Major AI News : OpenAI STUNNED! Metas New Humanoid Robots, New Quantum Computing Chip...And More

So one of the first stories I actually want to talk to you guys about is the fact that apparently $500 billion is nothing. You remember Stargate, that project that Sam Altman and others, including Oracle, are building out in terms of the data centers? Sam Altman recently said in an interview that apparently $500 billion for an insane level data center is apparently going to be nothing in the grand scheme of AI. And the reason that that statement is so wild is because I remember so many people were contemplating, is AI really that crazy? Is it really that powerful to the point where we need to invest 500 billion do? But apparently Sam Altman is saying, Look, $500 billion is nothing in the grand scheme of things, and probably in the future we're going to have, you know, $1 trillion worth of investment into data centers. So I don't know whether this is hype or not, because I think that is clearly, clearly a massive statement. But I think you guys need to take a look at this because it shifts my perspective on how big things are getting in terms of overall investment. Yeah, Stargate is the much bigger project. Stargate is the Stargate is a $500 billion project to build a very large training inference system. It sounds crazy big now. I bet it won't sound that big in a few years. And, you know, if we get to do this again, which I hope we do, you'll be like, you're raising $5 trillion for a cluster. What the. It'll be like, you know, you got to keep going. And if we're going to take a look at how other countries are also starting to think, wait a minute, if usa, you know, if the United States of America, if they are investing $500 billion into just one data center and a few data centers and their product, Stargate, maybe we need to start investing into some ourselves. So you can see here, the EU has launched an Invest AI initiative to mobilize $200 billion of investment into AI infrastructure. So this is going to be something that's absolutely incredible because I'm guessing the EU is finally realizing that, look, maybe we shouldn't just be focusing on bottle caps, but more so actual technological progress. So you can see right here, the Commission president said that AI will improve our healthcare, spur our research and innovation, and boost our competitiveness. We want AI to be a force for good and for growth. And we are doing this through our own European approach based on openness, cooperation, and excellent talent. But our approach still needs to be supercharged. That's why with our member states and with our partners, we will mobilize unprecedented capital through the Invest AI for European AI gigafactories. And this unique public private partnership, akin to a CERN for AI, will enable all our scientists and companies, not just the biggest, to develop the most advanced, very large models needed to make Europe an AI continent. So this is good. I remember for quite some time there was the theory that, you know, the EU is just going to be left behind. And I got to be honest, whilst the EU doesn't really have any state of the art models, I mean, Mistral is only the current company and I guess, you know, the UK with Google DeepMind. I'm not sure if you could say that that's a UK part because of course it is kind of like the UK offices are in London. But of course it's. The UK is not part of the eu. But the point I'm trying to make is that it is a good thing that we are currently looking towards more growth and improvement rather than just a complete slowdown. So this is of course really good. Now there was this interview clip which I actually dubbed with ElevenLabs, actually Heidi President Macron actually talking about the $1 billion investment. Now that brings me to today's sponsor, Marika. This isn't just another AI music generator. This is something completely different. Let me actually show you guys what makes this amazing tool so different from Suno and Yudio. Now one of their killer features is the VOC vocal selection. This is basically you can pick from different AI vocalists. If you want pop, rb, rap, you can use all of these specific characters. And the craziest thing about this is that you can actually use your own voice and then let the AI use it. Which means no more generic AI vocals. But my favorite feature is actually this. Okay, style reference. If you have a specific sound in mind, just drop a reference track in and Marika's AI will analyze everything from the genre to the rhythm, to the instruments, to the mood and and then creates music that matches your specific style. It's basically like having an AI producer that actually gets your entire vibe. Now this is something that is absolutely crazy because it's kind of like having an AI music producer for yourself. And it cuts down the production time for music by 99%. And I actually made this track that you're about to listen to in two seconds. So if we actually want to create the song, what I do like about the software is that they actually have these categories right here. So for those who are musically untalented, like myself, we can go ahead and click edm. We could actually use a template right here. Then we could go ahead and click Create. This is super useful because already has all of these presets so you can easily find out what song you really like. And of course, if you want to do it the advanced way, you can, you know, enter a song and title. So I'm just going to use random because my brain isn't working at all. And then of course, if I want a specific vocalist, I can actually pick someone like James for example, if I want to reference, I can also upload some stuff and then I can simply click Create. Sunbeams dance like golden shimmering shades Fireflies. So whether you're making music for fun or trying to launch a career, this is definitely a new tool that is worth checking out. Don't forget to click the link in the description to try out Marika yourself. Trust me, you're going to want to play with this one. Now, remember the very infamous quantum computers? Remember Jensen Huang and what he recently said? So let me give you guys a bit of a backstory. So of course, with AI pumping, many investors were looking at ways to gain outsized returns after Nvidia just completely shattered the stock market. And one of the ways that people were looking for returns was of course, through quantum computing companies. Now, crazily, Jensen Huang actually said something that was a bit bearish. He said that practical quantum computing is likely 15 to 30 years away. And he made these comments at the 2025 CES in Las Vegas. Now, he basically said that this stuff is far, far away. And I guess that this means that Nvidia probably isn't even working on quantum computer computers. Because recently we actually had two other major companies talk about quantum computers and basically state that they are actually five to 10 years away. Take a look at what Sundar Pichai said at the World government summit in 2025 on quantum computing. And this is really, really interesting because it goes to show that whilst yes, some companies are like, look, this thing might be further and further away, other companies and if you remember, Google actually made a breakthrough with Willow, which is of course, you know, their state of the art quantum computing breakthrough. And they're talking about the fact that, you know, they're still working on some stuff and they can see this happening.

