AI Search Innovations, Grok 2 Large & More Use Cases

This week there have been many long awaited releases like GPT4.0 fine tuning or search GPTs finally rolling out to some people. But it might not be as good as you think and I'll show you some better alternatives. Plus, the AI image and video space is seeing a literal space race to getting the best models with the best tooling possible. And for this video, we research, tested and summarized everything for you, so you can easily stay on top of all the AI news that you can actually use. All right, and I want to start off by talking about search this week because AI search is seeing some major developments and as you might have heard, OpenAI has been releasing SearchGPT to select users. But the competition is not asleep at the wheel, with Perplexity leading the pack and applications like GigaBrain that specialize on only Reddit search also adding upgrades. So let's briefly talk about this AI search. What is the status of it right now? Well, their goal is to disrupt Google search and different people have different opinions and preferences which search is better than the other? And while you might not have access to ChatGPT search, which I don't either right now, as you can see in this email that many, many people who signed up to the waitlist received searchgpts on a waitlist and they're not giving out access to anybody. You absolutely have access to some of the alternatives like Perplexity or gigabrain. And the good news is these tools are actually better than SearchGPT. And I say this confidently because various people got access to SearchGPT already and reviewed, tested it and shared their opinion with me. I'm Talking about various YouTubers or AI Advantage community members and there's a consensus. Everybody pretty much says that Perplexity is just better. It's more feature rich, the model is more advanced, there's references to the various pieces of information you receive in the text, not just in a separate tab. So you can see where this info is actually coming from and you have so many more features surrounding it. This is not a deep dive on search. I just want to point out that it's slowly but surely becoming better and better. And if you're looking for a specific answer rather than a list of links and resources, AI search is probably the way to go most of the time because it just gives you the answer rather than giving you multiple resources that probably probably have the answer. If you're looking for Reddit results, this is certainly a better way to search all of Reddit than going to Google. So if you're looking for an opinion which Reddit is really great for, you just go to Gigabrain and you look for something like Best Smartwatch and it's going to give you all the Reddit opinions compiled into one thing. And yes, this includes all the comments, which is incredible. If you want to go even wider than that, you can go to Perplexity and ask the same thing. And it's not just going to look at Reddit, but all of the Internet and process all those results through their large language model and give you a digested result with with all the references. I really think there's value to having both of these bookmarked. That's why I'm showing this to you. And if you were just like me and worried about the fact that you don't have search GPT, don't be. I honestly think they didn't roll this out more widely because they're not there yet. That the product is not good enough for primetime, where some of these competitors are becoming really, really good, even on the free plan. But just because the status quo is that OpenAI is sort of behind on search doesn't mean that is going to be the case in the future. Because they released a brand new blog post outlining one of their many news outlet partnerships. So they might just be getting started because if you look at all the news outlets that OpenAI partnered with, it's starting to get ridiculous. You have a lot of the big names in this list and their live feeds are going to be integrated with OpenAI services sooner or later. As of now they're just building these partnerships and SearchGPT is just a prototype, but expect this to get a lot better over time. And another AI search engine that released this week is actually the full model of growth Grok 2. We talked about the mini version last week, but now the full version is accessible if you have a Twitter premium subscription. So these other ones are free, this one is paid. But this does plug into the Twitter firehose. Now if you give it something like tell me today's headline, it does look at all the Twitter data and uses their quite capable Grok 2 language model to present you with the answers. The problem is there is no option to really focus this. You would have to prompt it, which is not just extra work, but you need to be aware of what kind of news you want to be looking at because by default it's going to give me news about the Nigerian government and Polish headlines. Now this just released, so I think over time it will learn about your preferences and your Twitter feed and what you interact with and then integrate that as context with the LLM. But as of now, this is just very generic and you would need to ask specifically to get specific answers. And that's why for this news and research use case, I really do prefer something like Perplexity or Gigabrain. One more note on Grok 2 is that it is really solid. Xai has become a serious competitor to some of these other companies, but it just doesn't have the