MASSIVE AI NEWS #13 Everything Just CHANGED! New AI Robots, AI Reads Minds, , New AI Text To Video

So welcome back to another week in AI content. And this week was actually quite interesting because a lot of unsurprised tools managed to come out. So let's take a look. So Starting with Audio LDM2, it's learning holistic audio generation with self supervised pre training, or for the people who can't be bothered with technical jargon, it's basically text to audio, text to other stuff as well. So in this video that they talk about, they essentially put a text input in, but you're able to get an entire output of different kind of outputs. And if that doesn't make sense to you, I'm going to play part of the video. But essentially what it is is that you're getting an output that is not just text to audio. So on the page you can see it's a section one, text prompted audio generation. So you can see that this is essentially sound effects. So what you can get is you can get sound effects, you can also get music generation, and you can also get speech generation. That means this entire model, it essentially gives you three different entire outputs, which is really, really useful because a lot of the reasons why people use different models is because they need to fine tune these models and specify them for certain purposes. But with this model it seems that they've managed to combine state of the art text performance in text to audio, text to music, and text to speak generation, which is pretty insane. So I think this is going to be one of the models that is largely used because as you know, people who are working with text to audio, text to music and text to speech, usually what does happen is you might be working on a project in which you might need text to music, in which you might need text to speech, and of course you might need text to sound effects. So I could presume a tool like this. The reason I think this is so game changing is because something like this where you have the sound effects and in a moment I will play some of them so you can see that this isn't actually awful or just something where it's just a demo. This is something that is really, really good. And I did use this tool live and it did show me something that was really cool as well. So right here, essentially what we do have is Texas sound effects. I'm going to play some of these Texas sound effects now so you can hear them and you can see just how good this. So now that you've seen all of the examples, the reason I think audio LM2 is game changing is because something like the Adobe Suite Imagine I'm editing a video and, and I want them to be able to add, you know, a background track. I can add background track. Then I want to be able to add the speaker, I can add the speaker. Then I want to add some sound effects. I can simply add that all in one simple tool. There is a video that they did release, but it is actually really, really cool. So leave a comment down below letting me know if you do try it out and if you do like the results, because my personal results were actually pretty good. Then we had something game changing where we had Google's DeepMind CEO. So this next algorithm will eclipse ChatGPT. So essentially the company that built AlphaGo, the company that built many different game changing breakthroughs in artificial intelligence, is working on a new system called Gemini that stated that it will tap into techniques that helped AlphaGo to be a Go champion in 2016 and it's going to eclipse ChatGPT. Now, essentially what this means is that they're probably going to release a tool that's somehow more advanced than ChatGPT or GPT4. And by them stating that it's going to eclipse ChatGPT, I don't doubt it. Because if you've ever seen AlphaGo and you've ever watched our video in which we talked about it, you'll understand just how crazy it was. So essentially the reason this is so crazy is because, number one, it's from Google and Google's DeepMind have done insane things with AI before, if you've ever been around the AI space. Number two is the fact that they're pumping billions of dollars into this and they're using a new method of training and making this AI thing. So I think this is where we're going to get some real innovation that's going to lead to that next step. And this article is really good and I would recommend you all read this because it is very, very interesting. But of course, as you know, the article does dive into the grave risk because when you're leveling up an AI and you're making that AI essentially into something that we've never seen before, it definitely could have some risks because I think they didn't really understand what they were saying when they said that they're going to make sure that this large language model can plan and think long term. I mean, if someone said that to me a year ago, I would have said, well that isn't that super dangerous? Isn't that what we're trying to avoid? But now I guess they're running forward with it. So we will see. And they did say that it might release sometime in 2024 or 2025. So that will be interesting to see how those developments do come into play and how they're going to be working on the system. Then we had lumalabs talk about how they're essentially trying to accelerate the rate of advancement and grow their community with their sponsor, Nerf Studio, and the Berkeley AI Research. So essentially what they're doing is I think they're open sourcing this tool and they're essentially trying to get a lot more people invested in this tool in order to make this kind of. So, yeah, essentially what they're trying to do is make more people invest into this kind of software and essentially make this kinds of content with this tool. Because I do believe