AI News: Nvidia Stuns CES While OpenAI Teases AGI

There was so much AI news this week, mostly because this week was CES out in Las Vegas. This is like the biggest tech event every single year. And so many companies hold out to make their big announcements during the week of CES because they know the world is kind of tuning in. This year was no different, but there was really just one company that overshadowed all the rest with the amount and the sort of level of announcements they made this year. And that company is Nvidia. Now, before I get into the news that Nvidia announced during ces, I do need to disclaim that Nvidia did sponsor part of my trip out to ces. This video isn't sponsored by Nvidia, but Nvidia was one of the companies that paid me to be out there. So with that being said, let's just dive right into it. Starting with the fact that Nvidia finally announced their 50 series GPUs. Gamers, video editors and AI enthusiasts like myself have all really been waiting for this new series of graphics cards to get released. And when they finally showed them off, they did did not disappoint. Because these new GPUs are built on the Nvidia Blackwell architecture, they can run generative AI models up to two times faster than the previous generation of GPUs. We can see they perform pretty much two times as good as the current gen GPUs that are out there right now in all of the most graphically intensive video games, as well as generative AI, like using Flux to generate AI images. Now these new GPUs, they come in four models. There's the RTX 5070, the 5070 Ti, the the 5080 and the 5090. And according to Jensen's keynote, he claims that the RTX 5070, the least expensive of the new 50 series models, the ones that go for $549, run just as well as the current generation 4090s, which go for about $1,600. So right now, today, if you were to buy a GPU, the best you can pretty much buy for gaming AI and all of those sorts of things are the RTX 4090s. That's a $1,600 GPU. When these RTX 5 come out, it's a 550 GPU that performs pretty close to as well as the 4090s. Now there are some people online that have complained saying that it's not super comparable because these new 50 series gpus actually use generative pixels and they generate some extra frames to make it seem like it's getting a much faster frame rate and to fill in the gaps of some missing pixels. But I'd say for 99% of the world, they're going to see it as performing just as good as the 4090s. So pretty soon, when these graphics cards are actually released to the public for about $550, you're going to be able to run generative AI models on your computer, play all of the top of the line video games, and edit 4K videos really, really smoothly. And that is what I'm super excited about. Pretty much all of that I'm excited about. I'm a gamer, I'm a video editor, and I'm an AI nerd. So all of this is great news to me. So I'm excited to get my hands on these 50 series. I'm personally going to spring for the 90, but I'm excited about this whole range of GPUs. This was just the very first announcement that Jensen made at the CES keynote. Now, I'm not even going to get into everything he announced. I'm just going to talk about the things that I found the most interesting and that I think the majority of people watching this video will find the most interesting. The next thing that I was really, really excited about is that he showed off this little computer. You can see it in his hand here in this picture. And it's called Project Digits. It's a personal AI supercomputer. Now, technically you can use this as just a computer that you use normally, but it's really designed to be like a cloud computer that's sitting on your desk. So in the same way you would use something like CLAUDE or CHAT GPT and you enter your prompt and it's sending it to OpenAI servers or anthropic servers and then giving you the response back. This little box that you can just put on your desktop and wire into your main computer will be like that server that you're sending it to. And these use the Grace Blackwell hardware, which is pretty much the same type of hardware that those big companies are running their AI models. So it will be similar to having your own personal AI supercomputer sitting on your desk that you can wire in from another computer and you can use it without Internet because all of the inference and training and whatever you're doing on it would all just run on device without even needing to be connected to the web. These little super computers are going to run for about three grand and they're expected to be available starting in May. So I'm super excited for these to come out because once these come out, I'm probably also going to try to get my hands on one of these and we'll put it through its motions in future videos and really find out what it's capable of and what we can do. Having our own AI supercomputer at home and yes, the cost $3,000, that's pretty dang high for most people. But this is just the beginning. If what we've seen out of the GPUs like the 50 series, little AI supercomputers are only going to get cheaper and cheaper and cheaper over time. And I can see a world where in the future everybody is doing all of their own AI on device at home without needing to send it to the big companies like the microso us and the Googles and the anthropics of the world. It'll all just be handled on your own home supercomputer. And as prices come down, that becomes more and more viable for pretty much anybody to be able to do. During the keynote, Invideo also talked about their agentic AI blueprints that use their NIM microservices. Now a Nim, short for Nvidia