HUGE AI NEWS #16 DALLE-3 , New BARD, Midjourney 3D, New Self Driving, Youtube AI and More

So one of the things that was just recently released was Dall E3. Now Dalle 3, if you don't know, is similar to Dall E2. It's essential image generation software, quite like Midjourney. But this one is a bit different. You see, OpenAI actually did something that I didn't predict they were going to do. And what they were able to do was they were able to add text. As you can see on screen it says, Larry is so cute. What makes him super duper. So essentially they've integrated this, you know, chatbot with dall e3, which is, of course, image generation software. So you can see they're able to design some stickers, they're able to talk about many different things. And I think this is probably GPT 4.5. And you might be wondering, why are you saying that this is GPT 4.5? Because previously we did talk about how ChatGPT ads it's going to get incrementally upgraded, more things are going to be released, and we know that it is going to be more multimodal than it is going to be anything else. So, of course, this is technically Dall E3, but it's also an upgrade to ChatGPT. So of course everyone's wondering, when is GPT5? Yada, yada, yada. It seems like we're getting these updates to ChatGPT that are essentially becoming slowly and surely towards GPT5. So this is a major update for ChatGPT because something like this is really cool because you can now actually talk with the text and ChatGPT can now give you an image. And of course, we do know that this isn't exactly like the previous image capabilities where you can analyze images, but in order to create these images quite like midjourney, you can. This is really, really cool. Now, we aren't sure how good this is compared to Mid Journey in terms of the realism, but we do know that in terms of stickers and various artistic things, that this is pretty, pretty good. So we don't know when this is going to be released. I think OpenAI did say that it's going to be released sometime in September, but it will be really, really interesting to see how this does work. So if you did want to see some comparisons between ChatGPT and of course, Mid Journey in terms of Dall E3 versus MidJourney, the top one that we can see here, this is of course midjourney. And the bottom one that we can see, no top one, is actually Dall E3. And of course, the bottom one right here is of course midjourney. Now this one you're supposed to have like a heart with a universe inside of it with clouds coming from the sky. And I do think that this one that they've shown us does actually beat Mid Journey in this area. But once again, I do think that what we're starting to see from certain image generations is that certain image generations are best at certain styles. I don't think there's one image generation that is best at all styles. Even in Mid Journey, you have to use v4 for more artistic and v5 for more real and version 5.2 for more realistic and I guess artistic combined together. For example, this is an example of some leaves dressed up as dancers or anthropomorphic leaves as some country folklore singers. And this is Dali, which looks pretty good. And of course this is Midjourney. But like I said before, I'm not sure which engine the person did use in midjourney, so I can't really judge that. But this is what I'm saying. Various different softwares, depending on which version is released, they're going to generate a different specific result. And then of course we have this one right here, which is a close up of a hermit crab nestled in wet sand. And the top is of course Dali3 and the bottom is of course Mid Journey. Like we said, Mid Journey does look more realistic, like insanely realistic. I would believe this is a real picture. And then this one of course also does look realistic as well. But I do want to say that this is actually a hermit crab and this is actually just a normal crab. So I mean, it's up to you which one you think is better. I'd love to know your thoughts down in the comment section below, but let's move now. This was something I did want to talk about last week, but the video was like 28 minutes, so we just didn't have time to put it in. But this is nextgpt any to any multimodal LLM. And this is generally where I think GPT 5 and all these other large language models are going to go. I know that many different people in the industry are thinking that we're going to decide one large language model that can do absolutely everything, and maybe that's going to be possible. But I like to think that what we can see here with NextGPT any to any multimodal LLM like previously, like Microsoft's Jarvis and other systems, is that this is much more realistic. So I think what we're going to have is one large language model in the middle. And then this large language model is going to call on every single other large language model for the specific request. So, for example, when it needs to generate an image, it's going to use midjourney or maybe Dall E3. When it needs text, it's going to use ChatGPT. When it needs to do audio, it's going to pull on the 11 Labs API. When it needs vision, it's going to pull on the best vision generator out there. And I think with all that combined, that's going to be more like an AGI rather than doing one AI that can do absolutely everything. I mean, I think it's going to be similar to how the human body works where you have the brain that pulls on, you know, the nose for smelling, you know, the tongue for tasting, the eyes for seeing, rather than the brain just absolutely doing everything. So I think it's going to be kind of like a body. And this was a paper. So, you know, it says, as we humans always perceive the world and communicate with people through various modalities, developing an any to any large language model capable of accepting and delivering content becomes essential to human level AI. So I think that is going to be very interesting