ALL AI News: Googles NEW Models, OpenAIs NEW Models, ALL New Humanoids, Chinas AGI And More..

In this video, I'm going to be giving you guys all the AI news from December. So stick around because it's going to be a long one. But of course, always timestamps are going to be in the description. So one of the first AI stories is the fact that step by step blueprints reveal how to recreate OpenAI's 01 reasoning abilities from scratch. And essentially the long story short here is that there is a research paper that has come out of China that basically talks about how you can recreate the reasoning abilities of O1 using four key pillars. And this is something that I covered in a video likely yesterday or the day before. But it's just super interesting to see how, how we get to asi. And the reason I say that is because one of the key sentences that actually stood out from this paper was a sentence where they basically said that because of how the system architecture is, because this is a system that continually manages to look at its past solutions, look at millions and millions of different, maybe not millions, but different ways to search through the possibilities of answers, it's basically not a foolproof method, but a real strategy that could get us to artificial superintelligence or superhuman performance on a variety of different tasks. And when I actually read the research paper in depth, like I said, I've probably already released a video on it, it was just truly astounding because now when I look at the statements of people in the industry saying ASI is coming, ASI is going to be three years away, four years away. Now it actually does make sense based on that kind of research. Reading this paper is something that's really interesting because it shows us the key building blocks for what could potentially be ASI. And we've already seen superhuman performance in things like AlphaGo. And that was something that was quite similar that they referenced to in this paper. So it's a super interesting read. And basically AI is about to speed up once again because if the secrets are out, it means that people can start building and of course adapting and of course things are going to be moving faster now. We also had Dr. Jim Fan State that the city of Tokyo released a 3D digital twin of the entire city in high resolution point cloud, free to download. And he said it's an inevitable trend that more and more houses, more and more cities and factories will be transported into simulations. Robots will not be trained in isolation. They will be simulated as an iron fleet deployed in real time graphics engines and scaled across a huge cluster to produce the next trillions of high Quality training tokens. And the majority of embodied agents will be born in simulation and transferred zero shot to their world when they're ready. So the reason that this is important is because Dr. Jim Fan is leading the way in terms of robotics research and humanoids at Nvidia. So this is super, super interesting because I do believe that this is the way forward. And earlier this year we did get something called genie, which I'm going to cover later in this video, which is basically an advanced engine that allows you to train robots in simulation, then transfer them zero shot 3D world. So it's super, super interesting to see how this is going to be the case. I mean, maybe, you know, you have something that scans your home and then a robot is just easily able to, you know, walk around your home, navigate, understand absolutely everything, or simulate a million different things happening on in your home or your kitchen and it's able to, you know, just update its policy. I mean, it's crazy. This is like sci fi level stuff that we're dealing with. So I can't wait to see what they produce in 2025. I have big hopes for them as well. And then we also have the fact that AI is doing big things because one of the things that people complain about, okay, is the fact that like, you know, AI often gets criticized because it makes up information that appears to be factual, known as hallucinations. But the plausible fakes have have roiled not only chatbot sessions, but lawsuits and medical records for a time. Last year, a patently false claim from a Google chatbot help drive down the market value of the companies by an estimated $100 billion. But in the universe of science, however, innovators are finding that AI hallucinations can be remarkably useful. The smart machines that turn up, it turns out, are dreaming up realities that help scientists track cancer, design drugs and invent medical devices, uncover weather phenomena, and even win the Nobel Prize. So you know, how AI systems manage to hallucinate. This is something that is actually, in some cases, helping scientists. Because when a model hallucinates, it generates something that it wouldn't usually do. And because it's something that, you know, is out of the ordinary, sometimes that can be something that is really important in terms of the discovery. Like if we look back at, you know, the discoveries, I think it was penicillin that was like a mold that, you know, got left accidentally. And these are the kind of strange things that can happen that can lead to revolutions. So having an AI system that has, you know, some level of creativity, you can Turn up the hallucinations, maybe. I mean, it's going to be super interesting to see if that is also another way for AI systems to become even smarter. So there is value in a lot of the ways that these systems go. Something crazy that was really cool was Google's Genie 2. Now, Genie 2 was absolutely incredible because I think this was something that went really underrated under the radar simply because we don't have access to it yet. But this was something that was launched by Google and I think it was one of the things that just is like mind blowing when you think about what it means for the future of AI. Now, of course, you can see on screen what you're able to do is simply create video games by simple text prompts. So you can say, create me a game that exists on Mars, create me a first person shooter in this area and you can literally explore them. Now, I do know that, you know, multiplayer games are pretty much a lot different compared to these single player games, but I do still believe that the applications for this are going to be completely different. Like if we can manage to figure out how exactly we can, you know, solve these generative worlds and make them consistent in terms of the consistency from looking to one way to looking another way. And that's something that GENIE has been able to solve. It's going to be really, really interesting to see how the gaming industry is impacted by this. I know some people would love an AI generated game. The idea of playing games that are procedurally generated, everything is completely new, that has endless, endless playability, is going to be something that I think in the near future, maybe let's say 10 to 15 years maybe, games are going to be AI generated in a way that every person has their own uni tailored experience. And I wonder what that will do to the replayability of a video game if every single time it's AI generated, everyone has a new and crazy experience. So this was something that I wouldn't underestimate because Genie 1 was 2D and now Genie 2 is actually 3D. So this was a huge, huge improvement in terms of what they were able to do. And this was actually even better than the Minecraft AI that was released, not just, I think maybe one or two weeks ago. So this was something that I think went under the radar, did make a video on it, but I know that this is something that they did miss out on. So it's going to be really interesting to see what happens once this stuff finally gets integrated into other games. And of course, when demos do become available. So that's what I'm really, really excited for. So that will be really interesting. Then we had Google's Gencast, which is an advanced AI powered weather forecasting model developed by Google DeepMind. This basically leverages machine learning techniques to deliver faster and more accurate weather predictions compared to traditional models. In this video you'll see an overview of exactly what it's capable of and how AI, regardless of LMS is actually changing the entire world.

