OpenAI Announce The Most IMPORTANT AI Project EVER (Project Stargate)

I think this will be the most important project of this era and as Masa said, for AGI to get built here.

So that was Sam Altman talking about what is probably going to be the most important build out for the future in terms of infrastructure for the United States. And that is of course a really important statement that I think most people aren't really understanding the gravity of. And in this video actually break down to you why that isn't an overstatement at all when you truly dive down into the reasons and that he said that. So this video is basically going to cover Project Stargate, which is of course the 100 billion or actually 500 billion dollar AI infrastructure for OpenAI in the United States. Now I have to be honest with you guys, this isn't the first time Project Stargate was mentioned. I actually spoke about Project Stargate on several occasions around eight to nine months ago in which we actually spoke about exactly what is going on. And those were very extensive videos. And, and now we have even more details that most people are really missing. So let's dive into what's happening. So you see right here that they talk about how this is essentially the AI infrastructure for OpenAI in the United States and that they're going to be deploying $100 billion immediately. Now that is a staggering amount of investment for future technologies. As I was saying, this is basically epoch defining technology. So that's why I state that this, this statement, although there is a ton of hype in AI at the moment, this is something that just isn't actually hype when you truly understand the end use cases. So now they actually talk about the fact that this infrastructure will secure American leadership in AI. And that is of course something that's really important. If you aren't aware and most people aren't currently aware, there is a silent war going on in the background. And what I'm referring to here is essentially the statement that these countries, okay, and by these countries I mean America and China, I don't want to say they're at war, but they are in a race for leadership in AI. And it's been said by many former OpenAI employees, in fact a notable one, that one actually left OpenAI due to concerns about the arrival of AGI, basically stating that whoever gets to AGI first is probably going to have godlike powers over those who don't. Now if that is something that former OpenAI employees are stating, it's clear that of course you need to secure American leadership in AI. And one way to actually do this is by making sure you have enough COMPUTE to actually run the AI systems effectively and get to superintelligence first. And later on in the video, I'll explain to you why such a massive amount of money is actually needed and why COMPUTE is largely the bottleneck that we are facing from getting increasingly more powerful AI systems. Now, of course they talk about how this is going to create hundreds, thousands of Americans jobs, generate massive economic benefit for the entire world, and this project will not only support the re industrialization of the United States, but provide a strategic capability to protect the national security of America and its allies. That is pretty true. It's quite likely that by 2030 or 2030 onwards, large parts of the economy are going to be running on AI technology. It might not seem like it right now, but many systems already currently run on very narrow neural networks that achieve remarkable superhuman performance, performance on tasks that we can't even fathom. And as time moves forward, it's quite likely as this project Stargate, this massive data center, which is essentially not just one massive data center, but an actual series of massive data centers, is going to be behind the economic growth and prosperity of America. And other countries will likely have their own. Now what's also interesting right here, Noam Brown, who is the lead at reasoning at OpenAI, he said that this is on the scale of the Apollo program, which is of course the first time humans successfully landed on the moon. And that was a massive moment in space exploration. And then of course the Manhattan Project, which is of course if you've seen Oppenheimer, you'll know exactly all about the top secret project during World War II where scientists actually worked to build the first nuclear bombs. And this was basically the world's most serious science experiment. And quite like this, this was where they wanted to create something before other countries could. Because whoever was using know the first to create that kind of technology, they would have significant powers over those who didn't. And he talks about how it is like that when measured as a fraction of the gdp. And he says that this kind of investment only happens when the science is carefully vetted and people will believe it will succeed and completely be transformative. And I agree, it is the right time. So for all of you AI skeptics out there stating that look, AI is just a passing fad, but we are in an AI bubble. That might somewhat be true, but you have to understand right now we have our 21st century project which is going to truly change everything. Not an overstatement, but like I said before, I will get on into some more points in the video. So remember how I spoke about how I made a video on this in the past? And one of the things actually spoke about in the past was the fact that the only bottleneck for a lot of these companies is, is essentially compute now, essentially. This was essentially referring to the time where they were not sure whether or not these companies would have gone on to actually go ahead with such a colossal investment. But during that time, Sam Altman actually did state that publicly, the main bottleneck holding up better AI is lack of sufficient servers to develop it. Now, you might not understand why that is the case, but let me refer back to the original research. The recent paradigm that has actually changed the and has made so much progress in terms of actual raw reasonability is of course, searching at inference time or test time compute. And this is something that allows the model to essentially think before it responds. Before. Remember with these ChatGPT models, when you input your question, it immediately responds to whatever message you have sent. And that is something that makes the model, I guess you could say, not as smart as it could be. But recently the researchers at OpenAI and many other research labs have started to figure out that. And I guess you could say they didn't really just start to figure it out. They actually figured this out in late 2022 and 2023, but they've now been applying this technique where you actually have the model think before it responds using a variety of different methods. If you want to know the full details, I've covered that in another video. It's actually really, really exciting. But, but the long story short is that they think in a variety of different ways that are remarkably efficient. And because they can think for a very long time, the more they think, the better the results get. So with these graphs, every time you've ever seen these, essentially what you are seeing is you're essentially just seeing the increase in capabilities of these systems as they think for longer. And that is something that I do think most people won't realize. Because the crazy thing about this is not the fact that these models get better is the fact that when they actually did search time inference on a variety of very difficult questions, the thing was, was that the actual accuracy wasn't going down. As they managed to apply more compute, which is essentially just more raw processing power, to searching. And as they searched for solutions for a longer period of time, the results from those tests, they consistently got better and they didn't show any signs of slowing down. The only reason currently why AI systems Okay. And this is the real point that I want you to understand so you can understand why this is going to truly advance us. In a crazy way. The only reason why AI has, you know, somewhat slowed down in a sense that, like, we haven't hit this rapid explosion of capabilities is because they literally don't have enough compute to run, or I guess you could say to let the AI think for a very long time. And that's basically the only thing that's stopping potentially super intelligent AI. Now, of course, there are other things that I think we will need to get to super intelligent AI and AGI in terms of how an AI is autonomous and how it's able to do certain things. But this test time accuracy graph was one that actually spooked a lot of researchers and was why Sam Altman was fired in the first place. Remember that whole AI debacle? That was essentially why this happened. So now that you actually know that basically the reason why AI hasn't exploded yet to really superhuman capabilities on certain benchmarks is because they simply basically don't have enough data centers to, to power the inference time. Once you actually do get these data centers built out, it's going to be really crazy because by that time it's quite likely that we are going to be more efficient in terms of our searching capabilities and we're going to have potentially 10x or even 20 times the compute at that time. Now, another thing that most people don't realize is that this is for the next six years. So this is a series of installations that companies are looking to build over the next six years. So it isn't one giant supercomputer that they are building. I know that that's what the Twitter algorithm would have you to believe, but this is essentially a series of connected data centers that are really large that is essentially going to power really advanced AI. And this is something that these companies need to do now whilst they still can, because if other nations manage to beat them to superintelligence, then it is going to be probably pretty bad for the country that doesn't have superintelligence. So this is why a lot of people do think that this state right here, 2030, is probably going to be a really important year, because by then it's quite likely that we will have these phases of supercomputers built out. Now, what's crazy as well is that if you take a look at this graph, I know it does seem a little bit confusing, but essentially it just shows us these series of models and a really hard Benchmark. So this benchmark is the benchmark where a lot of people may have considered that OpenAI has cracked AGI because this ARC AGI benchmark was considered as something that was really tough. This. Then of course, eventually we started with 01 mini getting around 7%, then 01 preview got 13%. And basically what you're seeing here is that as these results go up and up and up, as you're seeing from 25 to 31 to 32, they're just increasing the amount of compute and efficiency of these models. And when we have more data centers, I mean, is it going to be possible that we get 99% on certain benchmarks? It is quite likely that that is the case. And you can see for the highest amount of compute that they had available at the time, they, they were able to get 88% on this benchmark. So this is going to be something that is indicative of why people are very bullish now. So one of the things we actually got to see was an interesting interaction between Sam Altman and the President where the actual President of the United States asked Sam Altman, can you actually just talk about what this technology will be able to do and why we're actually building it? And he actually talks about the fact that AI will be able to cure all diseases. Now, as I state that, that statement is pretty profound. Curing all diseases is something that is pretty difficult to do considering the number of diseases and the complex nature of them, especially for some that we haven't even identified any kind of rudimentary solutions as there aren't enough cases. But with AI, you could potentially simulate the patient, simulate the reactions to them. There are just a billion different ways to solve this problem. And that is why building this technology is so impactful.