because it can do any to any, which is even video. So I, I think that's where things are going to go and it will be interesting to see how that does work. But I think that this is where these large language model systems are going to go. And even when we just before saw how ChatGPT is already using Dall E3, I think that is exactly where we are headed. So if you're wondering what the future of AI is going to look like, this is a good, this is another thing I did want to include as well. This is called MV Dream Multiview Diffusion for 3D Generation. This is really, really good. I mean I've seen tons of different 3D model generations and this is really, really good. Like it's approaching that level of detail where you can actually use this stuff. Right now you can see Viking axe, fantasy weapon, AK blender. And this is just a text to 3D prompt. So I mean, look at this. I mean, Gandalf smiling, white hair examples. I mean this is really, really good. Like before you would see a lot of stuff as of course, like right here, this is where you're seeing every other project and all of the examples. I mean you can see how it looks. You know, this one's blurry, this one's, that one's not too bad, but these ones aren't as good. And of course you can see they've managed to fix various issues and various quality issues. And yeah, this is really, really promising because when you compare it and you see the level of detail on this one, it just goes to show that with 3D, like, I feel like every week we're getting increases. And here's the thing that someone previously said on Reddit, I'm not sure where the post is, but what they did say is that sometimes what people are looking for is one major breakthrough. And of course that does happen. But sometimes all of these smaller breakthroughs on the smaller level allow us to create that much more smoother transition into whatever it is that we might be moving towards. So of course you can see right here that 3D is getting maybe 10% better every now and again. Like 10% better every month. You know, eventually you can have something that's crazy. Remember Mid Journey? Every month it got better and better. And eventually now we have something that is pretty crazy. So this is crazy. Like Jack Sparrow wearing sunglasses. Boom. A 3D model. And that's pretty crazy for me. So, yeah, I think this is really, really cool. I think it's very, very interesting and it was something that I wanted to put out there. I'm pretty sure there is a GitHub page that you can use or a hugging face spot. But this shows promising, promising results. Because once this is perfected, guys, it's really going to change the entire. Then of course, you can see right here, tech tycoons combined with a net worth of roughly 550 billion gathered in the same room today for a Senate for a Senate forum on the future and regulation of AI from Bloomberg. So here you can see all of the people who are leaders and pioneers in the AI space. They pretty much own the decision on where this stuff is going to move. All the are discussing AI. Now, what was interesting was one of the conversations that they had. Now the conversation wasn't that great, but there was something that I did want to pick up on because I don't think enough people are paying attention to this. And I think once again, it shows that we are right now in a race to the ball. So you can see right here. According to the Washington Post, one of the 22 tech titans at that Senate meeting, Tristan Harris of the center for Humane Technology told the room that with $800 and a few hours of work, his team were able to strip meta safety controls of its open source large language model llama2, and that AI responded to prompts with instructions to develop a biological weapon. Now you have to understand that this is a problem because of course, Meta Chief Mark Zuckerberg reportedly replied that those instructions are available on the Internet. This is an example of AI doing sophisticated research. Fair enough, but like we stated, this is just the beginning. So remember, okay, this is pretty crazy, okay, because of course, as they state, DeepMind was able to develop AlphaFold, which is pretty, pretty crazy. That solved a lot of stuff that would have taken us years to solve. So the point here is that of course right now these AI systems aren't great. I mean, they're great, but they aren't great to the point where they can develop completely new biological agents. But the point is, is that if they're able to do that, you know, five, ten years from now, that is going to be crazy because if we have open source tools which anyone can access, of course this is good that anyone can access it. But that of course opens up to bad actors. Maybe someone wants to develop something that could ruin a whole town, ruin parts of the world. I mean, it's definitely something that we need to be careful of because if we don't have safeguards and we don't have regulations, then this kind of stuff is going to fall into the wrong hands. And of course, you know, they equate this to being somewhat of a nuclear bomb. And giving a nuclear bomb to everyone on the planet is a recipe for disaster because it only takes one person to set it off. And we know that with 8 billion people, there's definitely at least a few of those who are crazy enough to just do something, just to see how it goes. So it is definitely something that shows that although open source AI models are good, I don't think they're the best because the risk of people using them to, you know, create fraud and just do many different things, it's just too much. But it'll be interesting to know what you guys think. If these open source AI models should still be allowed or if you think that the risk bad actors is just so then what we have here is something that is very interesting. I'm glad that this is now starting to get more recognition. This is called autonomous driving with chain of thought, autopilot, thinking out loud in