Rather, rather soon, the progress in quantum is palpably exciting. We've been focused on it for a long time. I think we have the most advanced state of the art effort anywhere in the world. The latest breakthrough we had, we were able to perform a computation in five minutes, that would have taken one of the world's fastest supercomputers over 10 billion years. It's longer than our universe has been around. But more importantly, it shows that we can tackle error correction in a way that we can build large scale fault tolerant quantum computers. I would say to me, the quantum moment reminds me of where AI was in the 2010s, you know, when we were working on Google Brain and you know, and the early part progress. So I would say, you know, that's what it feels to me. In a five to ten year time frame, I think you will start practically useful quantum computers.

And of course Microsoft actually shocked everyone with their new chip called Major One. I don't know if it's Major One or Major One Majorana. Honestly, my pronunciation is terrible. But they basically said that that chip is going to enable the creation of quantum computers to solve meaningful industrial scale problems in years and not decades. Guys, they are saying that quantum computers are going to be coming in years and not decades. And we know why quantum computers are so insane, because of course they can exist in two states. They're just so much more powerful than a regular computer. Which means that it's going to be pretty possible that we're going to have some next level AI stuff going on in the coming couple of years. So this is going to be something that is absolutely insane that I honestly can't wait to see what occurs with this kind of chip. I will be making a video on the May Jorana one very, very soon, but for now we will have to move on to the next story. Now, if you're wondering about other AI timelines, there was also this piece of information that was stated. Apparently time is short and we must accelerate our actions to match accelerating AI progress, possibly by 2026 or 2027, and almost no later than 2030 apparently. And this is Dario Amadei, by the way. The capabilities of AI systems will be best thought of as akin to an entirely new state populated by highly intelligent people appearing on the global stage. A country of geniuses in a data center with the profound economic, societal and security implications that would bring. There are potentially greater economic, scientific and humanitarian opportunities than for any previous technology in human history, but also serious risks to be managed. So for those of you that didn't really understood what I just said there, you have to understand that by 2026 or 2027, this is a shockingly, shockingly small amount of time. You have to understand that this is literally just two years away and two years in AI is an extraordinary amount of time considering how quickly these developments take place. And apparently it's looking like it's going to be an entirely new state, but it's going to be no later than 2030. So currently we're in 2025 and that means in the next five years it's quite likely that there will be some kind of, you know, gold tier level AI that is just going to be changing the world every single day across a variety of different industries. You can see here it says a country of geniuses in a data center with of course just incredible society and security implications. So this is going to be something that I think is absolutely game changing. And I think it's really time for everyone to start realizing that, look, a lot of hype is in the AI industry. Like, of course there are a lot of hype, there are a lot of different things being said, but you have to remember that a lot of this stuff is true. If we can of course have intelligence, that scale of intelligence in a data center duplicated across millions of different servers, which we are going to be living in a completely different society very, very quickly. So you have to understand that society is going to be moving very, very quickly and we have to prepare for set changes. Now, of course, not wanting to be left behind, you can see that Meta is planning a major investment into AI powered humanoid robots and this company is aiming to be the engine powering the humanoid robot market. So this is super interesting because Meta, you know, we've known that for the largest time they've been focusing on open source models. But. But I'm guessing that maybe considering the fact that China has somewhat surpassed them in deep seat V3 and of course R1, and I do remember the fact that there was also a blog post to the fact that Meta was scrambling after the release of deep seek v1 or v3, and essentially maybe now they're like you. Look, we can't completely dominate the large language model space, but what we can do is potentially we can power humanoid robotics. So what we will do is we will try to invest there. You can see it says Meta plans to work on its own humanoid robot hardware with initial focus on household charges. Its bigger ambition is to make the underlying AI sensors and software for robots that will be manufactured and sold by a range of companies, said the people, because of course the initiative hasn't been announced. So Meta are eventually going to announce this. But essentially what they want to build is the underlying robotic AI sensors and software. So Meta is basically trying to be the backbone for the AI revolution. Like they want their, you know, chips and stuff to be used. That way, of course, they can make money for years and years to come. And apparently they've started to discuss their plan with companies like Unitree Robotics and Figure AI. And at least initially, it doesn't plan to build a Meta branded robot, something that could directly rav as Tesla's Optimus, but it might consider doing so in the future. So robotics is of course speeding up. This is something that they want to talk about. And of course it kind of does make sense to actually be the people that's supplying the hardware rather than actually building the hardware, because building the hardware is very, very hard. But supplying it is of course, much easier. It's basically the thing where in the gold rush revolution, the people who got rich were the ones selling shovels, not the one who were digging for gold. So I guess Meta is trying to basically sell shovels here, which is of course maybe actuators AI software. And basically if they of course manage to figure out some really cool AI software, then other companies can go out and build that stuff. Now, in terms of robotics, I was supposed to have seen this. I didn't actually get to see this, which is unfortunate at the time because I was ill at