7 Fascinating AI Tools You've NEVER Seen Before (Underground AI #4)

Hey, welcome to another edition of Underground AI. This is a series I started creating where I tried to uncover and show off some of the lesser known AI tools that are out there. These aren't the things from companies like OpenAI and Microsoft and Google. These are the solo creators and the smaller builders trying to find cool ideas to use this current crop of AI technology that's available now. In this episode, I'm going to try something a little bit different. I'm going to play a little game that I'm going to call should this be a Standalone App? And the reason I want to do this is because I run this site called Future Tools over at FutureTools I.O. and I curate all of the coolest AI tools that I come across. However, I get somewhere between 50 and 75 new tool submissions every single day, and I still personally look at every single tool. The problem I've been running into recently is that every tool kind of is starting to look the same. And most of the tools that get submitted, people want to charge you money to use them, but there's free alternatives. Or if you already have like a subscription to Chat GPT or Anthropics Claude, you can do it with that existing subscription. So a lot of these tools don't feel like they should be their own standalone app. And while trying to find some really cool tools to show off, I found some tools that are doing some stuff different. They're kind of novel concepts that I haven't seen other people submit to the Future Tools site. There's probably a reason for that. Quite honestly, most of these tools, I feel like they don't need to be their own app. And there's tools out there that you can do them either for free or with a tool that you might already have a subscription to. Now, before I get into this, my goal with this video is not to harm any small builders or software developers that are creating these AI programs. In fact, I absolutely want to see new novel ideas getting submitted to this site. I want to see ideas that I haven't thought of yet or that I other people haven't thought of yet. Now, some of these tools may be genuinely useful to people watching this video. Maybe you can do something with Chat GPT, but a tool that only costs three or four bucks a month does the one thing that you really need to do so you can save the money on Chat GPT and pay the lesser cost. I also want any of the builders that are creating these tools that might be watching this video to just kind of see it as feedback because if you can figure out how to improve on the tool or build a community around the tool or, or do something that differentiates it from everything else out there, you still have a chance to win. Now with that all out of the way, let's dive in and check out our first tool that was submitted to Future Tools called Locus. And you can find it over@locusextension.com now it does say across the top Locus Extension is free, but it also has a pricing tab. So if we click the pricing tab and scroll down, we can see they have a free plan. But they also have a plan that is $7.50 a month if you pay annually. But let's be real, it's nine bucks a month. There's also some really funky stuff about this website are not really doing themselves any favors. They have like an explainer video up here to tell you what Locus does. But when I press play, there's absolutely no audio. In fact, the first time I pressed play on this video, I jumped to another website to make sure my audio was turned on and there wasn't anything weird going on. But no, they just put an explainer video at the top of their page with no audible explanation. Like it doesn't seem that diffic to even take a script, throw it into 11 labs and get 11 labs to speak out the explainer for you. The other thing that I found a little funky about their sales page is if I scroll down here to the bottom, we can see their pricing. And the paid plan at $9 a month summarizes with GPT 3.5, which is odd to me because anybody can use GPT 3.5 for free. That's what ChatGPT's free version uses. And they're charging nine bucks a month to to use what we can all use for free. Like why aren't we using GPT4 here? Or Claude Opus or one of the more state of the art models that's available. They also put a little star next to summarize with GPT 3.5 and it says GPT summarization is limited to X tokens. Did you forget to put the amount of tokens there? But anyway, the idea behind this tool is an extension for Chrome that you install it and then it's supposed to be like a better find function. So if you're reading an article, you can search anything you want and even if the specific word you're searching isn't mentioned in the article, it will still find the context of what you're asking and answer the question. For you. So let's go ahead and try this out. I'm going to go ahead and add this to my browser. Put it up in Chrome here. I'll pin it up to the top so I can see it. And we do have an option to continue as guests, so let's go ahead and do that. Instead of logging in, let's go ahead and go to the Verge here. Here's an article called the Teens Making Friends with AI Chatbots. It appears to be doing something up here. I haven't pressed any buttons. All I did was navigate to this page, but it seems to keep on loading up here. So we have a few options. Summary. If I click summary, it wants me to sign in to do anything. If I click on study cards, sign in to do