These guys can maybe share more about some of the work they're doing there. I think they'll jointly be some of the leaders about driving progress here. But I believe that as this technology progresses, we will see diseases get cured at an unprecedented rate. We will be amazed at how quickly we're curing this cancer and that one and heart disease. And what this will do for the ability of to deliver very high quality health care at the costs, but really to cure the diseases at a rapid, rapid rate, I think will be among the most important things this technology does.

Now, if you do think that statement is a bit overboard in terms of the fact that maybe, just maybe, people like Sam Altman have a reputation for being a little bit too hypey, you could state that other individuals who Also share this same view might reiterate this claim and show you guys that this is not an overstatement. Now it does seem crazy to state that AI could potentially solve all diseases, but, but every single time we are shocked by AI's capabilities and we have to understand that anything is possible. So this is a clip from Demis Asabis who actually talks about the hype in AI and actually states that look, this AI thing, it is not some kind of hype train. This is something that is genuinely game changing technology that people shouldn't pass up on. And of course there is the potential for it to cure all diseases. Now I'm not saying that it's gonna cure all diseases in one day. They're gonna deploy the AI and it's gonna do all of that. Of course there will be humans in the mixture, but you have to understand that AI is going to speed up the research capabilities tenfold and that is something that people have spoken about for a very long time now. And it's something that I think is.

Really important of kind of sort of crazy hype on both sides of this equation. There's the sort of, you know, I think it's called like dooma camp now of people thinking, you know, for sure this is going to go wrong. And then there's, there's the people on the, on the sort of Pollyanna camp of like, look, it's, it's all just another technology. We've seen this before with mobile, Internet. You know, it's going to be big like that but, but you know, we, we're very adaptable as a society and, and as humans and it's going to be, you know, nothing special. I think clearly, I think clearly that's wrong. This is, I think far bigger than the Internet or mobile or something like that. It's, I think it's epochal, defining. I've always thought that, I think it's becoming clear to more and more people. But I've thought that for, since I was a kid, which is why I worked my whole life on this. If it could be done, it would be unbelievably impactful. Of course the reason I'm doing all of this is because I think AI is going to be incredibly positive for the world. I think we, we're, I think we're within shooting distance of, you know, curing all diseases with AI, helping with climate through material science and new energy sources and other things that I think AI can invent as well as, you know, in our daily lives, just improve, improving productivity and enriching our daily lives and making, you know, mundane admin things sort of be dealt with automatically. I think those are all amazing. And that's all coming very soon.

Now, with that being said, considering we now have an epoch defining technology being built right in front of our eyes, AGI or ASI seems like it's not far away and the infrastructure buildout is underway. Are you guys excited or pessimistic? It seems like there might be a brilliant future or potentially some mass dystopia, but either way, this future of AGI and ASI is coming faster than we can all imagine.