MAJOR AI News : 500 Billion Investment, China RACES Ahead, OpenAI Breaks Out, Agents Join Workforce

so let's actually take a look at the crazy AI stories that have happened this week now notably one of the first AI stories that was absolutely incredible was the Stargate so if you don't know I'm pretty sure it's been everywhere even outside of the AI ecosphere and essentially what we have here is the fact that Stargate is the project that is going to be basically for AGI artificial superintelligence and I think

the reason that this is getting so much social media attention is because of the sheer size of the investment so you can see right here that it is $500 billion over the next 4 years to build out AI infrastructure for open aai and I think the craziest thing about this is not even the fact that they're putting $500 billion the fact that you know this thing that they say that this $500 billion is going to just be for openi in the United States like I don't know how samman and open AI have managed to do that but securing a $500 billion AI INF structure with the government that is an insane thing to be able to do and I'm pretty sure the you know other companies out there are probably I wouldn't say Furious but they're probably thinking wow opening eye has some serious power and that was a huge move for them so $500 billion is basically similar to the size of investment that the US government made in the Manhattan Project if we are talking as a percentage of GDP so this is something that is super fascinating to me personally because I truly enjoy enjoyed watching the Oppenheimer movie and I think this is going to be something that's quite similar to that so this is you know the thing of our time like this is a project that the United States are going to be building it's going to be built over the next four years and I think what this should show everyone even for the AI Skeptics is that you know people don't invest $500 billion into technology that isn't going to change the way we live so I think $500 billion over the next 4 years which is going to be essentially $125 billion per year invested into AI infrastructure I mean I can't even imagine how much money that is an incredible amount of

money and the thing about all of this that I'm starting to think about is that if they're investing $125 billion per year that is going to be a scale of compute so ridiculous most people can't even fathom it like opening eye on their recent round I know they raise several billion dollars but if every year you have $125 billion of course that's going to be paying for certain other things but the point I'm trying to state is that like the data center expansion is going to be absolutely incredible like there is going to be so much computer available that I think 100% now this investment pretty much confirmed that pretty much confirms that by 2030 the world is going to be in a completely different place because a lot of the times the experiments that you know these Labs want to do and the kind of things that they do want to research are simply hampered by the fact that they don't have enough compute and that is of course also due to the fact that they don't have enough for the investment in the infrastructure the AI industry essentially they're suffering from the problem of you know AI technologies have increased so rapidly but it takes a lot longer time to actually build out data centers so we're kind of trying to play catch up in that part so that is pretty

crazy now also want to include Le you could say a little bit of drama but Elon Musk actually commented on this and he said they actually don't have the money in response to the Stargate project and he says soft bank has well under $10 billion secured and I have that on good authority and then Sam mman says wrong as you surely know you can come and visit the first site on way this is great for the country and I realize what isn't always optimal for your companies but in your new role I hope you'll mostly put America first now of course Sam mman and Elon Musk are you know going at it currently they both hate each other Elon Musk is suing Sam mman if you're not familiar with the whole debacle Elon Musk is the one that initially created open Ai and he created it to be an open source company he's alleging that Sam Alman basically took the company and essentially just turned it into a for-profit company and is now going to get Equity from the company making him a multi- multi-billionaire and Elon Musk is like that's not the entire reason I started the company what is going on here and not to mention that Elon mus is actually a little bit behind on AI and I'm wondering if they are going to catch up so it it's super interesting of course this is a tweet about the actual money I do think that you know open ey are going to get that investment I do think that when we think about the kind of Investments That America going to be making in terms of