features yet that many of their competitors do. Like I literally can't even set custom instructions where if I'm looking at this news use case with something like Perplexity, I could actually go in here and set my custom instructions, which I actually have done in here. So now all the searches consider my personal context and give it a news based on my preferences. By the way, if you care to learn how to do this yourself, I did many tutorials and custom instructions on this channel and if you want to go in depth and learn the whole process, that's what we teach in the AI Advantage community. Next up, we have something that a lot of people have been waiting for a While and it's GPT4.0 fine tuning. I'm not going to go too in depth on this. If you want a full tutorial on how to use it, we actually did one in one of the previous episodes of AI News you can use. It's this section right here. We'll put a link to it in the description below. But basically they have been withholding fine tuning capabilities to their largest models like GPT4 and then 4o forever and now they're finally releasing it as the competition has forced them to do so. Again, fine tuning is only relevant if you have a super specific use case and then you want to be using the API to perform that use case. Something like classification or the generation of specific code. You could show additional examples to the model of how you would like these things to be processed and then you have a brand new model that can do that and now they unlock the ability to use their most potent model instead of just a mini model as up until now. So yeah, this is a big deal for people who have been waiting for this and for all others. Let's move on to the next one which is also a simple one. But I touched on this when it released but it was only accessible to the U.S. but 11Labs has application that is now accessible worldwide to everyone and it is completely free. It is called 11 Labs Reader and I gotta say some team members actually Reported back that from our extensive test testing of various AI apps for our video, where we looked at the most useful AI apps that are freely available right now, this one has some real stickiness to it. And what I mean by that, that once you start using it, there's some real value here and you might just find yourself integrating it into your everyday life. Because what this app does in a nutshell is it uses AI voices that sounds human like to read out anything on your phone and on the Internet to you, meaning you could have articles or text messages read out to you with these AI voices. And, and as of now, 11 labs made this completely freely available, which is quite generous. It doesn't take a pro plan, credits, nothing. And the news here is that this app is now accessible worldwide for free. So go give this a try and see if maybe you could be listening to something interesting on your way to work or while going for a walk instead of listening to a podcast might be a great tool to add. And now let's move on to the next one. As you can see, various tools are getting more powerful by the week and it's becoming easier than ever to create videos with AI, especially if you find a platform that lets you take advantage of multiple tools at once. And one fantastic platform that lets you do exactly that is the sponsor of today's video, Dubdub. I'll quickly give you an overview of what Dubdub offers, but then I'll go in and show you the step by step of this particular video creation use case. So first of all, they offer lip syncing of talking avatars. You can also generate AI voices, you can clone your own voice, translate your voice, transcribes video with AI, write and even generate sound effects. But what I really want to focus on today is this voiceover feature. So if you go in here, you can create a brand new voiceover like so. And what I really love about this is that you have a lot of detailed control and fine tuning ability to the voices. With many other tools you just generate the voice and then it is what it is. Here you can adjust the speed, pitch and all of these details like pause settings, the rhythm, the emphasis on certain words, and so much more. And beyond that, you could also create multiple voiceovers where you have a dialogue of multiple characters happening. But we'll just go with a single voiceover here to demo this for you. And what you can do in here is you can select your voice actors, just go with the default as it has a lot of emotional options here.

Today is the day to remind yourself that you possess the power and strength you need to.

I'm just going to go with Jensen here and here with tone. We're going to go with newscast. I'll use the AI writing feature here with a simple prompt to write a fictional short voiceover script about cats and hats all at once becoming a cultural phenomenon. Going from simple pet owners putting hats on their cats to it becoming a worldwide sensation. I believe one day this will happen. I'll add only include the voiceover transcript, exclude include all formatting and this should give us the clean script that we can then generate with this tool. One more follow up to make it shorter should do the trick. And there we go. Now that I have the script and I selected my voice and the tone, I can simply generate it and then afterwards I could go in and change, for example, the rhythm here to be a bit longer. Let's have a listen.

In a whimsical twist of fate, a trend arose as pet owners began sharing photos of their cats in various hats.