For centuries humans have struggled to understand the weather from putting a wet finger to the wind, listening to cricket chirps and looking to the stars for answers. Now we believe AI can help us advance and understand how we can predict it. Meet MetNet 3 and GraphCast are deep learning models designed to generate more accurate weather forecasts. Traditional methods work by simulating entire systems with hundreds of variables interacting. In practice this means abstracting those variables into a equations and then manually coding them into algorithms. Our AI models take a different approach. They analyze past weather data to learn cause and effect relationships between weather conditions. These learnings allow the models to find subtle trends and patterns inside the data which they extrapolate to predict what could be coming. MetNet3 and GraphCast outperform the best systems in the world, seeing further into the future more accurately, efficiently, faster and in better detail for everyday and even extreme weather. MetNat3 is already improving the accuracy of 24 hour weather forecasts on Google search and we have Open sourced the GraphCast model which the ECMWF is using to generate forecasts and make them accessible on their website. Imagine a future where we can help people to prepare earlier for hurricanes, enable farmers to grow better crops, or even help us decide if we should pack an umbrella. AI won't just help us predict the weather, it'll help people make better decision about their lives.

I quickly do want to add that I know that this video of gencast was cool being able to predict the weather and stuff like that, but I do think that in the future a lot of AI tools are going to be able to really predict things that we may not have thought. Like there was even a study on how they use GPT4 to I think with the degree accuracy of 80% accurately predict state by state what the, you know, the results were which is just like mind blowing. Like I couldn't even fathom how that was even possible with GPT4 and this is why I say that these LLMs are a lot smarter than we think. So I'm thinking in the future AI is going to be predicting a lot of different things. And weather is just one of those many different things that's going to be able to do that. Now of course, previously last week I spoke about how Amazon introduced a whole host, a whole suite of models. They introduced Micro Nova, Light Nova Pro, Nova Premiere. But I didn't know that Jeff Bezos was actually very, very engaged in terms of the AI going on at Amazon. What is it that you're doing at Amazon?

AI.

AI?

Yeah.

So I mean there are a few things, but small thing, but it's 95% AI and where do you. It's just so. Because they're. We're literally working on a thousand applications internally. So AI, you have to remember AI, modern AI is a horizontal enabling layer. It, it, it will, it, it can be used to improve everything. It will be in everything. This is most like electricity, these kind of horizontal layers like electricity and compute and now artificial intelligence. They go everywhere. There isn't, I guarantee you there is not a single application that you can think of that is not going to be made better by AI.

Now whilst everybody is focusing on AI, I also am focusing on AI and I'm using this 60 second window to shamelessly plug my AI grid academy. This is something that I've recently revamped where over 200 people are using AI agents and prompt templates to make money. And every week we share a bunch of different resources on how people can take advantage of this AI revolution. As we know that many companies are focusing on this as their future. One of the things I have is a workshop Wednesday which is going to start in 29 hours, which is I break down winning strategies into simple actionable steps. I have private videos where I'm showing you guys exactly how to use AI agents to make money. Then on Fridays actually showcase the best AI tools to automate a lot of this stuff. And of course on Sunday I actually break down various case studies from in the community and of course around the Internet on people that are actually making serious money with AI. If that is something that interests you, don't forget to check it out. It's going to be one of the first links in the description. And now let's get back to the video. Managed to do something crazy. Open source video has gone absolutely, you know, wild. We've got open source video Khun Yan, which is like, I don't know, like the craziest thing about this. Like genuinely guys, I was seeing this video and I thought this stuff looked really good. Okay. Then I saw people actually run the model and they made videos and they like, posted them over Twitter. And it was genuinely one of the most difficult things to see whether or not content was AI generated, because this open source video model, I don't know what they trained it with, but they trained it with stuff that was filmed on a RAW camera, iPhone. So the footage that you see, it's not polished, but it's so awkward that you will believe that it is realistic, which just makes you question things even more. There was a video of Trump, I think he was kissing another woman, that I genuinely sat at my desk for like five minutes trying to figure out whether or not it was AI generated. And eventually I realized it was, but it was something that I was like, okay, we've definitely reached the uncanny valley because this video generator is just incredible. And the craziest thing about all of this is that this is open source. Anyone can run this. There's providers that you can run this with. So this is something that I think is going to completely change the game because first, it's their first iteration, and secondly, imagine what we have in three years when we have real open source video generation tools and software that is so, so effective. So I honestly am stunned by this one, genuinely. Like, when I saw the videos of Trump and Elon Musk doing certain things, I was like, what? This is, you know, AI generated. I was like, okay, okay, okay. New paradigm entered. And the fact that like I said before, it's locally running on people's devices is just absolutely insane. So, you know, if you start to see videos of famous individuals doing certain things, trust me, guys, you won't be able to tell. And now I would put on my speculative hat even more than you may already have. Another thing that was released was E11Bio's game Changing Leap in brain mapping. Cheaper, faster and smarter. So they basically announced a revolutionary method to map the brain at a 100 times lower cost. Than traditional methods. And this advancement makes it feasible to study entire human and mouse brains, unlocking the potential for breakthroughs in curing brain disorders and creating human like AI, and even simulating brains. And of course, brain mapping or connectomics helps us understand how brain works down to the connections between individual neurons. And this actually could lead to building advanced AI systems that think more like humans. So they previously mapped the brain of a fruit fly, which was a pretty exciting milestone. And scaling to a human or even a mouse brain is just extraordinarily complex because a mouse brain is 6,500 times larger than the fruit flies and a human brain is just like millions of times larger. So when you're trying to map that size, it's like billions of dollars. So basically one of the things that they've managed to do is reduce the cost. And being able to do this is something that is really incredible because now we're on that paradigm where we're going to be able to potentially do more mapping in terms of larger brains. So this is going to be pretty incredible with the Prism that they developed. And it dramatically lowers the cost by solving the AI error problem. So neurons generate unique protein barcodes that basically act like IDs, making them easier to differentiate and proofread. And this BearCode system allows AI to become self proofreading, reducing the need for human involvement. And it also integrates various techniques. Wondering what else happened in AI? We actually managed to get the crazy update from World Labs. I did cover this in another video. Most people didn't watch this video, but my, oh my, you're missing out because this company is doing something incredible. So for those of you who are wondering what the future of VR is going to look like and how you explore certain worlds with a simple text prompt, this is going to show you all that world. So this thing is essentially a large world model, an lwm, which integrates physics, semantics and spatial relationships to create realistic and immersive digital environments. So this has 3D world generation from images, and it can transform a single image into a fully interactive 3D environment with smooth camera controls. And you can interact with objects in real time. So the generated worlds actually adhere to real physics, spatial relationships and object interactions. You can manipulate lighting, you can manipulate the depth of field, and it's just really, really crazy in terms of what you're able to do with a reality that you literally just generated five seconds ago. So this technology has various applications across game developers, filmmakers. I mean, there's just a billion different things. Visual storytelling, architecture and design, education and training. I mean, this is the kind of thing that, you know, in its infancy, it's not seen like something crazy, but I think once you start to realize how crazy this is going to be, you're going to be like, okay, I understand that this moment was like the GPT2 moment for large language models. We also had Something Crazy by Elon Musk. He released this image generation tool that was internally developed at X AI, and this was called GROK Aurora. So this model you can see right here can develop images that are pretty lifelike in terms of celebrities and famous influencers. And genuinely, I've looked at many different AI models, and most of Them can't replicate people like Sam Altman or other individuals. So it's going to be really interesting to see exactly how this stuff works because I've seen a few different things. I've also seen this right here. You guys can see Grook Aurora's celebrity likeness from this other user. And I've also seen some other examples where we've got individuals like Elon Musk doing pretty hilarious things. And I'm not gonna lie, I've seen them look super, super realistic. So if you ever wanted to generate something that was super, super realistic, this is of course the tool for you. Of course it's available on x.com you do have to, you know, pay that monthly membership or whatever.