their military defenses I'm not sure what America's military budget is but I know is absolutely insane and if America are essentially the country that's known for spending a ridiculous amount on their military budget You could argue that considering the fact that artificial superintelligence will probably be the most powerful military tool in existence then they're probably going to be spending every single nickel and dime on artificial intelligence so that is why I'm actually thinking that they are going to be investing this amount and probably even more off the books because they're probably going to have some secret government facilities anyways so 500 billion although it seems wild this is basically securing ing the future of the country so for me personally this genuinely does make sense but I think we're in for a wild ride over the next 10 years and of course it's all going to be documented on this channel now another thing that happened in the AI industry that didn't actually surprise me I'm not going to talk about this for too long but um you know sunno has actually been sued by gima the German organization representing music record holders and essentially what I am wondering is I am wondering if there is going to be a real chasm in the sense that you know these generative a companies may have to pay Hefty fees to those that can actually reproduce some of their work that you know exists within the training data so if you don't know how these generative AI systems actually work a lot of the times what tends to happen is you know they train on a lot of data and some of that data is publicly available data

which they don't have explicit permission to use because of that sometimes they essentially you know in a gray area they'll just download the content train the AI model on it and because of that sometimes in some cases you can actually get very similar content to the training data so you could create a song and then it sounds very similar to something that someone else already created publicly and this is something that happened with mid Journey there were images that you know certain artists had created but individuals were able to recreate those images by using very specific prompts so for me I think this is going to be super interesting I've already spoke about this in a video a very long time ago but I do think that the outcomes of these decisions are really going to shape the future of AI and to be honest with you guys I'm like 50 on this on part of me I do want these tools to be available because they are really useful for the entire industry but at the same time I do feel bad for creators whose work was you know just plagiarized and then people are out there creating stuff that is quite similar to them so for me I think that this is going to be super interesting with as to how these you know courts do rule because essentially they did steal but at the same time they're not really reproducing that content directly it's something that sounds somewhat similar so now that brings me to today's sponsor so most people in AI might actually want to build a professional website without technical knowledge and there's actually

a new AI powered solution that can create your dream website within minute so this is where hosting as AI Builder comes in while other platforms overwhelm you with complex customization hosting as AI Builder focuses on Simplicity and speed this actually perfect for beginners who want professional looking websites without the Steep learning curve and luckily for you guys they actually have a 85% off sale where you can actually grab the deal for just $299 a month or even $199 a month you can get started creating your website so let me actually show you guys how easy this is then I'm going to say let's create a website then let's click next then let me go on the hosting website builder so let's click next so let me enter a brad name let me actually call this one the wedding photographers and then what I will do as I'll say we are a photography agency in London focused on taking spectacular wedding photos please use the color red on the website and then immediately we're greeted with this amazing website here so this was literally done in about 10 seconds maybe so this is something that's really crazy you can see right here that this is in incredible detail and it's able to have everything that you would need if you were just getting started with your business now what's crazy about this is that you actually get access to powerful AI tools on the left hand side and you can see that there are many different things that we can actually use one of them being the AI logo maker so of course if this is a new business what I'm going to do is I'm going to create a business for my wedding photography thing so let's say a photography studio and then I want this to be luxury or elegant so I'm going to go ahead and create this and with these AI tools we can go ahead and use this logo so for example I could use this one to start my business and I could put it as you know the wedding for photographers and then for the font I think I'd just want this one that one looks really cool this is something that you can easily use especially if you are a beginner now if you're ready to build your website don't forget to use the discount code AI grid for a massive discount on any plan that is above 12 months when you sign up through the link in the description below now we also actually had openai release their operator agent and this is of course a research preview unfortunately the AI