Okay, excellent. I think you see the point here. We'll export this as a wav file and now that I have my voiceover, I can head on over to the video editing tool and create a brand new video right in here. But what we ended up doing here is actually taking the voiceover and creating a video of our own with a combination of Midjourney Runway Gen 3 and the Dub Dub voiceover that was created here. And without further ado, here it is.

In a whimsical twist of fate, a trend arose as pet owners began sharing photos of their cats in various hats, from cozy beanies to fancy top hats. This innocent pastime exploded into a global sensation, with millions joining the catsinhats movement, showcasing their stylish felines and creative headgear. The charm was infectious, attracting influencers and celebrities and prompting fashion designers to create cat accessory lines. Cat hat parades emerged worldwide and local businesses thrived with custom designs and cat themed cafes. Television networks even launched shows featuring fashionable felines. This delightful craze sparked conversations about pet care and companionship, reminding us that simple joys can unite us unexpectedly.

All right, so there it is. That's one example of how you could be using dupdub to create videos for social media or other purposes. So go ahead and try Dub Dub for free with a link in the video's description and create your first videos within minutes with the power of AI. A big thank you to Dub Dub for sponsoring this video. Now let's get back to some more AI news you can use. All right, next up, let's talk about all the innovations in the AI, video and imaging space. And let me tell you, these top companies here, they're shipping. Namely you have a new model from Runway, a cheaper alternative to their top tier Gen 3, and a new and improved model out of Lumalab's Dream Machine that is also top tier. And then on the image generation side, two of the top tier companies there also release brand new things. Namely, the midjourney web UI is finally up to par with the Discord Experience experience, even with the inpainting features. We'll have a look at that briefly. And then Ideogram, the previously state of the art text image generator, meaning it was the very best, thanks to Flux at generating Text, now has a brand new model, Ideogram. Two little side notes. The developers of Ideogram were kind enough to give us early access to this, so we got to begin our testing on it even before it released, which is wonderful. So let's quickly have a practical look at all of these tools and improvements and what we found in our research and testing process in all of these. Starting off with Runway's new model, Gen 3 Alpha Turbo. This is essentially a lightweight version of Gen 3. You get can can think of this as the equivalent of GPT Mini versus GPT4O. It's seven times faster and 50% cheaper than Gen 3. Gen 3 is quite pricey, but has the highest quality level. This is 50% cheaper. But what do the results look like? Because that's what really matters. We went in and really tested the image to video feature without a prompt attached as both Gen 3 and Gen 3 Alpha Turbo support this Right now on screen you can see various examples of these two different models. And while you look at these, let me briefly present you our findings and my take on what the difference between these two models are. Besides the obvious speed improvement, which is very noticeable, I think the 7x number does apply and it just feels super fluid to work with this new model. You just upload an image, hit generate and within seconds you have a result and you can iterate super quickly. It feels super good to work with this, but what about the quality? Because that's what really matters here. Well, first and foremost, all Gen 3 Alpha Turbo results are just a little more subtle. The main Gen 3 model is just a little crazier. It's out there. It drank one or two Red Bulls too many before generating the images. Now don't get me wrong, this can be a great thing depending on your Use case. If you really want those dramatic camera moves, Gen 3 is your friend. But Gen 3 Alpha Turbo is just a little more subtle. If it decides to zoom in, it still does that. It just doesn't zoom in so much. Now that's the movement side of things. But what about the image quality? Well, Gen 3 Alpha Turbo is generally speaking, a little bit brighter and more sharp and especially the sharpness is quite visible. I don't think it's a bad thing. It's just a different look and something to be aware of. And in the quality of the animation department. Well, because Gen 3 is a little crazier and a little more out there, it can also be more hit or miss and you can really see the difference. But out of a dozen images that we threw into both of these models, only on one of them we clearly prefer the Gen 3 over Gen 3 Alpha Turbo. Namely, it's this image from the Kingdom of Deposits, Planet of the Apes. The animation here is just way better in the full gen 3 model. But this was literally the only one where we really preferred Gen 3. Other than that, you're probably better off saving time and money with Gen 3 Alpha Turbo. And then if you're not satisfied with a result, you can always switch over and do the same thing in Gen 3. But using this as the default model just seems to make sense. And also