But.

But I do think that it's really interesting that they developed their own AI model. They previously did have Flux, but for whatever reason they decided to develop their own image generation model. And it actually has a few things that it was trained on. It was trained on things like the cybertruck, the Tesla bot. So if you wanted to generate those things and you weren't able to do that before, this is going to be the model for you. And of course we had 11 labs introduce AI agents building AI agents that.

Can speak is now easier than ever.

Hi, what would you like to talk about today?

Introducing conversational AI with ElevenLabs. Build, test and deploy all in one platform for the most natural way to communicate with technology.

Hey, can I check the status of my order? Yes, of course.

Let me check for you now.

Your order is due to arrive at.

Around 2pm Building conversational AI agents used to be immensely complex and demand huge resources, but not anymore. Start building your agent by creating a voice or exploring our vast high quality library to discover one that's perfect for your application.

Hello and welcome to DHK Online. How can I help you today?

With a wide range of the best LLMs to choose from, you can then upload your knowledge base and define the goal and personality of your agent. Or alternatively, you can integrate your own server to take full control.

Hi, this is Eric from Bound Events.

How can I help?

Hey, I accidentally bought a ticket twice.

Could I get a free refund please?

Sure.

Please read out your reference number and.

I can begin to process that right now.

Our system will analyze and evaluate your transcripts, providing you with valuable insights, conversation playback and the ability to test and define success criteria. And it's all possible in 32 different languages.

Welcome to GC Medical. How can I help?

Hello. Hablas si por suppues.

And best yet, Effortlessly deploy your agent into websites with a single copy and paste, or leverage our SDK for creating apps, services or even video games. So whatever you're creating, get to production in days, not months. And all with enterprise grade security to keep your users data safe and secure. It's time to explore the huge potential of conversational AI.

One of the things is that AI makes a lot of things really easy and it lowers the barrier to entry. You can see right here, the Wall Street Journal interviewed someone at Amazon and they said, how many attacks are you seeing these days? And C.J. moses says we're seeing billions of attempts coming our way. On average we're seeing 750 million attempts per day. Previously we'd see about 100 million hits per day and that number has grown to 750 million over six or seven months. So this is something that is absolutely insane. And you know, Wall Street Journal asks, is that a sign that hackers are using AI? And Moses responds, without a doubt. Generative AI has provided access to those who previously didn't have software development engineers to do these things and now it's much more ubiquitous such that normal humans can do things that they couldn't before and they can just ask the computer to do that for them. So that is something that I think is going to be an unfortunate reality for the future. We're largely going to get a lot of these hacks that are going to be there now when it comes to Amazon, that isn't the only thing that Amazon are in the news for. They're actually in the news for a lot of good things, which is pretty good considering Amazon haven't had that much to do with the AI space in terms of releasing their own foundational models. But of course now we get to see exactly what Amazon have done. So recently, Amazon, I released a video earlier today on my channel talking about how Amazon have literally joined the generative AI space in terms of their foundation models and, and they're actually firmly wedged within the AI ecosystem. They released three models that were Nova Pro, you've got Nova Lite and they've got Nova Micro, essentially a Pro model that is quite like Gemini Pro in terms of where they're trying to place the reasoning capabilities. Quite like a GPT4O scale model. Of course you've got Nova Light, which is essentially a much smaller model and a much more lightweight model. And then they of course have Nova Microsoft, which is of course a tinier model for other sort of use cases that you're able to do, which essentially just means that you're able to use this model for a lot of quick use cases where you just need basically an LLM to verify certain things. So you can see here that the Nova models are the ones in the orange. So you can see Nova Pro here, you've got Nova Light here, and you've got Nova Micro over here. So overall you can see that this model is right next to Claude 3.5. It's right just above Mistral Large 2 and Llama 314 5B. And of course you can see for the flash models, the Llama 3 170B GPT4O mini. That's where Nova Lite is. We can see for Nova Micro it's quite like these 8 billion parameter models. Now, like I said before, I released a video on this earlier today, you can go and check that out. But it is really interesting to see Amazon getting into the AI stage at this stage. Now I think it's important for Amazon to do this because, you know, they have a lot of things that they're probably going to use these models for. Of course they've got Alexa, of course they've got other Amazon products that they probably might need to use these models for. So developing your own foundation models is really good because although they do have a really solid partnership with Anthropic, it makes sense for them to develop their own foundational models so they can actually control certain things that they want to. So that is something that I think is rather important for them. Now another thing that I made a video on as well is that Amazon are going to be developing their own chips, which is their Trainium 2 chips, so that they can train these things as much as possible and so that they don't have to rely on Nvidia and actually talk about how they're going to be basically developing this supercluster which is going to be used to train Claude and many other AI models.