operator agent is $200 a month that is at you know open's most expensive tier of course it's an AI agent that can use its own browser to perform tasks for you so this is going to be something that can you know do a little bit of research it can book you flowers you know in the demo they showcased someone being able to do their groceries someone was able to find really specific information about flowers and stuff like that so it was something that I think is really important in order to move systems to more autonomy and essentially I do believe that this is something that you know is going to change rapidly over the next couple of months I really am not sure how they're going to solve some of the you know issues because they actually spoke about the fact that security is a very big one this is essentially like you know the internet Theory where you're going to have millions and millions of AI agents running around the web but with this one I think the really important thing here is to essentially have something that works and is reliable because they talk about how this AI agent sometimes it forgets the names it might send the money to the wrong person sometimes it does get stuck in certain loops and I mean of course you know with any AI systems what we've seen before if we are taking the current trajectory of how AI systems have been in the past it's quite likely that in 2 years this will look really rudimentary maybe there's going to be some different scaling laws I'm not entirely sure what AI agent scaling laws are going to look like that is something that I haven't really looked into I did see a tweet from the Google AI Studio lead Logan Kilpatrick saying that multi- a agents are going to factor into this but it genuinely isn't something I've seen talking about so I do want to know how they're going to scale past these problems I don't think it's going to be a major problem considering all they need to do is realistically be better than human but I think one thing that they will need to have to Ace if they want companies to actually use these AI agents in practice is of course the reliability as that is a main factor for many individuals I don't think anyone could really use an AI agent if they knew there was a 5 to 3% chance of it doing something that could be potentially catastrophic like imagine it sends an email imagine it clicks a links and downloads a virus like I don't think anyone would truly want to you know maybe even put that agent on and then go about their day knowing that a ticking time B Could Happen of course it probably wouldn't happen all the time but for an individual like myself I really wouldn't even want to chant it I'd have to watch it every single move and in that case it would kind of defeat the purpose because it's slower than me and it's not able to do things I am going to say though that if you potentially are disabled this is something that could allow you to access those devices and do things that you wouldn't otherwise be able to do which is of course something that the majority of people using these Technologies tend to forget so next we actually get to

take a look at a paper/ video of something called UI are so this is a nextg software agent designed to interact with computer screens just like how a human would so unlike all the systems that need you know a lot of Specific Instructions uias only needs to use these screenshots in order to figure out what to do now this is built to be adaptable and autonomous making it super effective so it's actually great at anzing the GUI so the elements like the buttons the fields the menus basically all of that stuff it can do directly from the screenshots and then it process these elements with pretty much Precision making sense of even you know really complex layouts which is really important for certain websites and then you know it basically then simplifies these actions like dragging crawling or clicking into a unified systems into a unified system basically across you know platforms like desktop mobile or web and you know it actually combines quick reflex like decisions you know the decisions called system one thinking with slower step-by-step planning which is of course system two thinking so it actually makes it really good at handling both easy tasks and complex multi-step workflows now this one can actually also you know look at what went wrong and UI Tas can improve itself over time and it cause and actually uses a process called iterative learning where it trains on past errors to avoid repeating them now this system actually performs popular AI systems like Claude and GPT 4 on benchmarks that do test things like reasoning and task execution and it's particularly strong in tasks requiring real-time adaptability like navigating apps interacting websites and even handling complex computer environments now I think this you know is something that's pretty it's a pretty big deal because this kind of software could you know automate many repetitive tasks saving time and effort for a lot of people I mean like imagine you have an AI that can book your flight organize your files navigate complicated software do additional research just do the very you know boring tasks on your computer for you this is going to be something that is truly autonomous now essentially what we have here is the base llm for UI tars is actually you know quen 2vl which is part of you know a series of Advanced Vision models and the paper basically you know like the paper behind all of this stuff it basically you know talks about how it was built on a 7B and a 72b model which mean that which means that it leverages a highly capable multimod model that processes all of this effectively what's great about this also is that this is also open source it's hosted in the GitHub on bite dancers repository and I think once again considering how I spoke about earlier in the video opening I did release their operator agent I think the landscape is definitely heating up as we're looking at many companies releasing very similar

tools and price is decreasing rapidly in terms of the open source tools are actually catching up to closed Source I think this is good for everyone though developers you know they can experiment with these tools if they want and individuals who just like to use shiny products like myself can use those products and you know the general public can actually get into AI in a very simple easy way now if we're talking about AI agents I was supposed to actually talk about this but I think this might be on opening eyes next road map so essentially this is where samman has actually spoke about you know his goal to automate software engineering and release a new powerful product that handles coding problems involving multiple steps aiming to replicate the work of an experienced programmer according to a person who heard him speak about the product and such a product is commonly known as an agent a type of system that aln said could recently join the work for so one of the things that open AI they haven't really spoken about that much is the fact that they are working on what I believe to be an automated software engineer I've seen a few articles where they've spoken about how they may have an internal one but this is something that I do believe is going to be gamechanging and I think it's probably why open AI haven't really focused on GPT 40 many people are wondering where is GPT 5 we need GPT 5 and I think openi is focusing on two things number one they're focusing on the AI agents because of course that is new ground