on a similar note, LTX Studio, which we talked about before, added even more features. They're the AI video generator with a lot of amazing features, but the base model is just not very good at all. And I'm showing you this because I hope that features like this will come to something like Gen 3 or Luma Lab soon or surely LTX Studio will upgrade their video generation model over time. But having something like face motion capture or detailed character animation or character dialogues where it actually lip syncs would be a very welcome feature in these top tier models. Alright, next up we have a new version of Dream Machine 1.5 and they released a brand new guide which is quite detailed. So if you want to learn about how to use this, just go to this link in the video's description and you can learn about all these techniques in the guide that they created. But what's new with this version? Well, their main claim is that the text to video is way better and what that also implies is that their emoji to video is also way better. So that's what I went ahead and tested. I had some fun with this. Here's a few examples. This is what happens if you Give it a light bulb and this mind blown icon, it just creates a little visual. Like this could be interesting for some sort of presentation. Or maybe something like this Reddit post here that uses it as the basic background of a website. I think this is actually one of the prime use cases to use these videos. But hey, if you're doing a website like this, might as well put something proper there like this generation of a film, camera, thunder and a chicken. Wouldn't this be an incredible website? I surely think so. And it was really easy to prompt it. It's just free emojis. Little pro tip. The film camera usually adds the cinematic movement to it because if you included film as a part of your prompt, it would add a little more cinematic characteristics to the shot. So let me show you my favorite of this bunch. The prompt was a camera fire and an eagle. And you got this eagle sort of flying through a fiery sky. Insert eagle. Sound effect, you know, like okay, I don't know, I kind of butchered that sound. And interestingly, some users even reported back that the text to video is better. And I just ran some of the same Gen 3 tests that we did in here and this is really night and day compared to gen 3. This is just not even close to being usable. Same for the other attempts when I played around with the same settings. It's just not suited for this use case. Dream Machine really is best for these animation type use cases where it's just a little more subtle. But I think for things like this, Gen 3 is still best. Dream Machine is really good for these animation style use cases. Stuff that you would animate in after effects like this. And among some of these other new and improved features, I would just generally say yes, it does feel like an improvement. Not a major one, but it's certainly very welcoming considering this tool came out about a month ago, I think, and we're already getting updates like this. It really is a bit of a space race in this entire AI video generation niche and I'm here for it. Especially while people use it for random creations like this. If you haven't seen, this is a really fun one. But also on a More serious note, McDonald's is actually using this for commercials in Japan. Not kidding, this was generated with Dream Machine and they're running this as proper ads. Also, many Olympic ads were AI generated. So this is just becoming more and more present in our culture and that's just the way it is. Hey, if you're watching and enjoying this video, I would really appreciate if you hit that like button. It really helps this channel out. And with that being said, let's move on to the next piece of AI news that you can use. All right. In the image generation department, there's actually exciting news, because if you wanted to get your hands on Midjourney, but were always deterred by the paid wall, well, now is your time. They actually enabled a free trial for everybody for seven days starting Wednesday, Aug. 21, meaning it should end Aug. 28. So depending on when you watch this video, Midjourney might just be free. And you don't even have to use Discord to use it, because now their web experience is available to everybody, not just people that generated a lot of images and Discord, and they even brought some of the essential features from Discord to this. So understandably. So I went in here and right away generated a cat with a hat. You can just hit enter here. And as you can see in my browser, I'm generating midjourney level images. Just to reiterate, there might be other models that excel at something specific like the text generation, but overall, I think Midjourney is still the most robust and the most sophisticated image generation tool with an artistic quality to it that is just unmatched. The styles in here, the default generations, it's just excellent. And if you need something generated with it, well, this week you can do it for free. And what is the new feature here? Well, it's this editor that hides down here because now we can actually do inpainting. So I could do something like paint over the entire hat loosely, and then, you already guessed it, I could reprompt it and change it. This is probably the