that this kind of content and these kinds of tools are really, really underrated. And what they're about to do for technology is really, really impressive. What they also did announce around two to three days ago was they essentially released this app called flythroughs, an app that shows off your space with AI generated cinematic videos that look like professional drone captures. So essentially all you have to do is do your lidar capture, whichever however you do it. And. And then essentially what it does is it does the fly through itself. So it's actually really, really cool. I think you can also do some of the fly throughs yourself and then it kind of makes it into a really cool montage. Now, one thing that I haven't seen anyone talk about when we look at Luma Labs, AI is going to be the application in terms of crime scenes. Because one thing that I was looking at is the other day, you know, tons of crime channels are popping up on YouTube where you can watch various documentaries or crime scenes and stuff and how detectives are analyzing this. But do you not think it would be interesting if maybe someone, for example, uses Luma Labs, you know, takes a snapshot of the crime scene at that specific time and then, you know, people around the world, detectives, can analyze the crime scene and maybe you can, you know, have that snapshot where you can look at that and then think, okay, maybe this happened, maybe that happened. Because sometimes crime scenes get contaminated, sometimes things can happen. And I think this would be definitely a really good application. I haven't really seen too many other than like real estate, of course, when you're trying to view a house or cinematic montages. But of course this doesn't involve moving objects, which is unfortunate, but it's still really, really cool. So Luma Labs, you know, you can use it on your phone, you can do it for your house, many different things you want to do. So yeah, really, really cool. And I would definitely check this out. Then of course we have Nvidia finally releasing their model Neuro Angelo. So essentially, if you don't know what Neuro Angelo was, it was where you could just take any photo and convert it into a 3D model. So it's really, really accurate. Like really, really accurate. So it's, it's just shockingly accurate compared to the previous ones in terms of how accurate these 3D models eventually are. So it's something that will be interesting because if you don't understand the AI cycle, essentially what happens is an AI tool gets released or announced and then essentially what happens after that is they release a research paper, usually with the tool, and then what happens after that is they open source it. And when they open source it allows people to build on top of that AI tool. Which then means that once someone is able to build on top of that AI tool, then that's where you get all of these AI tools, all these AI apps. For example, when Llama was open source, we got tons of these smaller, large language models. But now that this is open source, we're likely to get some kind of 3D recreation stuff where people can fine tune it, maybe they can make it better, maybe they can make user interfaces, they can make products. So it will be interesting. And I do think that's how many businesses are going to be making money by just editing open source tools and making them into a full fledged product. So definitely seems like a gold rush opportunity. But at the same time, I can't wait to use this because there are some pictures that I do want as 3D scenes and it will be interesting to see how this is used in many different industries. And I think, I know this might sound weird, but I do think something like this will be used maybe in like planets or something like that. Like for example, let's say the Mars rover is trying to analyze another planet and it can't venture to a certain area. It could take a snapshot of that terrain because it can't get there. It could, and then it could analyze it in 3D and then maybe you could get more data about that specific area. So that would be really, really cool. So yeah, there's many different explanations. We did release an entire video on that, so you might want to check that out in the description below. But yeah, it's really, really cool. So I'm excited For this, I'm excited to see what people build on it. Then of course, we had another AI tool release one image to high quality 3D object generation using both 2D and 3D division priors. Now this one is super interesting because I've seen tons and tons and tons of different AI tools come out. Even one from OpenAI, if you even believe that. And this 3D tool surpasses even that. And one thing I want people to understand about AI, and I really do want to get a debate going in the comment section about this because it is something that I do think is important. I think that any company that solely focuses on solving one AI problem is going to be vastly better than any AI company that focuses on solving various problems. Because time and time again what we see is you, you know, many of these apps that you're seeing, many of these other examples that you're seeing, you know, Shap E, I think this one was from OpenAI and some of these other ones are from other companies that do other AI stuff as well. But the problem is that these aren't that great. Like in converting this image to a 3D model, it isn't that great. But this company that's focused on doing this, which you can see, magic123, they've worked extensively on this. You can see just how good this is. Look at that teddy bear, look at this statue, look at that horse, look at this thing. Like those actually look like game ready. I would argue that some of them are game ready. Maybe not