text, it says lingo. One is the most interesting work I've read in auto driving for a while, but before perception, then driving action, then after perception, textual reasoning and then action. So if you don't know what chain of thought prompting is, it's essentially where you ask a large language model a question. Okay, so for example, let's say I asked ChatGPT, what's two plus two? It might just say back four, but let's say I asked it, what's two plus two? And then I say, let's think step by step and show your reasoning and explain your reasoning before you give me your answer. Then it's going to say two plus two is four, because when you add two and then you add two, it's going to be four. And of course, it's meant for more complex questions. But then you. They're now applying this to, I guess you could say driving. So it says lingo one trains a video language model that comments on the ongoing scene. And then you can ask it to explain its decisions as to why you stopped planning and what are you going to do next. And of course, it shows you, okay, why it's made these decisions. So at the start, we can see here, it says, I'm edging due to the slow moving traffic. And then of course, as you move on, it says, I'm overtaking a vehicle that's parked on the side. So I think this is interesting because it gives us an insight as to how these large language models are making their decisions. I'm accelerating now since the road is clear. I'm remaining stationary as the lead vehicle is also stopped. And I think this might be, it might be a breakthrough. I'm not too sure, but I think Elon Musk did make a comment on this because he did talk about LLMs. But yeah, I can't find the actual tweet from Elon Musk where he does talk about LLMs. But I do think that this is going to be interesting to see if this is more successful than other decisions. Contextual reasoning, as we know, does improve a model's response by around 20 to 30% or in some cases even five times. So it will be interesting to see how. Now, what's also very, very interesting as well is that Elon Musk actually said something literally an hour ago that means he has inside information about midjourney. So he said, okay, that midjourney will be releasing something significant soon. Okay. And this was in response to a tweet where it was talked about. Dall E3, once deployed, will improve at a faster rate than midjourney. And Elon Musk says midjourney will be releasing something significant soon. So that means, of course, Elon Musk has insider information about what's going on at mid journey. And it's no surprise. I mean, if I work there. But I'm trying to think, what exactly is Mid Journ releasing? Is it going to be finally 3D because apparently that's what they're working on, or is it going to finally be the desktop browser area that they were working on? That was leaked. It was leaked and I did see it. I do have the screenshot show you because they don't want anyone to see that stuff and I'm guessing they're just waiting competition and they want to be the first debut, that kind of stuff. Now there's a tweet here that does say, according to David holes, mid journey v6 will be a bigger jump from water v5 with better image quality and text prompting. And mid journey 3D should come out in the next six months. Now, if you want to know what 3D might look like, we do have an image trailer from this account. Nick Floats. Now, I'm not sure if he made this himself. I'm pretty sure it was him because I didn't manage to find this trouse. But this does look very interesting and this is possible because the software and technology does exist to do this already. Okay. And I'm not surprised if this does actually happen with Mid Journey because like we stated before, being able to do this, I know looking at crime scenes, as we discussed before, like a snapshot of a crime scene or something like that, that would be pretty interesting for detectives or managing to do real estate or, you know, try to virtually explore and figure out where you're going to be, where you place certain things. I definitely think it's going to, you know, be that next level. Then of we have Robo Fab introducing the world's first factory for humanoid robots. And essentially this is pretty crazy because of course, as you know, humanoid robots are becoming more and more popular. But this is a factory for the, you know, humanoid robots that are basically going to be everywhere and this is what they want. So they're trying to reduce the cost of these automated robots. And essentially, what's even crazy about this as well, something I forgot to mention, is that they're going to be using the robots to actually assist in the factory in which they're building robots. So I guess you could say this is somewhat kind of exponential and very interesting because I didn't expect to see this technological advancement happening so quickly. So these robots, I'm definitely sure that these companies are going to be getting more funding and more crowdfunding because of course, investors definitely want to benefit from this huge, huge industry that is projected to have billions and Billions of dollars or be worth trillions. Will be interesting to see how this does play out then. Essentially what we had was a surprise. We had this platform, YouTube, saying that they're going to release a bunch of new AI tools and it's going to be helping creators and anyone that does want to become a content platform. So it will be interesting to see exactly how this stuff does work, how it does change the entire platform in terms of content creation, if it's actually going to be good or if it's going to be just pretty bad. So it says AI images for shorts and stuff. So what I'm going to do is I'm going to leave some of the video in here so you guys can see exactly how it looks and how it works. Because this explanation is vastly better than mine. Yeah. Are you actually rolling? We're actually. Oh, we're rolling. Okay.