anything, Brainstorm, sign in, recent sign in, favorite sign in. And somehow I've already used two of three of my daily documents. This is the first page I've looked at and I guess I need to sign up to really use this thing. Okay, so I just went ahead and went through the process to sign up real quick. Now let's try the summary button. I will click Generate and see what kind of summary it gives me. All right, so we've got like a one paragraph summary here. Many young people have found comfort in AI chatbots like character AI, etc. Let's look at what Brainstorm does. I don't know why it says regenerate. I haven't generated any brainstorms yet. All right, it gave me some bullets here, so I mean, it's okay. But this is the free version. I don't know what model it's using behind the scenes. Let's go into our settings here. Okay, so it looks like if I click on the generative AI service, there's Locus, OpenAI 3.5 and OpenAI 4. I have no idea what model Locus is. I don't know anything about it. If I select 3.5, I can't enter my own API key. If I select four, I can enter my own API key. So it looks like I can use OpenAI's 4.0 GPT4. They've got the naming conventions wrong here, but you know, we know what they're talking about. I'm assuming this is GPT4. I'll save the changes, I'll regenerate our summary, see if it looks any better. And it looks to be pretty similar. Now here's the thing about this tool. There's already a really good, really well known tool that does this pretty well. It's called Perplexity. I have the Chrome extension installed right here. If I click the Perplexity Chrome extension for free, I can click Summarize. And I'd say this gives me an arguably better summary of what we're looking at. It's got bullets, it's broken up a little bit better, it's a little bit cleaner. I can also ask questions about this page that I'm looking at. So this Locus tool just seems to be just what Perplexity already does. Also, if I have a Claude account for free, I can just go control a Control C, copy everything on the screen, open up Claude AI, paste in what I just copied to the clipboard here and say summarize this article. I have it set on Claude 3 opus, which is the paid version. So I'll set it to Sonnet so we can see what it would do with the free version and submit it. And once again, an arguably better summary. I guess my kind of thoughts on this Locus extension here is I don't understand why it needs to exist. The next tool I want to talk about is called Intellise xyz. This is a real simple concept. You basically speak into the app and tell it what you want to do for the day and it will turn it into a task list for you. If you don't do it in that day, it sort of knocks it down to a previously uncompleted items and you have the option to bump that item back to today's task list. There's some other features like adding it to a Google Calendar, but a fairly basic concept. Now ChatGPT and Claude and so these other tools, they won't give you a checkbox that you can actually select and check off. So this does provide that extra little bit of usefulness. But this one does cost money, right? It says three bucks. And this is a little nitpicky here, but one thing I'm not a fan of is this is actually three bucks a month. But they just put three bucks. I feel like if you're going to put pricing on your page, be transparent with it and say it's $3 a month and not just $3. Because this, when you look at it quickly, most people would probably assume, oh, it's just $3 and I've got the tool. But if I click subscribe to Premium and hit checkout here, $3 per month, not a big deal. You find out the price when you click on the checkout. I just wish it was a little more transparent on the homepage. Again, I'm Nitpicking. All right, so I just paid the three bucks. They've got some recommended tasks that you can add. They've got a microphone button, a plus button, and an account button with a manage subscription. And that's really it. Pretty bare bones here. So I guess the idea is you click the microphone and then talk about the things that you need to get done today. So let's go ahead and try this out, see what happens. Today, I need to make a YouTube video about underground AI tools. I want to make sure I tweet something. I need to check my email. I haven't checked my email in a few days, so there's probably a lot of emails I got to respond to. I've been meaning to read a book I'm going to read for 30 minutes today and I also want to do a workout. So I'm going to do a workout of body weight exercises later today and that's really my main set of tasks I've got to do today. Let's go ahead and let it process that and see how it turns it into a checklist. Not too bad. Make YouTube video. It didn't add any additional details about what the YouTube video is about. That would have been kind of nice, but it simplified it real good. Tweet something. Check email, read book, do workout. Exactly what it was supposed to do. Then if I check it off, it actually has a little bar up at the top. You know, fairly handy, if not pretty basic. Now I'm going to jump into GPT 3.5 on my phone here and this is the free model. This isn't GPT4. This isn't. You don't have to be a plus member. Use this. This is the free model. Let's see what we can do with this. I'm going to start by giving it some instructions and then I'll talk about what I need to do today. I'm going to talk about what I need to accomplish today. I would like you to make a checklist task