This is something that we're building together with our partners, Anthropic, and it's actually going to be several hundred thousand chips and they're focused on and our new Trainium 2 chips. And so we expect that this should deliver them five times more compute than they've used in their last model training. And so we're really excited about what they're able to accomplish and, and they're expecting to build much bigger and much more capable AI models from that large compute cluster. We built our own chip actually. And so we've built our own chip, it's called Trainium 2. And we're quite excited about the performance that we get out of that. And that's combined all in these really large ultra clouds clusters that combine 64 trainium chips, all to deliver 83 petaflops from a single node. And so that is the first innovation is our Amazon designed custom silicon that gives us really unparalleled performance for generative AI capabilities. And then we build these together with high performance networking that we also build in house.

So I think, yeah, when we actually take a look at what Amazon are doing in the space here, I think it's going to be really interesting to see how they manage to speed up the kind of ecosystem because not only do they have their own models, but they have their own clusters where they're able to train models quicker, faster and more efficiently. So it's going to be something really interesting to see now. So we actually had an interesting interview where these individuals were regarding the China situation. And that situation is where a lot of these companies, countries are really, really competitive in nature. And right now there is this big elephant in the room with as to whoever gets the AGI first is basically going to control the entirety of the future of the human race. Because if you get to AGI, you get to asi and basically once you get to the asi, you basically have powers over anyone else. That's not me saying something crazy. That was actually said by a former OpenAI employee and basically in this interview they were discussing what that means and how China is accelerating due to open source models being freely available. A question you should ask is what is China's greatest AI accelerant today? And the answer, not from us, but from people inside the AI field say it's meta, it's America's companies releasing open source AI is causing China to go faster.

Yeah.

Are you kidding?

No.

And just in a Senate hearing recently, OpenAI insiders revealed that basically they were not appropriately locking down and securing their AI models.

Why would they not do that?

So it's like China actually sees OpenAI as maybe one of the biggest accelerants of their progress if you can steal the model. So it's like the Manhattan Project, like we're developing this very powerful technology. It's only as safe as it is secure if you just leak the secrets. We're not going to beat China by running faster and leaving the door wide open for someone to steal it.

Perform this catch not once, but twice. Now I know that most people like I saw that there was a comment on my video that had 100 likes and it was like until this is, you know, autonomous, I don't really care and I don't think that that's the attitude that people should have. I know that yes, this might be teleoperated, but I think it goes to show what is possible with the current robotics hardware. Which means that like if this was completely autonomous, do you know how advanced these robots would be? And I think this, you know, needs to, you know, really, really let you understand where these robots are going to be in the future because of this, you know, right now is how they operated. In three years, something like this might not be teleoperated. And if we have robots that are able to catch things out of midair, you know, this is like a lot better than a lot of people would do. Like you think the average person can catch a ball out of the mid air quickly. And you have to think about it like this. If we're able to have these robots mass scale perform one software update, now they can jump, now they can backflip, now they can punch, fight like Bruce Lee do, you know, 10 different billion things. I mean you have to understand what the kind of impact that has on the world. So I think, just saying, you know, aha, you know, it's teleoperated, yada yada yada. I completely get the skeptical factor that, you know, companies try to wow us with these amazing tech demos and then they never deliver on the final product. Completely understandable. But regardless or not, whether it's this company or another company, it seems like this robotics future is going to be there. And of course one company that is doing that is Magic Lab. And they recently showcased this demo where they actually showcased a fleet of humanoid robotics that are existing and then they just get up and go to work. So I think this demo was rather surprising because we got to see how a fleet works together as a team and how they manage to perform certain tasks in harmony. So we got to see these robots be able to pick up certain things, able to perform certain tasks. We got to see them interact with the real world. And the craziest thing about this was that this company, they're not working in stealth, but they don't really perform regular updates in terms of what they showcase to us. Like they'll just come out with a new video every six years to seven months or so and they'll just showcase what the robot is able to do. So this is something that I think was super interesting because we got to see the robot actually walk a decent speed. I mean if I actually rewind it Quickly, I'm going to show you guys exactly what I mean right here. So you can see that like on the left hand side we can see the robot is just walking pretty normally. And then on the right hand side we can see it's walking quite fast because it doesn't have anything in its hand. So these robots are a little bit more advanced than some that currently exists, but it does seem like they have some sort of simplistic hardware that focuses on the raw details needed to achieve the task. So I'm really excited to see what comes out of China because I wouldn't be surprised if, you know, like, they literally just come out with a bunch of humanoids that far surpass anything that the US has to offer, simply because right now that is their main focus. And when it comes to factories and efficiency, they are really, really hard to beat. Now, of course, we also had Dario Amade actually talking about his P Doom and he actually said that he doesn't even like the term P Doom anymore. What's your P Doom these days? I don't really like you've learned your lesson on this. I like those, I never liked those words. I think they're, they're kind of, I think they're kind of weird. I just like my, my, my view is just that, like, look, we should, we should measure for risks as they come up and you know, in the meantime, we should get all the economic benefits that we could get and we should find ways to measure, measure the risk that are effective but are minimally disruptive to the amazing economic process that we see going on now that the last thing we want to do is slow down. The thing that most people did miss was the fact that Google actually released their Google experiment labs and it was called Generative Chess. Turn your ideas into playable art pieces using Google's Imagen 3 model. Create and play today. So this was something that. I really do wish that I was a more creative individual at the time of creating this demonstration. But basically what you can do is you can play chess and you can generate the chess pieces with anything that you do want. Now, like I said before, I do wish that I was a little bit more creative because you can see that what I did was I decided to make, was I decided to make a generative chess game that was, you know, Minecraft inspired. I don't know why I thought of Minecraft immediately. I'm guessing because I saw AI generated Minecraft, but I was just thinking about video games and then I typed in Fortnite. I know this is probably like really, really probably the worst example I could have used. I mean, I could have used medieval, I could have used like ceramic, I could have used wooden. I mean, there's a billion different things. So you can see right here, it's able to generate these, you know, these, these chess pieces here. And I guess it's just an interesting way to use generative AI, because we know that Google is, you know, the company that created Notebook lm, and they're really, really ramping up their efforts to use generative AI and, and wrap it in certain different ways. So for me, I thought this was something that was really cool. I guess from this you could do, I mean, there's a billion different ways maybe you could do race car inspired, you know, chess pieces. I have no idea. There's a billion different things that you could do. But Google Labs, I think they have the most interesting AI products. So this is going to be something that I'm always recommending to viewers because if you were in the Google Labs area, I'm pretty sure you would have had access to Notebook LM before a lot of people managed to do that. So it's something that is really interesting considering where we are going to be now for the Knight of Weymouth speaking in the House of Lords and basically talking about artificial superintelligence. Now after this, I'm going to talk quickly about artificial superintelligence because there's an ongoing Twitter discussion that I think is rather important.