Yeah, we're good to go.

Let's go.

Okay, so YouTube just announced this whole new set of AI and editing tools that are going to revolutionize the platform, making creation easier and more fun for everybody. The aim is to unlock more creativity for more creators than ever before. The most exciting part to me is that what was announced is supposedly just the beginning. Let's get into it.

Okay, first up, let's talk about dreamscreen, the new image and video generation experience experiment that's making its way to YouTube shorts. Powered by amazing AI technology, DreamScreen lets you bring your imagination to life by simply typing in ideas as text prompts. It then generates super fun images and videos that you can use to set the scene. All right, let's see this in action. I kind of don't want to leave. This is nice.

These new tools are expanding the boundaries of digital art. And that's not all. Meet YouTube Create. This is a new app that YouTube is building to make editing easier for everybody.

And it's free of charge. It includes access to thousands of royalty free tracks and sound effects, and you can automatically create captions for your video with just one tap.

Cleo Jade, are you not mind blowing?

I'm very excited about this in a genuine way.

Last but not least, and my personal favorite, there's a feature that lets you to clean up and remove any background noise.

I live in New York. That would be really helpful.

The beta for YouTube Create is available first on Android to creators in select countries right now. So go and check it out.

These announcements show that YouTube is really starting to transform the way that content is created, helping more creators make more content in more ways than ever before. They're shrinking the gap between our wildest ideas and what we can actually create. Until then, just keep making things. I'll see you on YouTube.

So then of course we had Microsoft announce your Copilot, which is the Windows 11 update. And it's going to include most, pretty much just a ton of different AI updates, including Paint, Photos, Clip Champ and more, all to your Windows PC. And of course, well, they also talked about how was that Bing is going to be adding support for the latest Dall E3 model from OpenAI and deliver more personalized answers based on your search history. And of course, just essentially a whole update that just makes everything a lot better. So it's going to be really interesting because of course Microsoft is pushing this entire AI into the entire system. Now I do think that this is going to be interesting because of course, as you know, Apple's operating software doesn't utilize any of this at all. And it seems like Apple is currently being left behind in the AI races. They haven't even spoke or even anticipated anything just yet. So we'll be interesting to see how Microsoft manages to gain ground because of course things with AI do move a lot quickly and a lot quicker than people do expect. And you can be left behind because of course Internet Explorer did actually catch Google off guard. Now of course I would show you this entire video. I probably might make a dedicated video on this. But of course you can see Bard can now connect your Google apps and services. So it says use Bard alongside your Google apps and services, easily double check its response and access features in more places. So what does this actually mean? So you know how you have the Google Drive, you have gmail, you have YouTube, you can actually check and use Google Drive. Or I mean you can actually use Bard to check your Google Drive, to check your Gmail, to check your YouTube, to check Google flights, to check pretty much all your anything. So this is like actually having a great personal assistant. And this is really, really, really cool because this is something that's more valuable than ChatGPT because of course ChatGPT is great, but I mean you have personal. One of the large problems with ChatGPT is that if I need information, I have to give it that information. And it is pretty time consuming. If you're using chat GPT to respond to email, to do work with your business, with your company, having to constantly feed ChatGPT that new information, especially after updates, is very, very time consuming. This is where Bard comes in. And it's already connected with Gmail. It can already help you with shared conversations, just so much stuff. And it's really, really cool that now they have this. Okay. And this is going to be something that, once again, like we said, is a step up. Of course, as you can see, when you go into Bard, this is exactly including Bard extensions, of course, if you click next, then you're going to see Bard meets Google Workspace, because it has all of your stuff. And of course, it's really cool. Is that Double check Bard responses. You can check how accurate Bard's responses actually are. So if you're not confident about Bard or Bard's not confident about something, you can click a button and it's going to show you how accurate that response really was. So, of course, you can see right there, it's giving you a lot more stuff. And of course, as you know, it's now, of course, you can upload images. So one of Bard's new features, I took a random picture of a car that I found on the Internet and then I literally just said, YouTube this. So if you don't know what it is, you can say, Google this, YouTube this. And then essentially it's going to give you, like, YouTube videos that are about that. So I think this is going to be really interesting to see what kinds of cars it chooses and what kinds of things are linked in here. So it's really, really interesting. And it also says YouTube video views will be stored in your YouTube history. And so, yeah, I think this is going to be used a lot more than people did expect. Then, of course, we have Amazon's Alexa voice. It's allowed to become a lot more natural and a lot more better and a lot more clearer. So this is going to be interesting because this is something that we did expect earlier on the year. We did Bedrock, which was a bunch of different foundation models to all of their different APIs and services, but this, of course, to their lead product, which is, of course, Alexa, or as many of you may know, Amazon's Astro. So take a look at this clip, because I think it showcases exactly what Amazon is doing. And of course, Amazon Alexa hasn't really had much of the spotlight since Siri and other things like Google have all. But I think whatever company manages to get a home device that is really good first off the ground, that is integrated with a large language model that sounds natural, is actually useful, probably can make jokes and stuff like that, it's definitely going to take this next wave by storm.