list of the things that I need to get done based on what I say. I need to make a video about underground AI tools. I would also like to tweet something today. I want to make sure I get caught up on my email. Let's get to inbox zero today so I clear all that out. I want to get a workout in today. Specifically, I'm going to do a body weight workout today. I want to read for at least 30 minutes. I've got a good book that I've started on. I want to get 30 minutes of reading in. I also want to make sure that I have time to cook dinner later. I'll just close out of that and look at that. I got a checklist there. Create video exploring underground AI tools. Tweet about AI advancements. Clear. Email inbox to 0 body weight workout. Read for 30 minutes. Allocate time for dinner preparation. Now I don't have the little checkbox. I don't have the little status bar. It tells me how far I've gone with it. And I don't have the option here to save it to my Google Calendar. So there are some extra features that, that you're getting with that tool. But again, this is the free version of Chat GPT. I didn't even pay three bucks for this. But there are tools out there like Checkly, which is a free checklist generator. So you can just copy and paste them into here and get your checklist. So it is two tools, verse one. And while I do like the idea of this tool, does it really need to be its own standalone tool? I don't know. What do you think? All right, next up, I want to talk about Blogfox. This one's pretty interesting. Convert YouTube videos into high quality blog posts. So I can plug in a YouTube URL here. I can do up to three blogs for free and it will turn it into a written blog post. So I'll just go ahead and grab my latest AI news video. Let's go ahead and place this video here and try it for free. All right, so it's being finicky about the formatting. So I will grab this version of the URL. A lot of people use this version of the URL. So if I was this website, I would consider figuring out how to make it work with either variation. But let's go ahead and use this version. All right, so it's starting by transcribing the video. Now it's making an outline. All right, it wrote the blog, now it's putting the final touches on it. It's been taking quite a while. Does say this could take up to a minute. It's definitely been longer than a minute. All right, so here's what it gave us. This is the featured image. It generated a tank shooting a robot who's playing chess. AI news roundup, GPT4 video robotics and more. It's got our video actually embedded in it. You've got a little sort of beginning paragraph up here. And then we've got the rest of our blog post here, including an faq. So this isn't bad. I mean, it's fairly impressive. I'm fairly certain this was generated with the Dall E3 API. I'm just guessing, but that's kind of what it looks like to me, which makes me think this was probably written with GPT4 as well, because they're probably just going to use all of the OpenAI APIs behind the scenes. This one's fairly helpful because it does multiple steps where if you were gonna do this on your own, you'd probably use multiple tools. This kind of does it all in one shot. I wish it actually generated more images for throughout the blog post and not just a single image. So really, in order to do the exact same thing directly with one of the free models like ChatGPT or Cloud 3, we'd have to go to the video, come down to our description, scroll to the bottom where it says show transcript. Show the transcript on the right. Let's turn off timestamps. I don't think it really matters if you turn off timestamps or not, but I'm going to just to keep it clean. And then we'll go ahead and select all of this transcript text, we'll copy it, head over to ChatGPT, paste this whole thing in here, and then I'll give it a couple lines and then I'm going to give it my actual prompt. I'll say, please convert the above video transcript into a blog post with title and subheadings. I want to make sure it knows that everything I pasted above was a video transcript. And it did. Okay, it actually gave us some bullets here. It didn't give me a title for it. I guess we can say these bullets would be the subheadings, but it didn't really break it up into subheadings for us either. And if I do the same thing with the free version of Claude, paste in the transcript here, give it the same prompt, convert the above transcript into a blog post with title and subheadings. I can see already it did give us a title. They called it AI news roundup, the mystery of GPT2 chatbot and other major AI developments broke it up into subheadings for us here and wrote a nice little blog post for us. Now, if I want to get a featured image for my blog post, I got to use a different tool. I could use Dall E if I'm using the paid version of Chat Chat GPT, but if you don't have a paid version of chat GPT, you'd go to bing.com create and then we need to think of a featured image. Let's just type AI News and see what kind of image it gives us just based on AI News and we get some images that look like this. Obviously we can reprompt until we get what we're looking for. But I essentially just recreated everything blogfox does for free. It's just a lot more steps. This blog Fox tool does save you a whole bunch of steps. This one, to me seems fairly useful. The pricing, on the other hand, seems a little excessive for what it does. If you want three blog posts per month, it's