number two they're probably focused on the reasoning kinds of models and number three they're probably also focusing on the coding area because I think there had there has been just a rapid expansion of users who simply use claw for many different coding products projects and I think it's quite likely that in the future open might release a dedicated tool I'm not entirely sure if they will just stick to building the base models but I think that since openai are also a product focused company it would make sense for them to do something like that considering that some mman you know he's been in YC a lot he's not really the kind of you know AI researcher type just to focus on good AI models of course you know you do need good AI models but he's more of someone that just you know focuses on the business side of thing and I think in some elements that is you know rather important for the company like one thing I see all the time a lot of people talk about how you know all this FL fancy new tool it beats everything in the benchmarks but a lot of times people don't realize that if that isn't baked into an act good product it doesn't really matter how good the reasoning is people are going to use something else that's much easier and they just know what it is so I think you know this is going to be something that they can do better than a lot of people so considering Sam Alman has been saying you know AI agents are going to join the workforce he's been saying that for some time I'm going to be really intrigued to see exactly what they do come up with now this is probably you know my favorite part of this video because I saw this and I was like okay this is probably the coolest Benchmark and I think once AI can manage to break this Benchmark this is where things are going to get super interesting so Ronan tamari tweeted this and it came up on my timeline and this paper I just had so much fun reading this but it said uh video models actually don't equal World models and there were some papers that spoke about this there was a paper that came out of China that basically spoke about how um you know they they conducted a small series of tests and the the kind of results that they got were just super surprising and essentially there was a paper that came out of China that was super interesting that basically backs up this claim but they basically said that look you know models like Sora actually have no idea of what a world model is and of course

people were like maybe controversial maybe it's just China trying to take a shot at open AI but guys take a look at this okay this is from Google Deep Mind and it basically talks about how we actually need a new path to AGI now because in the original Sora paper they talk about you know how it could be used as a path to scaling AI but I think this is going to completely change the industry and I know just been yapping on about what this is but essentially it's do generative video models learn physical principles from watching videos and essentially a lot of for like not the longest time but recently A lot of people have been saying that you know since you know video models like VO2 and so since they're so good at you know the motion when they're recreating those videos from a text prompt some people would argue that those models have some kind of understanding of a world model and I guess that could be the case but this paper and I'm going to show you guys this website right now cuz it actually have a GitHub and if you actually take a look this is called the physics IQ Benchmark so it says do generative video models learn physical principles from watching videos and essentially they don't they actually don't like I thought you know on some level they would because if you look at all of that footage you would have some you know understanding of physics but I'm guessing that even you know really good video models just truly don't understand physics so basically what they did was they took like the initial you know uh frame right here and then you can see like this is a real video of what really happens in the Basse reality then you can see what actually happens in the test realities you can see that it's you know it doesn't really do the same thing now for a better example of this I think this one is pretty difficult though to be honest with you guys like I think if you show if you showed you know the average person what would happen here maybe they wouldn't but we can see here like so we can see this uh you know this Roomba looking device that actually has paint attached to it and then it's spinning from left to right and you could see that you know it's got paint on the edge right there so you would you know expect the paint to streak across and go across but you can see here that these models like if we actually take a look at you know what these models are you know doing we can see that some of them just truly don't understand the physical reality some of them go completely off some of them just do a variety of different things but I mean it's super interesting research because when we also take a look at this one right here as well we can see that you know this uh you know this pot rotating on the same device we do take a look at you know what you're able to