AI, microservices are little pre built tools that you can use. Like let's say you need to use Vision AI for something you're building, there's a NIM for that, you need a language AI, there's a NIM for that, you need a to speech, there's a NIM for that. These NIMS are little pre built AI tools. Now the blueprints are when you start stringing multiple of these things together to complete more agentic workflows. So some examples they give here Daily's Voice Agent Blueprint uses Nvidia's Riva Automatic Speech Recognition and Text to Speech Nim microservice along with the Llama 33 7D NIM. This helps it achieve real time conversational AI. So it's taking multiple NIMs and and stringing them together into pre built workflows. So you can just grab one of these blueprints and use them in whatever software or tool you're building and you don't have to figure out how to connect all these dots on your own. Another example of a blueprint is Nvidia's PDF to Podcast AI blueprint. You've probably at this point played with or at least heard of Notebook LM where you can give it documents like PDFs and text files and URLs and then it will actually turn it into an audio podcast that you can listen to. That's another blueprint that Nvidia is going to make available so you can have your own personal built notebook LM type tool using Nvidia's little workflow that they pre build for you. Another blueprint that they talked about that I'm really, really excited about is a blueprint for AI agents that can analyze video. Now in this example here, they're talking about surveillance footage. Like imagine a company that has 100 cameras on their workflow watching all the production that's happening. Well, nobody's probably watching that footage unless an issue comes up and then they need to go hunt through the footage to go and look at what happened with that issue. Well, with this NIM that Nvidia's building, it's like having somebody constantly watching and analyzing all of the video footage as it's coming through. And if something pops up, it's more proactive. It will actually alert the company that, hey, there's something maybe you should look at here. This article goes on to talk about how these NIMs can also be used for things like sports, where you can use the NIM to analyze videos of like your football throw, your golf swing or your baseball swing, and then get feedback from the AI on how to improve it. One thing I'm really looking, looking forward to on this, and I would love to build a blueprint for this, if it doesn't already exist from Nvidia by the time these are available, is one where I can upload a ton of B roll footage. Maybe it's footage that doesn't have any audio, it's just clips from me walking around ces and it can analyze all of the footage and help me uncover the coolest footage to share in future videos. A blueprint like that would be amazing for my own workflow. And blueprints that they're building are going to enable things like that, which is really, really exciting for somebody who makes videos like me. Nvidia also announced some new large language models. These models include their Nano model, their Supermodel and their Ultra model. The Nano model is a smaller, more optimized model that's designed to run fast, but maybe it's not the smartest model in the world. It'll do very, very simple tasks. And then you've got the supermodel, which is kind of the middle ground model. This is like Your, let's say GPT4O that most people use when they use chat GPT and then you have your Ultra model, which is their most powerful, smartest model. These are all built on Meta's Llama models, but they're optimized and designed to work with this new Nvidia hardware that's coming out. And the last little thing that I want to talk about from the Nvidia keynote was when they went into their Cosmos World foundation models. But what makes this really exciting is that this was trained on 20 million hours of real world video and it was designed to create like a digital twin of real world environments so that you can actually train things like robots and self driving cars inside of digital twins of a real environment instead of in the real world. So you're not actually risking, you know, real human lives or real machines on the open real world. You can actually simulate what it will be like inside of these virtual environments, train them in these virtual environments and then once you're confident that these machines are ready for the real world, because they've produced good results inside of these digital worlds, they can now then deploy robots in these autonomous vehicles in the real world. So this is really, really exciting. It's going to help really accelerate the training of new robots and the training of autonomous vehicles and things like that. It's also going to help with production planning for manufacturing companies. They can sort of simulate real world manufacturing scenarios and optimized for flows for their factories inside of virtual environments. And then once they feel they have the optimal flow for their company, they can then build it out into a real factory. This also gives these machines the potential to simulate any possible scenario you can imagine. I mean some of the scenarios they talked about where autonomous vehicles driving through the snow and an animal runs out in front of it, well, they can simulate the car running into that scenario and they can simulate what would happen based on different actions that the car would take to see how it plays out. And then once a whole bun of simulations have been run, now the car knows the best outcome based on all of these simulations. Really, really, really cool