nine bucks a month. Basically get what we can do with free tools. If you're using a tool like this, you probably want more than three blog posts per month. You're probably going to go more for like 15 blog posts per month. Now you're looking at 30 bucks a month. That's 2 bucks a blog post. And you can do it all for free with existing tools. So this is another really interesting one called geospy. You can find it over at geospy AI. And this is one that claims it can figure out where a picture was taken just by seeing what's in the image. Now this is one I'm not actually sure I can do with other free tools, but we'll find out. First, let's see how well this one works. I have a few images here that I saved on my desktop. These are actually photos that I took. I didn't find them online. These are my photos. And we'll see if it can tell where it is based on what's in the image. This first one I think should be fairly easy. This is Half Dome in the background. This is taken in Yosemite. But let's go ahead and toss this in here and see if we can figure it out. Country United States State California City Yosemite Valley. The photo was taken in Yosemite Valley, California. The rock formation in the background are Half Dome and El Cap. So yep, it got this one based on what's in the picture. We got El Cap and we got Half Dome. This next one was actually taken in Maui. Let's see if it can figure that out. Let's see if it thinks it's just Hawaii or if it can specifically figure out Maui. I don't know. United States, Hawaii. Kihei. This photo was taken on the island of Maui in Hawaii. So it got that one right as well. So far, two for two. And then this third one here is actually Milford Sound in New Zealand. I took it a long time ago. So the resolution isn't amazing. Let's see how it does. Country New Zealand. This photo was taken in Milford Sound, New Zealand. It got three for three. But you can't do it with the free version of chat GPT because the vision is only in GPT 4. But you might be able to do it inside of Claude. Let's try. I'll go ahead and upload the first image here. This is the Yosemite Valley image. Where was this image taken? The image appears to have been taken in Yosemite national park. And then some details about why it believes that. So it got that right. Let's try our Maui shot here. Based on the scenery with the palm trees, a coastal shoreline and a view of a mountain island or volcano across the water, it appears to have been taken in Hawaii. So it says I cannot pinpoint the exact island. The other tool actually figured out it was in Maui. And then finally, let's try the Milford Sound image that I took. Yep, this image appears to have been taken in the famous Milford Sound of Fjord located in Fiordland national park on the southwest coast of New Zealand South Island. So this one got two out of three. It did catch that this is Hawaii, but it wasn't as specific as Maui like the other one. So this geospy tool, pretty cool tool and it's free, so it's hard to say like, hey, you can do it with other free tools because, well, this one's free, so why not use it? And it does seem to work better and be more impressive than what you get out of free ChatGPT or Gemini. Next up, I want to play around with Recraft AI. This is one that I actually shared on my underground AI livestream that I did a couple weeks ago. But the live stream is watched by less people than that watch these recorded videos. So I thought I'd show this one off again because this is actually a legit tool that I think is very worth playing around with. If we click on Try Recraft Free, we can click on create new project. And this is like a giant Figma board that you generate AI images on. Let's go ahead and create a new vector image. We have a whole bunch of options here for vector art, seamless vector line art, cartoons, flat images, etc. Let's do line art and let's do a wolf howling at the moon. Let's go ahead and click Recraft and see what it generates. And it generated a kind of clip art looking vector graphic of a wolf howling next to the moon. But the cool thing about vector graphics is that they're infinitely scalable. So if I zoom out here, I could scale this up as big as I want and it will actually never lose quality. And then when I go to download it, I can export it as an SVG file here and use it in tools like Adobe Illustrator. So this is actually a great platform for creating things like logos. You can generate great images with Midjourney and Firefly and Dall E and Stable diffusion, but none of them are generating vector graphics like this. I also really like the layout, how I have this sort of like infinite canvas. I can move this over here, create a new image that's not a vector. We have a bunch of different styles. Let's do this Kawaii style here. And also do a wolf howling at the moon. And we can see that this image is still really big because I stretched it out. This image is tiny in comparison. But we get an image like this. Now this one, because it is a rasterized image and not a vector, if I scale it up really big, it should actually lose quality over time. Although this one's holding up pretty dang good. But you can start to see some of the pixels start to show up a little bit more when I really, really grow this image in size. Now, this one, if I go to export it, I don't have the option for an svg. It won't make it a vector. However, there is a button here that says vectorize. I can click this vectorize button