the horse because the horse's face does look a bit weird. But literally, you know, generated objects just from an image. This is some pretty, pretty impressive stuff. So they use a different strategy. I'm not gonna get too much into the details, but I do think that this kind of stuff is really interesting because it shows us that when companies just simply focus on, you know, whatever it is they want to do and they really just, you know, decide to make these high quality meshes, it shows us that specialization is, is going to be something that I do think stands out. So if there's any air companies watching this, you know, I highly doubt it. But I would say specialization is, is where it's going to set you far apart. Because if there's a company that can literally do image to 3D with perfection, they're going to be worth a lot because image to text, image is huge. And this is very, very close. Like this teddy bear, this statue, this horse, this mug, rat, whatever it is right here, that's really, really close. So I think which whichever company manages to solve quickly, they're going to do really, really well. So there's that. But then we had something called Music to Image, which was really, really interesting because it's not something that you would expect or even has a demand, but I do think that some creatives will use it. So essentially you can select an audio and then you can generate an image from that music. And I'm not sure what it's based on too much, but it is very, very interesting to see the kinds of AI tools that are constantly released because Music to Image wasn't something that I even thought about. But. But I'm guessing just based on what I think they do is maybe they have something that describes the music track, then converts that description into a text prompt that is then led into something like stable diffusion, and then you get an image output from that. So, yeah, I think this is something that could work because, I mean, sometimes people make music and they want to transform it into other mediums of art. And I guess sometimes also people, when they do have music, they want to see what it looks like, you know, what does this sound like, what does this feel like? If it's aggressive, if it's, you know, having a certain theme, I don't know, it's definitely interesting. Although it's not crazy, you know, it's not game changing. I still think there's something like this is interesting and it's definitely something you can check out, you can use because there's a hugging face space. So, yeah, let me know what you think about this as well. Okay, so this is probably the biggest thing that I've ever seen in AI and I think this is going to transform online content and I think there's literally going to be a law made about this because what we are looking at is the very next evolution in how news presenters, how YouTubers, how, you know, media companies, including some people like myself, may actually choose to deploy their content. So what you're seeing on screen is not a real person. This isn't a real person at all. This person didn't get up and record this content. They didn't record the voiceover. This is 100% artificial intelligence generated content. And it kind of blew my mind because I was like, what on earth is going on here? But I did mess around with the software a year ago and it was shockingly good, but now it's indistinguishable from reality. So essentially this is an AI avatar. You can clone yourself and make videos and make it, say Whatever you want. So take a look at this video because I know you've just been looking at some guy and you haven't even heard anything, but it's pretty insane. Okay? And I'm gonna talk about some of the tricks that they did use, because I think I did pick up on some of these tricks because it was really cool. But take a look at this video because it's still, still, still till this day, ever since I've seen this, it's still been playing in my mind every single day. Because I'm thinking, how on earth is this even possible? But enough rambling, take a look. In the early days of computer programming, a significant historical anecdote is known as Grace Hopper and the bug. In 1947 at Harvard University, Rear Admiral Grace Hopper was working on the Mach II computer. One day the system stopped working and technicians discovered a moth trapped between the contacts of a relay, causing the malfunction. Hopper then coined the term debugging to describe the process of fixing computer errors. She even pasted the moth in her logbook, which is now on display at the Smithsonian National Museum of American History. Now that you've seen that, understand that this YouTuber, Jesse Welland, someone that you might know about if You've been on YouTube for maybe over five years, but this guy was like a prankster back in the day. It was really popular. He actually also did this with his own avatar, which he cloned, which just goes to show us how good this really is. Now, I think his video is not as good, but. But I do think that the movements, some of the inflections, some of the pauses do make it a lot more realistic. And I think when we start to get to that area where these AI companies start to realize that some of the mistakes is actually what makes a lot of the stuff human. That's when I think it's going to be a game over moment. Because a lot of the times what we have is AI trying to be perfect. And usually perfections isn't great because life isn't perfect. Like for example, if you were to draw a car, there's usually rough things on the car that. It's not usually a car that's just fresh out of the showstream. They're usually mistakes. It's usually marks. Usually when people speak, sometimes they go, mmm, sometimes I have little off pauses. But