technology. Also probably fairly oversimplified by me, but really exciting stuff. If you haven't watched Jensen CES Keynote, it is actually pretty mind blowing stuff. I'm super excited for where Nvidia is going and was just really excited to be a part of this whole experience and be a part of this Nvidia keynote in person at ces. I can't wait until it's all sort of deployed into the real for us to be able to get our hands on all of it and use it. All right, moving on beyond Nvidia Now I was talking about world simulation models. It appears that Google DeepMind is also planning on building their own world simulation models. Tim Brooks, who works in videogen and world sim at DeepMind, is looking for hires to help build this stuff. He says DeepMind has ambitious plans to make massive generative models that simulate the world. So right now Nvidia seems to be leading the way in this area. But it also looks like Google DeepMind is going to be going down this path as well pretty soon. And since we're talking about Google, they are currently rolling out a new feature called Daily Listen. This is sort of like Notebook lm which creates a podcast for you, but instead of you going and feeding it documents of what you want it to create a podcast for, it actually is just watching your search history and it knows what kind of things you look at and it knows your interest because it's Google, they kind of know everything about you if you're in their ecosystem. And it can create these Notebook LM style podcasts, but just based on your interest and it can create a new one for you daily so you can kind of get a rundown of what it thinks you'll be interested in. Now this is apparently rolling out right now. When I try to go to it, it says it's not available for me right now, so I don't know how quickly it's rolling out, but supposedly it's rolling out now. Once it's rolled out, we'll probably play around with it on a future live stream, hopefully by search history and whatever else it's looking at. Is this team enough to show off on a podcast? We'll find out. Fingers crossed. Microsoft had a little update this week. They made their 54 models open source on Hugging Face. Says here Microsoft is releasing its 5.4model as a fully open source project with downloadable weights on Hugging Face. If you're curious about 5 4, it scored over 80% in challenging benchmarks like math and MGSM. If you're curious what MGSM is, it's the multilingual grade school math benchmark, a benchmark of grade school math problems. So from an open source standpoint, it's probably one of the better models that's available for doing math. This article also claims it's got impressive results in human matter of time before. It's available on other platforms like Grok, Groq, Grok and within tools like Perplexity and Mind Studio and some of these other tools that are more model Agnostic. Since we're talking about large language models, X launched their Grok as an iPhone app this week. I went ahead and downloaded it on my phone here and it's pretty bare bones. You can have it draw pictures right now. Apparently this is what it thinks I look like when I use the suggested prompt of draw me. You've got the option of chatting with Grok 2 or a temporary chat, and that's pretty much about it. There's not much else on this app. It's pretty bare bones. But I think that Grok believes if they really want to compete with the chatgpts and clods of the world, they need to have an app that's separate from X, because there's some people that just refuse to use the X app. Maybe they'll use the Grok platform knowing that it's a lot more uncensored than a lot of the other platforms that are out there now. The other really big news that's been circulating this week is this blog post from Sam Altman about how he believes that they know how to get to AGI and they're now shifting their focus to asi. AGI being Artificial General Intelligence, ASI being Artificial Superintelligence. Now, a lot of this blog post is just his reflections over the past year. The ups and downs of running OpenAI, getting kicked out, coming back, that whole story. But if we scroll down towards the end here, this is where it gets really juicy. He says. Down here, we are now confident we know how to build AGI as we have traditionally understood it. We believe that in 2025 we may see the first AI agents join the workforce and materially change the output of companies. Now, again, AGI is Artificial General Intelligence. It's like an AI that is pretty much as smart as the smartest. Theoretically, it can answer any questions on any topic and answer them accurately. Moving down the post a little bit, he says we are beginning to turn our aim beyond that to super intelligence in the true sense of the word. We love our current product, but we are here for the glorious future. With super intelligence, we can do anything else. Super intelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own and in turn massively increase abundance and prosperity. So when they're referring to ax, SI or super intelligence, they're talking about an AI that is smarter than the smartest humans. Like that is moving into a point where the AIs are so smart that we may not even understand their logic or reasoning around what they're doing because their intelligence has exceeded our own intelligence and that is what OpenAI is now starting to work towards. Another quick but interesting note about OpenAI, Sam Altman tweeted on X Insane thing. We are currently losing money on OpenAI Pro subscriptions. People use it much more than we expected. So the $200 a month plan is actually losing them money because OpenAI's 01 Pro model is actually being