and now I do have a vector graphic. And if I go to export it, I have the option to export it as an SVG file. And it should be pretty infinitely scalable. And this tool is totally free. So if you need to make logos and things like that with AI, this is probably the first place that I would recommend. And there aren't really any great alternatives out there that make vector graphics this good, other than maybe Adobe's Illustrator that now has Firefly AI built in. It will also generate vector graphics inside of Illustrator. But then you got to pay for Adobe Illustrator. This one, I've done everything you've seen so far for free. Next up is one I came across called Brainy Docs, and this is another one that you can do it with free tools, but it will take a lot more steps. So this is one that just does the whole thing with a sort of single shot. But if you wanted to recreate it with free tools, you essentially could. It'll just be a lot more steps. So this is one that you give it a PDF and it turns it into an explainer video using AI. If we check out the pricing here, they do let you make a one PDF into a video and it has to be a five page PDF, which I mean not super great. To test how it goes, free PDFs a month is 10 bucks a month. That only gives you 30 total pages a month. I mean some of the archive.org like research papers are more than 30 pages. You're paying 10 bucks a month and you might get one video out of it. If you're trying to make it out of some sort of complex research paper. 30 bucks a month for 10 PDFs 100 bucks a month for 40. I mean you're paying 3 bucks per video that you generate. And that's assuming each video is like a 10 page document or less. So I think their pricing is kind of weak. I not a fan of their pricing at all. Let's go ahead and try it and see how well it performs. So when I log in, I pretty much have this screen. It's got a document area and a summarizer that's coming soon. So really not a whole lot to it. Very, very, very bare bones right now let's go to archive.org let's search machine learning. And as we can see as we look through some of these documents here, this one's 33 pages long, this one's 9 pages long, this one's 6 pages long. Almost none of the research papers on arXiv.org are going to be short enough to do the free version. Just out of curiosity, let's find out what happens if we try to upload one that is too long. So here's one that's 33 pages on multi space alignments towards universal lidar segmentation. I come back to Brainy Docs here and create new video slides so I can drag and drop the document here and let's go ahead and generate video slides. Please upgrade to get more document pages exceeded. Number of pages limit for the month available pages 20 interesting. So saying I have 20 pages even though their pricing set up to five. Let's go ahead and find a different one here. This one's only six pages. We'll go ahead and submit this one instead, see if this one works. All right, so video processing, please wait a few minutes. It's giving us our original document here. It does have a lot of visuals in the document so it'll be interesting to see if it incorporates those visuals into the slides themselves. Okay, so it finally generated the Video. It took quite a while actually. It took a good five, six minutes or so and it generated a video here and then we can also see the slides that it generated. A separability based approach to quantifying generalization and it put some information on the slides. It did add some images to the slides. So I'm happy to see that it actually pulled some images from the PDF. Clustering in different models, training and inference process. Actually pretty impressive. A lot better than I thought it was going to do. It actually pulled in the graphics straight from the PDF and gave us some details, a conclusion, acknowledgments, et cetera. Not too bad. Now let's see if we actually watch the video. This is what the video looks like here. This presentation explores a novel approach to quantifying the generalization capabilities of deep learning models, specifically focusing on which layer of network best facilitates generalization to unseen data. Generalization is a key aspect to deep learning networks. Not bad. I mean horribly monotone. They should definitely integrate with like an 11 labs or one of these tools that actually gives the voice a little more personality. But it's not bad. I mean it's a video that explains what's going on in the presentation as slides, has visuals on the slides. Pretty good actually. I wanted to do this with free tools. It would take a lot of steps. This actually condensed a ton of steps down into a single step. So for example, if I was to use Claude, could pull this PDF in here and then give it instructions like I want to turn this PDF into a slide presentation, break it out into what information should go on each slide. Then as we can see, it'll actually break out what information we'd want to copy and paste onto each slide. It's not pulling the graphics for us, so we'd have to manually do that, maybe screenshot from the PDF. But once we have all of our slides, I can do something like I want to narrate over the slides to explain this research paper. Please write the script for the narration. It then goes on to write the actual narration that I would talk over for each slide. Copy this, go to 11 labs, paste it into 11 labs, click on generate speech. Generalization to unseen data remains a challenging