most powerful thing because AI image generation always produces varying results. And with this workflow you get precise control. For example, in here, I could keep the rest of the image of the cat and turn this into a oversized wizard hat. Maybe let's give this a little more space. Space to really go for the oversized hat here. And there we go, we submit and it will regenerate only the region that we selected here. Up here you can select the scale and the size of your brush. And there you go. Here at the top of our generation stack, it should already be creating this. And as I'm speaking, it is creating our brand new image. And it seems like we don't even need to edit to the completion of this because there you go, we have the same cat with a brand new oversized wizard hat and a few variations. And we got to do this without losing our original cut with the bow tie and the green eyes in here. If you're familiar with inpainting, you know how powerful this is. Now it's available in the web interface too. And you can use Midjourney for free for a week. Have fun creating. And we got so lucky with this happening because actually this Sunday we're running a remix competition in our community for our creative district here. This is a thing we did recently. We split it into various districts depending on interests, and we'll be getting together with various community members and remixing various images with this Bit Journey tool. I'm excited for this because this is really how our community got started. Back in March 2023, in our Discord server, we were running a mid Journey remix competition every Sunday for like five months, I think. All right, enough nostalgic thoughts for today. Let's move on to the next one. Okay, then we have Ideogram coming out with ideogram 2. If you weren't aware, Ideogram was actually the first model that got text completely right before Flux came. And with all the open capabilities of Flux, people sort of preferred that. But now Ideogram strikes back with ideogram 2. And the level of manual control they brought with this release is actually amazing. And we actually had early access to this, so we got to test this a bunch. And I'll show all of that to you in a second here. But I just want to point out the fact that if you're generating with Ideogram, the level of manual control here is amazing. You have five different models, general realistic design, 3D, and anime. And especially this design model opens up some use cases that were unthinkable a few months ago. And a year ago people were stating that, hey, AI image generation might be impressive, but it's useless because it just cannot get these images assets like text and logos, right? Well, that completely changed now with models like this, because with the ideogram 2.0 model, you cannot just generate images with a long piece of text completely, precisely, like I'm talking 100% accurate. But you also get to control things like color palettes. You can upload your custom color palettes for the generation to be exactly on brands. Then you select the design model and off you go. You have an AI image generator that actually respects your branding guidelines. So as you can see in these testing examples, you can actually create movie posters, you can create graphics reading cards, anything with text on it in a custom color palette. But beyond that, you can use all the capabilities we already know and love from image generators, like putting up text in imaginative environments like on imaginary skyscraper in London, as this example here shows. Or you just abandon all productivity and generate images of AI cats with AI hats. Beyond that, I want to point out one more thing, and that is the fact that the hyper realistic capabilities of this are also excellent. So look at this. If I prompted with something that implies realism, like a photo of a young man with curly broccoli hair and a serious look, dramatic Rembrandt lighting, blurred background of a rustic building. And I generate this in about 30 seconds. We should see the realism of this. Not bad. Although it might have misunderstood the broccoli hair. If you're not familiar, that's the typical hairstyle for Gen Z and it's sort of a meme on the Internet. Let's run this one more time without the broccoli. And there you go. Yet another model that can fake reality. It's crazy to see how far we've come in just two years. And if you're sitting here just taking it all in and feeling like you might not be able to actually use some of these tools tools because it might be too complicated. Just consider this clip that made its rounds on Twitter this week of an 8 year old girl writing code with the assistance of AI. I found this super inspiritive because it shows what these tools make possible. If you just put your presuppositions of you maybe not being as technical as is needed for some of these workflows. And just look at it realistically. These tools have become so easy to use and so helpful that just by keeping a little bit of an open mind, even an 8 year old can start writing code and creating apps. And that's the deep end of the pool. Some of these other tools we talk about are way, way easier to use. And I just wanted to let you know that absolutely everybody can do this and that does include you. All right, that's all I got for today. See you next week.