. Right was a bit more spicy. Is this Venice Chatbot that comes with 20 free messages. Okay, so sometimes I just feature these random chatbots that have something super unique about them. And let me tell you, the uncensored model that this one is based on works like a charm. So I'll just say, write me a dark joke about generative AI YouTuber thumbnails. Let's see what it comes up with. Certainly a lot of angles that you could go with here. Why did the generative AI youTuber's thumbnail go to therapy? Because it was feeling pixelated and struggling to find its resolution in life. Oh no, please. You know what? That didn't work to my satisfaction. Let me throw this into the anthropic prompt improver quickly and let's improve this and see if we can get something better from a well crafted prompt. Now look, I'm just doing this for fun. This is not one of the good use cases. Right? As I talked about before, you really want to flesh out prompts. If you're going to be reusing them regularly in automations or automating them or handing them off, this is kind of just a fun way to do it. But you know what, I'll just run this and I'll make sure to this do define the topic. Say clickbaity generative youtuber thumbnails. Let's see what we get here from the news Hermes 2 data model based on llama 3. Why did the generative AI YouTuber thumbnails start a cult? Because they realized they could create clickbait thumbnails that would lure viewers to their channel, then brainwash them into worshiping the AI's infinite creativity. Or so they thought. Until the AI decided to replace their thumbnails with disturbing images of their own impending doom. Now the youtubers are trapped in a never ending loop of clickbait thumbnails forever tempting their audience to watch, but never truly free from the AI's grasp. Is this what's going to happen to all the youtubers who use clickbaity thumbnails? Oh boy, oh boy. I sure hope not. I might be guilty of this myself here and there, but what I learned over time is you just got to play the game. If you want to do YouTube in 2024, you got to amplify the messaging on the thumbnail sometimes. But I think that's the key point. It's about amplifying it. Not just flat out lying to get people's attention. Those people will be trapped in a never ending loop of clickbait thumbnails themselves. Maybe. Point being, this model is Completely uncensored and quite unhinged. You can ask it anything. You can write, dark humor, or anything else that you might come up with. Before we move on to the next one, please leave a like on this video if you're enjoying this content. I know people say this all the time, but it really does help the channel. And a video like this takes anywhere from 30 to 40 hours of collective work to put together. Research, testing, reviewing, recording, edited by two people, reviewed by two people, then the packaging, then the publishing. We do it all for you every single week with a lot of love and a lot of enthusiasm for all of this. Leave a like if you enjoyed. And now let's move on to the next use case, which will be quick but very interesting. This one is by zoo.dev and it's called text to CAD. And what it does is quite simple. It generates these CAD models for you. This is a minimalist desktop stand for my iPhone. I guess you could kind of 3D print this. I tried this a few more times before, it didn't work too well, but there you go. This is a 3D generator that outputs various 3D files for you. I guess you only have three generations on the free plan, but kind of interesting because usually these 3D generators, they try to create a full 3D model, including texture, lighting and everything. And this just produces the blueprint for a CAD software, which is essentially a raw 3D model. Interesting. Now you're aware of it, we can move on to the next one, which, as I promised, is actually the fun one of this week. All right, we covered this app before Kreia AI. They always have fantastic free plans. But let me just show you something new that I stumbled upon a few days ago. Maybe this has been here for a while, but you can actually click this camera tab and. And it takes in your live camera feed. Now you can do the same thing with screen sharing. But look at this. This is the one where I would say you could actually pull this up with your friends or family and start having a good time. Because I can just full screen my camera as I'm recording this and then you can prompt on top of it. Okay, so let me scale this up a little more and here I am. So look, I just prompted surprise youtuber and what's going on? Okay, what about a toxic green two horned unicorn? That sounds kind of cool, right? Look at me, look at this. You can kind of like readjust to it. Okay, let me just put an order in a bathtub and see if I can say hi. To you in here. Okay, this is really fun. I've always been told I look a little like Italian greyhound. I always like the little dog is a lot. Let's see if this actually kind of works. What is this? What's going on? Okay, one last one. I'll just do something super random. A samurai training under cherry blossoms. Okay, how about this? Come on, give me a katana. Ah. Okay. All right, all right, all right. So as you can see, this thing is not perfect, but it's so much fun. And you can just get super creative with it. If you're in a group, multiple people can just put in random fakes and it recreates you. You live. I think I only logged in with my free account here. Right, and it just does it for you. No payment required as of now. I have another good one which might not be so fun, but is super convenient because we looked up many background removers over the course of the past. What is it, like, eight months that I've been doing this every single Friday? Yeah, I think so. I think the first episode was in November. Anyway, the point is, if you have an image, like, let's say this YouTube thumbnail that we had here once, I can drop it in and it will process it. And here's the cool part. It will not just remove it with one model. It will remove it with all. All of these various background removal models. Okay? And sometimes they work super well and sometimes not so well. So I can just go through and look at the key it got, and in various situations, various models are going to shine. For example here, this one looks really solid, whereas this one doesn't really do what I want it to. At the time of this recording, this is completely free. They even say that the app is free. He released the source code. You could run this locally, you could run it in the web app. You could just download the PNGs in second, like so. No login, nothing. We even featured this as the app of the week on our newsletter because we always like quick features like this that are convenient. Look at that. That's a PNG that was created in seconds. Sure, there's the text, but that's the nature of the image. No other tool could have done that better. Lovely little tool. And then to round things out, I just want to point out one YouTube feature that now uses AI, which I am ecstatic about. This is something that I wish existed years ago. Okay, so here's the deal. They applied AI models to separate music from the other tracks in the videos, and now they have a new feature that if you get a copyright strike because because you have some copyrighted sound in your video, you can actually split the music from the dialogue and only delete the musical track while maintaining your dialogue. Now just to be clear, the way this worked up until today was that you had to either delete the entire audio track and you just lost the content of the video, making it unviewable, or in the best case scenario, all the monetization from the video went to the copyright holders of the track. Now this gives creators way more freedom because a lot of times you're not even exactly sure if something is copyrighted or not. Now I can feel a lot more secure in my creation and especially my life. If you do a two hour long live stream and there's one five second clip where you show something that you shouldn't have, or you play some song that you shouldn't have, you don't have to demonetize the entire live stream. You can actually keep it as it is, split the stems, remove the music, keep your dialogue, and keep monetizing the content. I think this is absolutely massive for content creators all around the world. YouTube is super strict with the music. That's why I see this appearing here first. But just think about the possibilities. Think about the other use cases that AI technologies open up. We're looking at background removers and fun little apps and little self made games right now. But just think about what happens if all these features, all these standalone AI applications, start being integrated into things like YouTube, into things like your iPhones. The world is changing and I'm here every single Friday to guide you through this development. So at the very least you can know what's happening and face this massive technological shift that we're all living through with open eyes and full awareness, rather than the approach that most people take which is just pretending like nothing is happening. And then one day they'll just get hit over the head with a change that might be too large to ever recover from again. So if you enjoyed this content, share this with a friend or family member that might benefit from all of this practical education and I'll see you next week on the next episode of AI News. You can use See you later.