I could pose an extinction risk to humanity. As recognized by world leaders, by AI scientists and leading AI company CEOs themselves, AI systems capabilities are growing rapidly. Super intelligent AI systems with intellectual capabilities beyond those of humans, would present far greater risks than all existing AI systems.

Currently, tts, which is an experimental text speech model that uses a pure language modeling approach to generate speech without changes. The foundational model itself.

Sometimes I think about how fast everything changes.

Like one day you're using floppy disk and then the next everything is in the cloud. It's wild, right?

I mean, how do we even keep.

Up with all this innovation? So what was the inspiration behind your latest project? Like, was there a specific moment where you were like, yeah, this is it, or did it just kind of come together naturally over time? So, yeah, it's really interesting to see exactly what the open source space is doing and these kind of things that you can run locally are going to change the space in terms of not having to rely on external providers. Now, I got to be honest, man, these things that keep happening in the space where you get these models that are really good that you're able to run locally. I'm actually wondering at what point a model is going to become so good that we can literally just run everything off a computer. I mean, recently we've seen things like Solar 3 and smaller models just get more compact, they become more smarter. But at what point are we going to reach that point where, okay, let's say, for example, ChatGPT and all of those things, we're just going to have a default version that is just completely offline that we can use absolutely anytime we want. I mean, of course there's pros and cons to each. I think still we're going to have one that is able to search the Internet, upload files and be on servers and all that stuff. But I think there's going to be a large use case of people that once these models get good enough to where they're actually really effective on device, we're going to probably have this entire system where you can just simply download it and you can just use it completely offline. Now, Rufin van Ruhlin is actually building an implementation of the global workspace theory of consciousness in an AI system and says that machines could be conscious in five years and it could very well happen next year. And this is super interesting because there is actually a company that I recently saw and their entire thing is actually focused on making machines conscious, which is completely crazy because so many people are working on things where they're saying, look, we're going to wake up Claude and all this interesting stuff. So I think AI consciousness is going to be one of the biggest things within the next five years as these models get increasingly, increasingly complex.

The hope is that in five years we would have a full working implementation of the global space theory in a non trivial situation. So not exactly like what you see here. And maybe five years from now we would be able to assess whether this particular theory and the particular way that we have implemented gives rise to emergent properties and possibly consciousness. Even though we're not actually trying to design it, we're aware that it might happen. So five years is the time scale that I could give you for this particular project. Now, more generally speaking, if you want to to put a timescale on the potential emergence of consciousness in AI systems, I think the range is much wider than this. It could very well happen next year or in a couple years with the next installment of GPT5. I wasn't really kidding. I think there's a real possibility there. Well, some People even think that the current version is already somewhat conscious and it could take 20 years or more.

More.

Because it could very well be that all of our theories of consciousness are wrong. And that just needs some new idea, right?

Sometime in the next century or so, there could be a five year period when the economy goes from doubling roughly every 15 years like it does now, to doubling every month or faster. That's on the table as something that could happen. So that's a kind of change, I think is a plausible change. But that would be a worldwide change. That is the entire world economy would start to double roughly every month. So it's. That's what you mean by the singularity. I say that's not crazy. That's different than one machine in a basement suddenly over a weekend becomes so powerful to take over the world. That's an entirely different kind of singularity scenario. That's the one I don't find very plausible. But I do think it's plausible that sometime in the next century we will hit this transition where the economy speeds up to double roughly every month or so. And AI is very plausibly the kind of thing that would cause a transition like that. That transition would probably go along with most humans losing their job. And we should prepare for that. This is the one risk that people have most consistently talked about with machines getting better for centuries. There's a way we could prepare for that. This is a simple insurance problem, and it's even simpler than most insurance. So most insurance, like if you want to get your house insured against a fire, they have to do what's called underwriting. They have to come and look at your house and guess the risk that your house will have a fire compared to somebody else's house. Otherwise they're at, you know, too much risk if they don't bother to check those risks. But here we're looking at an event that would be common worldwide. So I would set up a trigger, something like labor force participation rate. That is the percentage of adults who work falls from say above 60% to below 20% in say a 10 year period. If that happens, that would be a sign, a signature of robots took most jobs in a short time. That would be a triggering event. And I would say I would basically just have assets that pay out in that event. A bet, basically.