used way more than he expected. And it's quite expensive to run apparently. Stability AI had some news this week. They introduced stable point aware 3D which is real time editing and complete object structure generation, also as SPAR3D. Now this is a tool where you can give it an input image and it turns it into a 3D mesh. So just one image from one angle and it figures out the whole 3D object. They've actually made the code available over on Hugging Face here, but there's also Hugging Face spaces where it's available to test and use as well. Here's an example of a toy car. You can see the original image up here on the left. It removed the background, turned it into like a point cloud here and then turned that point cloud into a 3D object that we can move around and look at from every single angle. Now, it's not perfect. You can tell the angle that the image was originally taken on looks a lot better than the opposite angle because the opposite angle is just trying to guess the underneath sides looking a little funky. But we're getting really close to the point where you could just upload one single image and then have AI convert that into a 3D image. And I'm excited about that because I want to upload a picture of myself, turn it into a 3D version of myself, and then use it in Unreal Engine, use it in Blender, make videos with it, and do all sorts of fun stuff like that. Pretty impressive. I'm excited to play around with this more, but mostly I'm excited to see it improve. This is the worst is ever going to get. Adobe also released something pretty cool this past week and there wasn't even really any big announcements about it. They kind of did it under the radar. They launched this thing called Trans Pixar AI which can actually generate videos, but the videos have transparent backgrounds. I'm probably going to use the heck out of this because I make videos and it would be cool to put little animations and overlays on my videos. We could see some examples here like a jet fighter and a robot that turns its head towards the screen, an explosion, a squirrel eating a nut. All of these have transparent backgrounds on them and they were all generated with AI and you don't have that little like green glow behind them that you get with like green screen videos. Now this one is available on GitHub here, but it's not really open source. They give you the right to use it, but for non commercial research purposes only. And they've also loaded it onto Hugging Face, so there is a space here where you can play with it. I just gave it the prompt. Fireworks explode in bright, beautiful colors. This only generates 13 frames, so it's a really short 2 second video. But you can see my little firework explosion here that it generated. Here's the alpha version that it generated. And I can Download these as MP4s and use them in my videos. Now, there were some other interesting announcements that came out of CES this week. I'm not going to go too deep into my experience from CES in this video. I'm actually going to do that on my Monday livestream at 11am Pacific. I'll share all of my experience, show some of the B roll that I shot, and talk more about CES during that live stream. But these are some of the bigger news announcements that had people talking at ces, one of them being Samsung and their smart fridge. It wasn't really just a smart fridge. They showed off smart washer and dryer, smart appliances, smart light bulbs, smart pretty much everything. And the idea for Samsung was they want everything to sort of be interconnected and preempt your needs. So, for example, maybe you're wearing Samsung's Galaxy Ring and it recognizes that you're really, really hot while you're sleeping. Well, it could tell that to your smart thermostat and turn up your AC for you automatically without you needing to do anything. They also showed off this smart fridge, which I just mentioned a second ago. And what makes it smart is that it actually can see inside of your fridge, notice when you're out of milk or out of eggs or something like that, and then automatically order them for you from Instacart to arrive the same day. Now, if you watch my live stream, where I used an autonomous agent to go order toilet paper for me on Amazon, racked up almost $700 worth of toilet paper before I shut it down. This kind of idea really, really worries me. Like, how is it going to handle if there's just items in front of an item and it thinks you're out of that item because there's something in front of it. What happens when you stack multiple things on top of each other? I have a lot of questions about this technology, but I do see the future that they're envisioning. I see that future where these tools sort of preempt your needs. So you don't even need to think about what you're ordering. It just knows you need it and orders it. I just think we're still kind of far away from me being able to trust technology like this, personally. This company, Withings, also showed off their smart mirror. You can see it's a mirror that actually gives you all sorts of health data. Just by looking into the mirror, you can see that it's got like heart rate analysis and blood pressure analysis, and it syncs up with your smart ring or smart watch and gives you your activity levels and the steps you made during the day. And it also supposedly will give you like motivational messages when you're looking into the mirror. Mirror. I did see this one in person. I didn't test it myself. There was kind of a line to do it and I didn't have the time to wait in the line because of everything I was jumping around and doing during ces. I did see it in person though, and I did see