gauge from these video models and then we can see that you know they just don't look pretty accurate and I think this research is really important because if we can understand that okay this meth me of training the model means that it doesn't have an internal World model that can accurately represent physics it means that okay we need to take a different method which means that we're probably going to be on the right path sooner rather than later so something that I also wanted to share as well was the fact that it actually isn't currently hype and the reason I'm actually adding this to the video was because a lot of times there will be pieces of news that Outsiders from other areas and other Industries would look at individuals within the AI community and would state that you know it's pretty much just hype that they're talking about but we have to look at what these companies are saying in terms of the CEOs who are I guess you could say a lot more conservative in their predictions and someone who is more conservative is of course Dario amod and take a look at what he says in this clip because I think it shows the perspective shift of how timelines are moving forward based on the trajectory of current models and the newer Paradigm I think until about 3 to 6 months ago I had substantial uncertainty about it I still do now but that uncertainty is greatly reduced um I think that over the next two or three years I am relatively confident that we are indeed going to see models that show up in the workplace that consumers use that are yes assistance to humans but are gradually get better than us at at almost everything um and the positive consequences are going to be great the negative consequences you know we we also will have to watch out for um I think progress really is as as fast as people think it is um one thing that I will criticize is I actually think it's very important now that fast progress is relatively likely um to uh to appreciate it with the proper gravity and to talk seriously about it um some of the other companies I won't name any names um you know there's just all these you know it's all these weird Twitter rumors like employees talk you know employees like you know have this kind of Sly winking like you know um nod to like oh there's these amazing things we're doing here um I actually think that's dangerous because someone on the outside looking at it is like oh man that's just hype that's that kind of communication gives the impression that this stuff is not serious and I think it's like really that's really dangerous

to do when it actually is serious um uh like I think the AI industry as a whole actually has an obligation to point to the seriousness of the moment that we're in if if we're really saying there incredible positive things that are possible and inevitably with any change this large there are risks we have an obligation to communicate seriously about it and to say what we actually think and I think he's actually right about that he's stating that look if we're really understanding where this technology is going to go then we have to stop messing around in terms of how we say this technolog is bad and I think this is a very important statement because he's basically stating that look the technology is real and the way that people talk about AI leads the public to believe that it might just be still hype so I think this is an important statement to just show you guys that look whilst yes there is hype sometimes about new technologies this is certainly something that is going to be revolutionary now another thing that was you know I was supposed to make a video on this but I didn't have the time at the time was the fact that GW actually talks about you know essentially the AI Revolution and what's going on and he actually speaks about the fact that there might have actually been a breakout he says if you're wondering why openers are suddenly weirdly almost euphorically optimistic on Twitter and before I get into this if you're wondering who gr is he's actually someone that predicted the scaling laws and I think he also had early information on GPT 4 so this is someone that has been pretty consistent in terms of the air predictions so it's not just like a random you know thread on Reddit but um yeah he talks about how their you know I guess you could say statements from the GPC 40 model 203 have been completely different and he said it's like watching the alpha go ELO curves it just keeps going up and up and if you're not you know familiar with what he's referring to I'm going to show you guys a picture now and this is a pretty low quality picture but I think you guys get the gist of this you can see that the curves just keep going up and up as long as there is more search and so far this is kind of what we've seen with our current model of AI and since alphago was essentially the smartest you know kind of model we're basically presuming that this is going to be the similar thing that happens with these llms and of course the levels of intelligence and if we can apply that framework to you know and make it more robust reliable then it's quite likely you could get AGI or even super intelligence now you can see here there it says there may be the sense that they've broken out and have finally crossed the last Thresh hold of criticality from merely Cutting Edge AI work which everyone will replicate in a few years to actually take off which is cracked intelligence to the point of being recursively self-improving and where 04 or 05 you'll be able to automate AI research and development and finish off the rest an Alman in November of 2024 said that I can see a path where the work we are doing now just keeps compounding and the rate of progress we've made over the last 3 years continues for the next 3 or 69 or whatever and of course a week ago they then said look we know how to build AGI as we have traditionally understood it we're beginning to turn our aim beyond that to Super intelligence in the true sense of the word we love our current products but we are here for the Glorious future so of course this is something that you know I think is really important because I think statements like these when we actually look back on them we're going to see that maybe this wasn't all hype and maybe this is something that you know give us an early heads up onto where the AI industry is going and you know after seeing those graphs after seeing what people are talking about you know this is definitely something that you should take into account of course I think I don't want to say maybe you should make lifestyle changes but I don't think that you should you know not plan accordingly in the sense that like maybe I'm not saying to switch careers I'm not saying that I'm just saying that you know stay adaptable understand the changes there understand the technology because trust me when I tell you guys I've been in and out of the AI sphere and sometimes I talk to people about AI like regular people and they really don't get it like I'm not saying that in a bad way I'm just saying that like the average person just isn't that you know they basically they you really