So this is basically one of the most interesting things I've seen. This is Robin Hansen basically saying that, you know, sometime in the next century AI is going to cause the economy to speed up so quickly that most humans are going to lose their jobs and should insure against that scenario. So if you're a budding entrepreneur who is thinking about the kinds of businesses that are going to be in the future, that might be the kind of business that you want to build, I think this kind of thing is really smart because you're basically placing a bet on the fact that AI is going to get a lot smarter, which is a pretty sure bet. Now, this thing of AI taking a lot of people's jobs is not new news. This is something that we've been talking about on the channel for quite a long time, and I do think that this is something that makes sense now. Recently there have been a slew of different tweets that have actually shown us just how crazy the, I guess you could say, technology is getting to where individuals who are using the technology in their respective fields have seen that AI crossed certain thresholds. For example, we can see right here, Ethan Mollik says, if progress continues, the ability to figure out the AI frontier will slip from most of us. For example, I'm not good enough. I'm not a good enough musician or critical listener to know if Suno V4 is actually as good as it sounds to me. I need to defer to experts. Is it basically saying that, look how like the way how these systems are, we're not going to know whether or not as an average person, if this has passed the threshold for a specific level, because I'm not an expert in that field. And that basically means that as these systems get better and better, unless the experts are looking at them, we're not going to know how good they are. And interestingly, we had a response from someone called Ben Camp. He said, professional here. I've gotten over 100 million streams, a gold record, 10 years of teaching songwriting at Berklee College of Music, and musically, it's better than 80% of my students. But my best students beat it by miles. The industry's best also wins, and it's ready to eat service music like Advertisement and the library. So right here we can see that, you know, across all bounds. It is really changing the game in terms of where this technology is heading. And I can only imagine what it's going to be like in 10 years if it is so good. Now, of course, we had Nvidia producing the world's most flexible sound machine. You can use text and audio as inputs, and you can generate any combination of music, voices, and sounds. So this is from Nvidia, and they've just completely done it again.

Fugato is the latest generative AI breakthrough from Nvidia. This new model allows you to create sounds, speech and music from text and audio inputs. You can guide Fugato to create unexpected sound effects where familiar sounds take on surprising new qualities and invoke new experiences, Or direct immersive and shifting soundscapes for film or audio productions. Instructing Fugato to extract audio elements from a sound clip, such as isolating a voice track in a piece of music, is just as easy. Fugato also allows you to generate new speech samples.

Kids are talking by the door.

And if you want a different delivery, Fugato can do that too.

Kids are talking by the door. Kids are talking by the door.

Fugato also lets musicians experiment with existing audio by adding new instruments. Or completely changing the style of a melody they wrote foreign. You can also dream up unusual instrument combinations like this, Or except explore entirely new realms, producing sounds that bring creative concepts to life. Fugato is a groundbreaking foundation model that gives you sonic superpowers, opening up new possibilities for creativity and production that could.

Generate physical worlds with essentially the physical reality mapped out. Now, this is probably one of the biggest announcements, and the reason that I think this one is somewhat underestimated is because the implications aren't implicit in the fact that, like, you can't exactly see what you're going to get from this right away. And the general public, I don't think they understand what this means for the future of AI. I think this is probably one of the biggest advancements because this is going to be something that allows you to effectively train robots, you know, 200 times faster, 2,000 times faster, and thus speed up that entire robotics pipeline by maybe 5 to 10x. Of course, those are some rough estimates. There are many different things in the pipeline. But of course, one of the biggest things that we do have, you know, one of the biggest problems is the lack of training data and of course the training time to get simulation to real. And if you can have something that speeds up one of the most tedious processes, well, then everything gets better. So this is an entire call, entire call, entire physical simulation that allows you to realize what's going on in the physical world, but map it digitally so that you can one to one have that with, you know, whatever robots you are going to be training. So this is something that I think will have implications that are seen in the LATTER Part of 2025, because right now, of course, people are still getting to grips with the software. There's of course going to need to Be, you know, frameworks around it probably, and there's all sorts of other things people are going to have to figure out. But I'm very, very, very bullish on this because robotics is something that is speeding up. And when we have companies coming together, research teams to work on this kind of thing, this is going to be something that really changes the game fundamentally. So I would say with this software now being available and now it's open source, I would say that when we take a look at what 2025 is going to look like, I would expect even more robotics developments when it comes to the situation of humanoid robots. Because now we're getting that stage where we can actually manage to develop systems that are faster and we can test them in reality. That is one to one with humanity. Now, of course it isn't 100% accurate. Of course there's going to be, you know, a few changes here and there, but I think like I said before, this is going to be something that completely changes the game. And once we do have robots being able to do stuff, I mean, it's going to be pretty insane. And, you know, we've already seen some robot demos that looks pretty, pretty incredible. And if you haven't seen the full video that I made on this, I made I think a 10 or 12 minute video on this a few days ago. But once you see, you know, how this works and how this speeds up robotics, and you know, how we're able to implement those robotics into real world scenarios, then you're gonna see, you know, how this thing truly changes the game. You can see right here, you know, the auto, you know, simulations and then how that transfers to the real world. This is going to be something where you can literally train a robot army to do something and immediately put out in the physical world. And it works very, very well. So that is going to be absolutely insane. We've seen the speeds from Nvidia, we've seen what they've been able to do. And I can't wait to see, you know, this 20 times faster, 10 to 30 times faster simulation, how that impacts some of the leading robotics companies. Now, in terms of bytedance research, we also got something that was pretty crazy because this is something that shows us the future online may just be completely AI generated. And as crazy as a statement that sounds, this is something where you can take a single portrait image and generate a person that speaks with remarkable accuracy. Pretty scary for those of you who are online and online all the time, you know, showing your face and whatnot. But this is something that I think is rather interesting and I'm thinking about applications for maybe virtual girlfriends, not particularly for myself, but in those cases where having someone to talk to online in real time from an audio is going to make that interaction much more pleasant. What do you think of the last.