the time. But this is a video of the very agile Unitree robot. So the Unitree robot, for those of you who don't know, and I think this one is so impressive, is because this robot is just extremely agile. And ever since they've made this robot, I've seen constant, constant updates in terms of the software that has enabled this robot to look incredibly realistic. Like when this robot was first announced, the way that, that it walked, the way that it moved, didn't seem really impressive. But after several software updates, this thing is honestly a remarkable platform. Like, it's so, like, you know, smooth in terms of how it moves, in terms of how agile it is. And since we are at like the very, very early stages of where robotics is, if the robot can literally dance and do this, think about what robots 10, 20 years from now are going to be able to do. If they can move, you know, quickly, autonomously and just do a variety of different charts, the world is going to look completely different. And remember, guys, this robot right here, the one that you see on your screen, that is dancing. And this video was posted by the CEO of Unitree. And the reason he actually posted this video, because the video that they posted before was so good that many people thought it was cgi. You have to understand that like, what are the robots going to look like in the future? So this is a case where we have, you know, robots looking so realistic, so effective, moving so agile that, you know, people don't even believe that these robots are real anymore. I think this entire thing is absolutely incredible. And it just goes to show you the pace of AI development. Honestly, in 10 years, I definitely think we're going to be getting some household robots because there have been several breakthroughs that I think are just going to be shared throughout the industry that are going to make this entire industry a lot more effective. Now, of course, if we actually look at more research, there was also Beam Dojo, a new reinforcement learning framework developed by Chinese researchers that actually teaches humanoid robots to walk on tricky terrains like stepping blocks. It basically uses a two stage training process, a specialized reward system and lidar mapping for precise foot placement. So this is something that is absolutely incredible. I mean, the robot you can see right here, this is something that like, you know, most people probably would struggle to do. I don't know about you guys, but I cannot walk backwards onto, you know, six different wooden pylons. That's not something that I can do. But once again, like I said, these robots are going to get continual updates, continual software updates that is just going to make them more and more effective over time. And that's exactly what we're seeing right here. It's truly impressive to see what these systems can do in just the early stages. And I can't seem to say the early stages because I don't think there has been the crazy, crazy breakthrough moment for robotics just yet. But this is definitely something that is remarkably impressive. Now of course we do have Thinking Machines. If you do remember Mira Moratti who was the former, I think, COO for OpenAI, she actually left to start her own company and today they actually announced what they're doing. So it says today we're excited to announce Thinking Machines Lab. Thinking Machines is an AI research and product company. And it says we are scientists, engineers and builders behind some of the most widely used products and libraries including ChatGPT, PyTorch and Mistral. And our mission is to make AI intelligence work for you by building a future where everyone has access to the knowledge and tools to make AI serve their unique needs. We are committed to open science through publication and code releases and while focusing on human AI collaboration that serves diverse domains. And it says our approach embraces co design of research and products, enable learning from real world deployment and rapid iteration. It says that this work requires three Core foundations, state of the art model intelligence, high quality infrastructure and advanced multimodal capabilities. And it says, we are committed to building models at the frontier capabilities to deliver on this promise. I honestly don't know how this company is going to outdo what OpenAI have done. Maybe Miramorati has some Kiki engineers that are going to be able to build different things, but I do think that they're probably going to tackle this AI problem from a unique angle that most people haven't seen yet. Honestly, this is kind of vague, but essentially they are just building LLMs, multimodal AI that interacts across different data types and of course, AI safety research with real world testing. I honestly think that when we actually start to look at things, I think that as the AI industry grows on, as years grows on, there's definitely going to be a variety of different AI industries that just spawn in. So companies like this, I think they're probably going to take their lion's share of a market that we probably don't even know exists yet, because those AI markets are definitely, definitely coming. And now, of course, I probably should have included this earlier on in the video, but if we're actually talking about AI investment, remember how Sam Altman said $500 billion is nothing? And if, of course, we're also going to talk about former OpenAI employees and, and what they're doing after their departure. Ilya Sutskova has actually raised money again at a $30 billion valuation. So Ilya Sutskever, once again, I think he raised another billion dollars at a $30 billion valuation, which is absolutely insane. But then again, if this company can manage to get to Superintelligence before anyone else, that valuation is actually going to be probably on the low side, because if you have super intelligence, there's literally nothing that you wouldn't be able to do, provided, of course, you know, it's within the laws of physics. But then again, when you have advanced technology, it looks like it breaks all of those laws of physics. So this is something that is extremely bullish. Now, of course, another thing that we do have here is we do have Yann Lecun. Now, Yann Lecun, of course, he works at Meta and he's working on what he would call advanced machine intelligence. And recently in this AI Action Summit, he actually spoke about, whilst, yes, superintelligence might seem like it's on the horizon, we're actually very far away from that and we're missing something really, really big. So maybe everyone's just, you know, got their head in the AI hype hole. But apparently we're missing something really, really big. So take a look at what he says here.