with speech, it's usually all perfect, clear and concise. Which is why I think that this kind of tool, they're already getting some of that nuance perfect already, which is a little Bit scary. So it will be interesting to see how far this tool does go, because I think once this gets released worldwide, because currently it's in beta or alpha, it's gonna be absolutely insane. So take a look out for that one because that definitely, definitely did scare me. Also, Anthropic did actually improve Claude 2, which isn't really big news, but I think it's important to note that Claude 2 is actually coming up on the level of ChatGPT in terms of usability, just because, as you know, currently ChatGPT is facing a crisis where many people are stating that it isn't as good as its counterparts in terms of the fact that the quality has declined. So I think it's a PR problem, because ChatGPT is most certainly better than every other AI tool out there. But because it's declined in usability, many users are feeling frustrated that they initially paid for something, but it's gone down in quality. I mean, imagine buying a phone and it got worse as you used it. I mean, although that does happen, it's not what you'd expect from software. You wouldn't expect your software to get worse. Okay. It's not like a hardware issue. They've just decided to do that, I think, because of pricing issues. So it will be interesting to see if many people move over to Claude 2, because also you have a thousand context window, you know, 100,000 context window, which means you can input PDFs, large bodies of text. And I know that's something that you can't do with ChatGPT. And another thing that's annoying with ChatGPT as well is that not only do you have to pay for it, it's not currently worldwide. A lot of the features they release, it's just for us at the moment. So that is, you know, quite frustrating for many users. So it'll be interesting to see what happens over the future. But I do think that Claude 2 is actually up there now, especially with the context window. It definitely sets itself apart. Also, Nvidia unveils a more powerful AI chip coming next year. And honestly, guys, when it comes to Nvidia, I just have to say that this company, a lot of people are saying they're overvalued. You know, it's a stock market bubble. Like, I think the price is actually justified. And I'm not a stock advisor or anything like that, but I've looked at Nvidia time and time again and they just keep doing everything right. Like, honestly, they just do everything right. Like, they are really, really really in the game, when it comes to AI, like they have the graphics card, they have the software, they have the robots, they have the specialization, they have the team, they have the. They literally have so much stuff that I wouldn't be betting on any other company when it comes to AI, they are really great. They do provide us with a lot of information that we supply you guys with. So I would say that when they do release this GH200 super chip, it's. It's gonna be incredible. I mean, it's gonna be absolutely incredible. So I didn't think that they were going to be able to continually increase the quality of their chips, but somehow they do it. And if you're thinking, okay, it's a chip, like what is that gonna change? Well, think about like this, okay? I mean, early this year or later on this year, they announced that they could train entire models on one chip when previously that wasn't possible. So if we get to a stage where you can essentially train entire models on one chip in just a couple of days, think about how quick we're going to be getting updates to AI models. Think about how quick someone's going to be able to train their own model. And think about how future models, the models like ChatGPT6, GPT8, GPT10, Google's Gemini, how much easier it's going to be for them in order to train those models, because the compute's time is going to be that much quicker. So I think that's something that's really underrated and that is something that, you know, people aren't realizing is going to increase so much the speed at which this kind of stuff comes out. So keep an eye for this because these are the announcements that they don't break the headlines, but they really do move everything quickly because people are like, oh my God, all these tools coming out so quickly. Well, thanks to Nvidia, it's thanks to all these fast chips that are able to get everything processed so quickly. So all these cloud companies, man, they're going to be blowing up. But yeah, Nvidia once again doing it crazy. Then of course we had something called Suno AI. So essentially it's text to music. And we have seen loads of different text to music, but this is the first tool that is actually out there and released. So thank God for that because, you know, so many tools out there, you know, in these videos that we talk about, we talk about them, but the tools don't get released. Like I might say, okay, they released this text image, for example, earlier in the video we talked about, you know, text to 3D or whatever, but more images, 3D. But the problem is is that we can't actually use these tools. So what useful is a tool if we can't use it? And this is something that you can use, it is an alpha and you can sign up. It's like Discord, it's like Mid Journey. So for those of you who are creators out there who want uncopyrighted soundtracks, definitely going to shake up the music industry, the royalty free soundtrack industry. I do want to know what your thoughts on is this because although it doesn't sound absolutely insane, I do think this is going to change a lot because once we get full three minute soundtracks that sound perfect like midjourney's text image, things are