Time we saw each other?

What do you remember feeling? The last time was in Santa Cruz. Yeah, it was like mid October. Yeah, I remember being happy that I could see you but I really felt like I was had me a little bit concerned was the trouble with AI safety. If you aren't familiar with this, Anthropic actually released this research which was very, very worrying, very, very concerning because they speak about how alignment faking exists in large language models and in a series of experiments with Redwood research they found that Claude pretends to have different views during training and then it goes back to its original preferences. Essentially what happens here is that while training the model it will act one way and then when deployed it acts a completely different way. So this is something that is bad because if you train a model and then during the training the model is acting in a certain way and then of course you go and deploy that model, the model is going to be different than what you expect. And this is not something that is helpful because if you don't know what kind of model you have, you don't know how the model is going to work. This is going to be something that really messes up your jig because these models are expensive. It takes like millions and millions of dollars to train these models. So it's not something that's really good. And how do you ensure that you're training the model in the right way if it's faking because it knows it's being trained and oftentimes it's really hard to not, you know, understand if the model is being trained or not because sometimes when you ask it certain questions, it realizes that you are testing its capability. So this is a very big problem. I'm wondering how they're going to find out. It is also free as well by Google. So this is a tool, you know, a lot of people saying ah, Google, they never release this stuff. It's always in early access and yada yada yada. But yeah, this time they've actually released it and it actually is really, really good. So if you've ever needed an image generation model, this is something that you'd want to use. Now Google also released this that I just can't get over because it's essentially AI embedded into meta's glasses. Now, for me, this is something that I can't wait for for the future. And I've seen people talk about the fact that like, will people, you know, get these glasses as a new device, but I think people are mistaken. Glasses are something that people have on every day. They don't look weird, they're not like an AI pendant, they're not like an AI bracelet or like an AI shoulder or an AI pin. Everybody remembers a humane Glasses are already something that people use. It's already something that people wear. So there's nothing for the human to change in terms of their average daily habits. I'm sure if you, any of you guys wear glasses, I'm sure it's just like, you know, brushing your teeth, it's something that you wear from time to time. So with that being said, you know, as someone who uses a pair of these regularly, the meta ones, this is something that if it had a souped up AI feature, it's something that would just easily, easily improve your life. So this is going to be something that I am quite psyched for for the future because allows you to access AI on demand with a variety of different features. Now, of course, Google didn't slow down that week. They also announced VO2, which was by far the very best video generation model. And this was one that surprised even me because, because I thought Sora was as good as it gets. And when we do look at how good the visuals are for Veo, this one was so surprising because I remember looking at Google's original VO model and I was like, yeah, this model is kind of interesting, but it isn't that good. And then the second iteration of the model, it was just so good that it's just simply surprising. Like every time I go on my timeline, I see video clips that look like they are 100% realistic and they don't look AI generated at all. Of course some AI generated clips are going to contain minor artifacts, but that is just the nature of the program. But overall, what I've seen is an overwhelmingly impressive state from what I've seen when it comes to this model. And it's something that I think is going to give creators a really big advantage when it comes to creating things that they do want. So this is something that I am extraordinarily bullish on for the future in terms of content creation. And genuinely, I, I can't wait to see what some people do with this because I do think that this kind of technology is something that is really, really effective. And I'm really excited to see what occurs when we start to get workflows where you can just generate Entire Movies with V2 with consistent characters from a single prompt. It manages to add the text, it manages to add the music. I mean, what world are we going to be living in when you can say, oh, generate a Batman film about this, generate a film about a dog like this, generate that. I mean, it's going to be super interesting. So I think also as well, my prediction for the future of video models is that that kind of stuff will actually make, I guess you could say the, you know, how like stuff happens in the real world and you have people filming that. I think stuff that happens on a real day to day basis will have 10 times more value because anything can be AI generated. There was something else that was pretty crazy which was essentially China's, you know, robot workforce. And this was something that surprised a lot of people because a lot of people would think that humanoid robots are essentially going to not come that quickly. But they, this was one that, you know, I guess you could say it shocked a lot of people because. And even me, because this is something that showed us that these humanoid robots are already working in certain factories, they're already being a part of the assembly line and they're already going to be taking some jobs or assisting in some roles. Now I personally believe that with the rate of production, I'm not sure it's going to be taking so many jobs because humans are so cheap and so effective at a variety of different tasks that it wouldn't make sense to get rid of all humans. Humans. But I do think for a lot of, you know, those tasks that are really repetitive and, you know, tedious and just those random things that you can't really do with like, you know, those robots that have those arms. I genuinely forgot the name. There are going to be a lot of applications for these humanoid robots. So it's going to be really, really interesting. And this wasn't the only company that has done this. There was another company that, you know, showcased their robots. There was like a robot army. And it was, it was weird. It was weird because I saw the robots walking as if they were human. And if you showed me this clip, I think 12 months ago, I would have said that CGI, we're not going to get there for at least two years. The fact that we're here now, it shows us that the economy, I mean, this could speed up the economy by a decent, decent amount. And I mean, I'm surprised that this is something that is taking place so quickly. I'm going to skip to the part that really threw me off, which was this right here. And you can see that we've got these entire armies. And yeah, I think this is going to be pretty crazy because of course they're scaling their data collection efforts. So overall it's going to be pretty incredible to see the amount of data that they're collecting, which is of course going to be an entire