other people using it. It looked pretty interesting. I don't really care about a mirror that gives me motivational messages, but a mirror that can daily give you all sorts of stats could be really handy and potentially even save lives. Right? Like if it preempts some sort of health issue and you go and see a doctor because the mirror preempted it and, and let you know it was about to happen, that could be a game changer for some people. This company Haliday, showed off their new AI smart glasses. What makes these ones different is that instead of having a little screen with a heads up display inside of the glasses themselves, they actually have a laser that beams the images straight into your eye. So when I was testing out the Astra glasses from Google, there's actually a little tiny square inside of the glass that you could kind of. You could barely see it. You had to look at the right angle. But there's a little thing there that, that's where the heads up display is, that puts images in front of your face. This one you could see in one of these images right here. It's got a little like laser thing. And this laser literally beams the light right into your eyes so nobody else around you can see what you're seeing. It is shooting the text or images or directions or whatever. It's showing you directly onto your eyeball. Now, my first thought is, is, is this safe for our eyes? I mean, we've been constantly told don't shine lights in your eyes and don't shine lasers into your eyes and things like this. And these are glasses that literally project stuff into your eyes. I think it's really cool technology. I think it's exciting, but I want a little bit more long term confirmation that this is actually healthy for our eyes. But I do find it really fascinating. I think this one was pretty cool and helpful as well. VLC Player. It's a media player for both Mac and PC. It's the media player I use to play videos on my computer. They're adding real time AI subtitling that happens on your device. There's no cloud or anything like that. So you can download videos, watch them using vlc, and it will automatically subtitle them, even if there's not a subtitle file with that video. That seems really helpful to me. And then I want to end this video by showing off something else that I came across this week that I found just absolutely fascinating. There's a new film coming out about Vladimir Putin and the whole movie basically, basically deep faked Putin. So I'm gonna go ahead and mute this video because I know that it's very likely it's copyrighted music in the background, but if I fast forward it a little bit here, we can actually see what he looks like in this movie. Now, this is an actor playing Putin. And then they used AI technology to replace the face using Putin's actual face. They actually uploaded tons of images of Putin's face, drained it into an image of his face, and then use that inside of these videos videos to get his real face onto this actor. That seems crazy to me. If the creator of this movie mysteriously disappears, well, we won't go there. But from what I understand, the movie's not a favorable movie towards Putin. So I'm not sure how excited he's gonna be about this actually happening. I just find it very fascinating and also a little bit ballsy that somebody actually decided to do this. And that's what I got for you today. I actually recorded an entire AI news video while I was at ces and because of a mistake that I made, the entire audio on that video got screwed up. I had a setting set wrong, and I tried to fix it over and over again, and I could not get the audio to work properly and to sync up and to sound good. So I end up having to scrap the entire AI news video that I made from my hotel room in CES and had to wait till I got home home to record the news video. That's why this video is coming out a little bit later than normal. But my plan is on my Monday live stream at 11am Pacific, I'll share my experience of CES. I'll give my thoughts on it, I'll share some of the video footage that I shot there and then there's also some really cool AI tools that have come out this week that I'm excited to play with on the live stream. So if you're free at 11am Pacific on Monday, I'll be going live. Come hang out, come chat, ask me anything. I'll share my CES experience and we'll play with some tools live and talk more about the Nvidia announcements and things like that. It should be a really exciting fun time. And if you like videos like this, you want to stay up to date on the latest news. Get really cool AI tutorials on how to build cool workflows for things like productivity and making cool AI videos and you want to stay looped in on the latest AI tools. Make sure you like this video and subscribe to this channel and I will keep you looped in and make sure more videos like this show up in your YouTube feed feed. If you haven't already, check out FutureTools IO. This is the site where I curate all of the coolest AI tools I come across. Keep the AI news up to date constantly and there's a free newsletter. You join this free newsletter, I'll hook you up with the AI Income Database, a free database of ways to make side hustle income using AI. It's totally free and you'll get the coolest news, the coolest tools delivered directly to your inbox. It's twice a week and we will keep you looped in there as well. Find it all over for free at Future Tools IO. Thank you so much for tuning in, hanging out with me and nerding out about all of the coolest and latest AI news. Hopefully I'll see you in the next one. Maybe I'll even see you on the live stream on Monday. Have a good one. I'll see you later. Bye bye.