problem in deep learning for tasks like classification and and foundation models. And then what I would do is I would essentially copy each bullet here, paste it on its own slide in something like Google Slides, export all of the slides, pull them into something like DaVinci Resolve, overlay my audio over each slide. It would be kind of a pain. It's a lot of steps. This Brainy Documents is actually something that I think would save quite a bit of time if you're trying to make a sort of automated slide presentation based on a PDF. Again, my biggest issue with Brainy Docs is how they've priced it. I mean, this is just kind of unreal to me. You really would need to use the $30 a month plan. 30 total pages is kind of useless, especially seeing how that's about the length of a single research paper. Over on archive.org with 100 total pages, you're more realistically getting about three videos a month. The pricing is not great. Even up to this one, you're probably going to get 11 or 12 videos a month out of it. So for 100 bucks a month you can make videos off of research papers. It's a little pricey for my taste. I probably still would do a lot of the manual processes or, you know, maybe hire somebody to help me do the manual steps on someplace like Fiverr and it would probably end up being a better value. I think it's a really great time saving tool. Just too expensive for what it is, in my opinion. In my opinion. And then the last one I want to show off is just a really, really fun game. It's totally free. It's something that I found myself just sort of playing with for a long time, longer than I'd like to admit. But it's a game that uses AI and you can mix elements and create more things. And the things you could create are pretty much infinite. Whatever you can think of, you can create. So if I go and make water and mix it with fire, it gets us steam. If I mix wind with fire, it gets us smoke. Smoke plus steam. We have cloud, earth plus cloud, we have rain, rain plus fire, rainbow. You can get it to create some pretty creative things by just continually combining stuff. I've seen all YouTube channels dedicated to playing with this tool and trying to get to really crazy words. If I mix rain with earth, I get plant, plant plus smoke, incense, plant plus water, swamp, swamp plus fire, dragon tree plus rain, forest, forest plus dragon, tree, ant. But you can get crazy stuff. I've seen people get like AI, robot and things like that. And it is using AI. I think a lot of these earlier ones are sort of cached in there, so it just automatically knows what the combination is. But once you've played for longer and longer and longer, you'll discover new words that nobody else has ever discovered inside of this game. And it's using AI to Combine the elements and decide what the output should be. It's really, really addicting. It's hard to describe without playing it because it seems pretty boring and lame. I've spent a lot of time messing with this tree plus Rainbow Unicorn Unicorn plus Fire Phoenix. Anyway, I'm going to end up going down the rabbit hole if I try to show off too much more. Check it out. It's over at Neil Fun Slash Infinite Dash Craft. Really fun. Really addicting. Check it out. It's fun. You'll like it. Nothing comparable. And let me know what you think of this video. I was planning on doing an underground AI and just showing you a bunch of really, really cool tools. But all the cool tools that I thought to show you as I was showing them to you, I kept on thinking of alternative ways to do the exact same thing for free, and so I figured I'd share those as well. But I also don't want to like hurt the companies that I was talking about in there. I pulled these ones specifically because they're on the right track. They're doing things that a lot of other AI tools aren't trying to do as standalone products. They just need to figure out their differentiator. They need to figure out a way to make them stand out and do things that I can't do with free tools already. Otherwise their business probably doesn't have a lot of longevity. At the end of the day, I make the videos for the viewers, not for the people creating the software. If you're watching this video and you came across a tool and you're like, I don't want to spend this money per month, well, hopefully you learned of some free alternatives to do the same. Ideas? Let me know what you think. I'd like your feedback. The underground AI isn't a new concept, but the should this even exist as a standalone app was kind of something that kind of became a theme of this video as I was actually recording the video. It didn't start out that way. Anyway, I just want your thoughts. Should I make more videos like this? You like this video? You want me to show you more paid tools and also the free alternatives? Or do you want me to just stick with here's a bunch of cool underground AI tools that nobody else is talking about? I am very open to learning and improving these videos for you so that I can provide as much value as possible. Thank you so much for watching. If you enjoyed this video, give it a thumbs up. And if you want more videos like this in your YouTube feed, all about AI news, AI tools, AI tutorials, all that kind of cool stuff. Subscribe to this channel. I'll make sure that stuff keeps showing up for you. Thank you once again for tuning in. I really, really appreciate you. Nerd. Not with me today and I will see you in the next video. Bye. Bye.