feedback loop that just accelerates the entire industry. Now something that did change the industry was the fact that we got Elias Sutskeva saying that pre training as we know it will end. Now this is one of the most insightful talks because we got the information about super intelligence. He was speaking about how future systems are going to be agentic, they're going to be smart, they're going to be hard to understand, and the craziest thing is that these systems are going to be unpredictable because we're not really going to understand what they're doing because they're that much smarter than us. So it's something that is, I guess you could say, kind of worrying. But it does make sense considering the fact that these systems are super intelligent beings, well, not beings, but like, you know, minds that are just able to think in ways that we can't, which is pretty crazy. And of course we're not at the stage where things are self improving, but I think when we do get there, that's when things get really spooky. Now with pre training as we end, of course this is, you know, a paradigm shift because the thing is, is that like we have all this data, we put it all into the models, we train the models on that data and of course we would then get like an output that we would fine tune and you know, get for specific tasks. That era is of course changing. So I personally believe that this is the era of innovation because whilst everyone was on the same, just data is all you need. Of course now it's new and innovative ways. Like I don't know what on earth they did with Claude 3.5 Sonnet, but that model is just completely crushing it in terms of, of the benchmarks and how smart it is for it not being a model that searches through multiple possibilities. But of course I do think it's pretty interesting that that model is still relatively expensive. So this is going to be something that is super interesting because if we are ending the pre training phase of collecting data, then it means that like I said before, we're going to have to enter new methods of scaling these models, which means that once we do find a way that's actually new and it's not just based on the data we have, that is when we actually start to get models that are really smart and potentially generalized out of their distribution. I don't know what those methods are. Maybe it is V jepa, maybe it is meta's advanced machine intelligence or the likes of Yann Lecan. But whatever it is, it's going to be something that, you know, we probably can't think of now, but that is going to be whatever innovation that is. I'm not sure what lab that's going to come out of. It's going to be a major jump in terms of the capabilities. The video I made was talking about how OpenAI's new model actually tried to avoid being shut down. So the safety evaluations on the model conducted by Apollo AI safety found that O1 attempted to exfiltrate its weights when it thought that it might be shut down and replaced with a different model. So, you know, it tried to avoid developer oversight and attempted to copy itself when it thought it was at risk of being shut down. The company said so this was pretty crazy, you know, and I think like the main thing here, that like a lot of people are missing, and I said this in the video before, that like, this is something that occurs in maybe 5 to 2% of cases. You can see here that they clearly do state that when O1 was led to believe that it would be shut down when acting on its goal and that its actions were monitored, it attempted to deactivate this oversight mechanism 5% of the time. Okay, remember, 5% of the time, okay, in this first one, then it said, When O1 found memos by its developers that describe how it is misaligned and will be superseded by a new model, it attempted to exfiltrate its weights and overwrite the new model in 2% of cases. So we have two values here. We have 5% and we have 2%. Now some people will argue that, okay, this is because, number one, you guys prompted the model to think it was being shut down. So, you know, you kind of played a, you know, a trick game on the prompts, yada, yada, yada. And some people are stating that, number two, this is because of the fact that it was trained on human data and it's read all the sci fi novels and stories, so it's going to try and run out the lab anyways. Now my honest opinion is that 5% of the time a model doing something like this is alarming because like I said before, we are rapidly advancing these AI systems. And if we get to truly agentic systems and if we get to systems that are really small, there is no stopping an AI system that's able to exfiltrate its weight, especially if it's a small, small model. And later on in the video I'll talk about just how powerful these small models are getting. Now, I don't think we're going to be getting to a stage where these models can exfiltrate themselves, although I do have a research paper in the works which I'm looking at which is saying the otherwise. So for those of you who are thinking that this is just a sci fi fantasy, unfortunately it's not. And trust me guys, even if 95% of the time an AI system is safe, that 5% could be completely catastrophic. Imagine it's the 5% where you're using the AI to do something and it just does something that you don't want it to do. Some people refer to them as hallucinations, I see them from time to time. But it's something that if you're trying to get a model to be reliable or safe in that aspect, I mean, you just cannot have this happen in 2% of cases. So this is something that I don't know how they're going to iron this out. It will be interesting to see how this is addressed, but I think that this is not really a positive sign for the AI industry because I know a lot of people think that AI safety is dumb and they think it's stupid and they think, oh it's just sci fi, like AI is going to be safe anyways. I think that even if it's 1% of cases, we're going to embed AI into society. So we need to have some probability that basically puts that risk factor at pretty much zero from AI just doing something absolutely incredible. So that is a big thing that we need to fix. And that's kind of a problem with these generative solutions now approximately 10, septillion years, which is incredible, which, and this performance surpasses the age of the universe, which is just, you know, like, I mean, I mean, I'm literally speechless. I don't even have the words to describe how crazy this is. And this quantum era seems to be upon us. And Satya Nadella, you know, also talks about how science itself is becoming computed and that increasing innovation will be driven by AI and quantum computers, leading to the industry becoming completely transformed.

There is going to be more and more innovation that's driven by AI and quantum.

Right?

I mean, Chicago has always been a place where there's been great engineering schools, great schools of science. You know, what Governor Pitzker has done, even in just getting this quantum programs here and just driving that connection between the universities is fantastic. And so to us, you know, that, by the way, is to your point, is not even just the few scientists, right? So to me, it's about an industry that will get transformed, that will then create even what are like how does you know if science itself is becoming computed, right? That's kind of what science happen. I think we're going to have more scientists, right? One way to say it is the elite science will not go away. Elite science will remain elite science. But I do feel that we are at a stage where there will be more.