We're missing something really big because, you know, never mind trying to reproduce human intelligence, we can't even reproduce cat intelligence or rat intelligence, let's let alone dog intelligence. They can do amazing feats. They understand the physical world. Any house cat can plan very highly complex actions and they have causal models of the world. Some of them know how to open doors and taps and things of that type. And in humans, a 10 year old can clear up the dinner table and fill up the dishwasher without learning zero shot. The first time you ask a 10 year old to do it, yes, she will do it. Any 17 year old can learn to drive a car in 20 hours of practice. But we still don't have robots that can act like a cat, we don't have domestic robots that can clear up the dinner table, and we don't have level 5 self driving cars, despite the fact that we have hundreds of thousands, if not millions of hours of supervised training data. Okay, so that tells you we're missing something really big.

And something that surprised me a lot was of course, the Grok 3. If you aren't familiar with Grok 3, that is Elon Musk AI chatbot. And honestly, the reason that this is so insane is because Grok 3 is now the state of the art in terms of AI model performance. And what's crazy about this, okay, what is crazy about this, okay, was that Sam Altman recently said something to the tone of Elon Musk actually needs to just make better models than us and stop trying to, you know, interfere with what we're doing in a legal way. And you can see right here that, you know, Elon Musk, they've actually managed to make their early Grok 3 model the number one on the LMSYS arena, which is basically a blind test between AI models and apparently prefer Grok 3, which is absolutely crazy. And there was a little bit debate on Twitter stating that they did like a consensus at 64, which basically just means that they fiddled with the benchmarks a little bit. But there was just a back and forth saying that OpenAI did the same thing. The point is that I actually do believe that Grok 3 is probably the state of the art and I do trust the LMSYS benchmark the most because it is humans that use it and humans don't know which model they're using. And whichever model they do prefer, it's completely blind. So I do think that now this is probably the best model that exists, but it's also super interesting to see just how quickly they managed to catch up to state of the art. And if they can catch up this quickly, I think that there is the potential for XAI to even surpass what OpenAI is doing. So considering they got a small nimble team, there is the potential for them to Even surpass where OpenAI are in terms of sheer raw model performance, which is pretty crazy. And they also even released something called Deep Search, which is of course basically like a kind of research tool that OpenAI have as well. So this is crazy. Like things are moving incredibly fast in the world of AI. It's super, super competitive. So many people are competing for. I don't want to say the same customers because they're all doing different things, but it's just super interesting to see how all of these companies are managing to ramp out products of such high quality in such a short amount of time. And I do apologize for the glitches. They're not sure what's going on, but I can assure you Deep Search does work well. And if we're looking at other companies as well that are doing some things, we also have Claude's thinking from Anthropic. So you can see right here that this button is here. I'm not sure if this is just like an early feature that some people now have access to. I don't have access to this. Maybe you can check if you have access to claw 3.5 sonnet thinking pro clock 3.5 sonic. I. I do wonder how smart that model is going to be, but I would estimate that it probably will be out this Thursday because there's been a slew of AI announcements and things like that that are going to be out very, very shortly. So that is going to be something, of course, to look out for because we all know Anthropic do not skimp when it comes to model performance. And considering we just had Grok3 just do absolutely insane. How are Anthropic going to respond to this? So that's going to be super interesting for me to see where their model performance.