just don't care that much and that's completely understandable but I think it's like being early to the internet which is of course a great feeling if you're here watching this video now of course there was also this crazy model and I probably should have just had the benchmarks here but basically there was a new model called Deep seek R1 and this has just blown up the internet like because uh you know fact let me actually get the benchmarks now what I'm actually showing you is something pretty crazy but when we actually look at the benchmarks here we can see that this is a model that is currently basically just as good as open eyes 01 and I mean there is I think this is the current you know uh I guess you could say feelings from the AI Community I think 80% of the people and the general public think that this model is great it's amazing and it's basically just as good as op eyes1 and then you've got 20% of people that are basically saying look all they did was you know steal the chains of thought from op eyes model and manage to replicate that that's why the model is basically the same so I mean I don't know I do think that this is definitely great for developers you now have a model that you can reason with apparently you can basically get near unlimited rate limits so this is going to be something that is going to unlock a ton of productivity for those of you who are trying to build stuff I know I'm definitely going to be trying to use this in some automations that I'm going to be currently building because this is just so much cheaper and there's not really any rate limits and it's going to be something that you know allows you to access a thinking model at a much cheaper rate than many others so it's basically so much cheaper so I think as long as they can maintain that it's going to be really interesting and I think the thing is as well about this that actually shocked most people is not that they got there I think the thing that shocked most people was the fact that they got there so quickly like people were expecting you know China to be 6 months behind but you know or even a year behind but if they are you know let's say 3 months behind that is not a short time frame that like that that's basically saying that look there is the potential that China could take the lead here especially if they are on the same Paradigm as opening ey like all of these Labs some of them make breakthroughs and certain breakthroughs are probably going to lead to bigger gains and we're now in a stage where you know previously AI research was shared globally all the papers everyone would have access but now you've got these private Labs working making breakthroughs that makes their model even better than others and I think some labs are probably going to hit escape velocity and I think it's going to be super interesting to see if the USA manages to do that first or China manages to catch up and somehow overtake them I have no idea but um it's definitely something that shows the AI race is probably neck and neck at the moment but it's going to be really interesting to see where those things lie now the thing I wanted to show you guys as well was this right here and um the thing that we didn't get from the open AI 01 model that you know I guess a lot of people they were kind of not upset but they just wanted to see was the internal chains of thought for how the model think so the models they think for a long time and that's how they get better responses and with that um one of the things that we got to see with R1 was that it looked at this thing called self-reflection and it wasn't really trained to do that but it started doing that when it you know wanted to figure out the question and this was considered the aha moment of deep seek R1 so it learns to think using an anthropomorphic tone basically thinking wait a wait a minute I can actually solve this here and it's just a very interesting way because of course these models are built on you know system 2 Thinking which is you know your rational thinking which takes effort versus the system one thinking which is basically like chat gbt where you ask a question and immediately responds and then this is where we have the actual lead of the reasoning at opening i1 so if you ever wondered who was behind the brains behind that this is Nan Brown he is the lead at opening eye on this and he's done a variety of different things in the past that were really intriguing and I think it goes to show exactly how incredible his research has impacted the field system one thinking is the faster more intuitive kind of thinking that you might use for example to recognize a friendly face or