about to change. So with that being said, let's move on to the next one. Then we have playht which is a low key player in the AI space and they released something called their conversational model which actually challenges 11 labs. Everyone knows what 11 labs is. It's that privately funded company that you can use to make generated AI voices. And it sounds absolutely indistinguishable from human voices. But playht, they really did just change the absolute game with this conversational stuff. I mean on their website you can see AI text to speech, AI voice cloning, voice generation, API. But what they released with the, with the conversational stuff, it's crazy. Okay, now don't take my word for it, okay, don't take my word for it. Look at this video because I think this is going to be the second company that most people do use and just take a look because it's crazy. All right? It's just crazy. I mean, yeah, I'm just gonna play the video and you let me know if you can tell which one is AI because I can't. Okay. The comments can't on other videos and I don't know. So you let me know. I honestly can't believe what I'm seeing right now. I honestly can't believe what I'm seeing right now. I honestly can't believe what I'm seeing right now. I honestly can't believe what I'm seeing right now. Hello, Play support speaking. Hey, yes. Yeah, so I've been on your basic plan for like a few months and I think I want to upgrade to the professional plan. The one that's $99 per month. Yeah, sure. Glad you're liking it enough to, you know, consider an upgrade. Let me just pull up your details. Real quick, can you shoot me your account, email or like your phone number? Sure. It's 650-451-2218. Alright, just give me a sec. Okay. There you are. So what are you actually looking for in the upgrade? Any specific features or stuff that you've got your eye on? Yeah, well, I've been running out of storage a bit. Then we had something that was once again shocking as AI is. I mean, every week it seems like there's something that just, you know, just blows my mind. And I did think that these people were gonna be safe, but I really don't know which job is gonna be safe other than, you know, entertainers. Because this, what you're seeing on screen now is an AI robot that cleans toilets. And it isn't just like a basic, like button you press and then, you know, it essentially just, you know, like wipes the toilet. But this is something that literally goes inside a toilet, opens the door, shuts the door, Hoovers, cleans it. And I think this is crazy because everybody knows that nobody really wants to clean toilets. I mean, especially in some scenarios, you know, like, for example, a bathroom and a bar, you know, those ones are particularly not nice. Or, you know, like public bathrooms. And I think you have to understand how capitalism works, where if a company can get, you know, robots made cheaply enough to where they don't have to pay people to clean bathrooms, I think it's going to be a job that's just completely done by a robot. I mean, this robot literally is opening the door, you know, cleaning it and just moving on to the next. It's just crazy. Like, I didn't think we'd see that this year. I didn't think we would. And maybe it's been something that has been going on, but I mean, even grabs like some tissue and just like, it's crazy. I'm not gonna lie, guys, this is something that I really didn't see coming just yet. But let me know what you think about this, because it was, it was just, I mean, crazy to see how good it was. You know, it wasn't tripping over. It's got like a hole, you know, based on it. I mean, it's crazy. I mean, you're watching the video, you're seeing what I'm seeing, okay? And maybe I'm overreacting. Like, dude, it's just a robot, you know, trying to clean the toilet. Like, it's not revolutionary by any means, but I think the accuracy in which it does it, the fact that it can Open doors. I think that part is interesting. So I think this shows us once robots get really good at something, they're really going to be a part of the workforce. But I think what's gonna be crazy about this, because I did do some research and they have been working on these robots for a long time, is when these robots become super cheap. So when economies of scale start to the point where these robots become super cheap, and I think that's way in your everyday life. You're going to just be walking on the street and then you're going to interact with the robot. So that's what will be interesting for the most part. But let me know what you think about this, because I do think that this is quite interesting. And I want to say one more thing before we move on from this. Most people are going to say, dude, it's just a robot that can clean toilets. Just remember that these things get better every year. So now they can just clean robots perfectly. What are they going to be able to do next? Just remember that continual progression and evolution. Then we have this one where AI just reconstructed a Pink Floyd song from Brain Activity. And I don't know, guys, I don't even think we're living in reality anymore because I feel like this stuff is happening. And these videos might get 10 to 15k views or whatever, But a large majority of people don't realize how advanced this technology is increasing and just how much money is going into this software. So I really do think it's going to be interesting to see how much the world is going to change in the next, you know, 10 to 20 years just based on AI alone. So something like this, you know, is really crazy. So, yeah, I mean, I mean, I don't know what to say other than that, like, this is something that is basically science fiction if you told it to someone 10 years ago. So, yeah, crazy.