laugh at a funny joke system two thinking is the slower more methodical thinking that you might use for things like planning a vacation or writing an essay or solving a hard math problem after this competition I wondered whether this system to thinking might be what's missing from aot it turned out that having the bot think for just 20 seconds in a hand of Poker got the same boost in performance perance as scaling up the model by 100,000x and training it for 100,000 times longer let me say that again spending 20 seconds thinking in a hand of Poker got the same boost in performance as scaling up the size of the model and the training by 100,000x when I got this result I literally thought it was a bug for the first 3 years of my PhD I had managed to scale up these models by 100x I was proud of that work I had written multiple papers on how to do that scaling but I knew pretty quickly that all of that would be a footnote compared to just scaling up system and I think the reason that this is something that you should pay attention to is because he literally discovered something that you know allowed for 100x more results with of course a slight change and the point is that and over time we're going to get more and more changes that you know aren't going to be drastic in terms of potentially the system architecture but they're going to lead to a crazy amount of gains that means that you know accounting for exponential growth is quite likely that and so this is where we have Dr Jim fan actually talking about the singularity and I think it's rather important to discuss this because Sam Anan spoke about the singularity of course if you know GN is talking about breaking out then the singularity I don't think it's that far away but people have varying definitions of the singularity so Dr Jim fan says the operational definition of Singularity is that we are not truly done until Transformers start to research the next Transformer a less fancy term is Auto a decades all CS topic Singularity is basically automl at the extreme and this is basically you trade capital for higher intelligence without human babysitting basically saying that look the model needs to read research papers collect in curate data manage a GPU cluster monitor the training job and select its own offsprings from the Silicon perhaps they can even peer-review each other and lead a virtual AI conference but I don't think we are far away from this and I think this is the point that a lot of people do state is going to be the singularity and I think we do need to watch out for this moment in the future as this is going to be the moment that things start to recursively self-improve not in a singular system but the entire system as a whole from start to finish from singular idea to getting the model to improve itself the quicker you can iterate is the quicker that you can fix mistakes and the quicker you can have that Loop is essentially the quicker you get to the next level of intelligence and with that being said I think it does mean that once this Loop is solved of the automatic machine you know learning researcher I think that's when things are going to go into the stratosphere so that's when I'm definitely going to prepare maybe it is the case that they're doing this now but I would say that would and like my prediction for this is that like when you start to see maybe the software engineering bench start to go about like to you know maybe even 80% 90% or when you start to see those benchmarks being really you know broken and we're not far away from that I think we're maybe just like one and a half years away from those benchmarks being crazy I think that's you know the moment that we're going to get a ridiculous level of a research done and if that happens things are going to speed up 10 50x and then you know we're going to start getting ridiculous levels of intelligence that means the exponential curve has started now of course there was a video about China's first humanoid training base established by the national humanoid robot Innovation Center they actually began operations yesterday and this is essentially the factory that produces data and that's what Jang Lee the chief scientist of the Innovation Center said and it's say 50t training Hub that will launch over a dozen tasks involving 100 robots from more than 10 companies and institutions including the AGI B and Foria intelligence and the data gathered here will be available for companies who are trying to you know embody AI

intelligence and by 2027 they aim to scale to support a thousand robots under training so once again this is something that is moving across all fronts even things like AI models are moving forward and of course the embodiment of these models are moving forward too now in more humanoid robots I actually got to see the xang humanoid robot moving in real life this actually not my video but I'm saying that I actually get to see this and the reason I say that is because I remember when I made a video on this a lot of people were basically stting that this was CGI and that it was fake so I'm going to show you guys the video that people claimed was CGI but obviously we can see here it is out in public it is walking around and it is moving I think in 10 years we'll probably look back at these and think that the movement was awful but for now it looks very incredible and remarkable so you can see right here exactly what it looks like I think this one probably had a bit more charge I do think that these robots are definitely the same platform but the point I'm trying to make here is that it isn't far like I mean 10 years of research and development with billions of dollars invested into these companies I think these robots are going to be really smart moving around you know completely fast I mean the vision the voice mode I mean this future is clearly coming and we have to understand that you know as I spoke about the start of the video with humanoid robots like this we actually do have the Stargate project which basically says that you know the American government is investing $500 billion what do you think other companies are and you know not even other compan these other countries are going to do so with that being said let me know what you thought about this video do you guys think that the AI hype is real do you think it's fake going to happen at the end of the day I'll see you guys in the next