Googles $100 BILLION A.I Masterplan, Secret HUMANOID Robot, Yann Lecun On AGI, Grok 1.5 AI

So, with an absolutely crazy day in AI, let me show you all the crazy stories that you may have missed. Coming in at number one, there has actually been a new AI tool that has somehow broken the mold in terms of getting out into the mainstream. So this is Viggle AI. It's basically a tool that you can use to instantly generate motion of characters, and it's been used to generate this recurring clip that you've pretty much seen everywhere. If you haven't, it's completely fine. It's. It just means that you actually go outside and have a normal life. But for those of you that have been spending way too much time online, you've seen this original clip versus the Viggle AI being used to generate various different social media personality styles, different characters and celebrities doing this same thing. And it's been memed millions and millions of times. And I got to be honest, it is pretty, pretty funny. But it does go to show that AI technology is continually breaking the mold. Now, what's crazy is that this technology doesn't even seem pretty good, but you can see across the board, things like these are getting millions and millions of views. And I've seen a variety of these different memes in many different categories, which is pretty, pretty hilarious if I got to say so myself. So I think once these memes even increase further, I think it's going to be something that does show us how crazy AI is, because content creation with AI is, is going pretty, pretty crazy. And, you know, content creation and mean. And of course, AI generated media is going to be something that really does get affected by AI. And you can see there are a bunch of different tweets about this. And this one, you probably only get this if you've been paying attention to the OpenAI drama. We are above them, below them, around them. And this is in reference to Satya Nadella, who basically saying that I wouldn't say he owns OpenAI, but that they pretty much have all the IP and the infrastructure. If OpenAI went away tomorrow, they'd still be completely fine. And this is him dancing to that. Of course, it's not actually him. Just so you guys know, it's not actually him. This is done within a couple of seconds. It's basically like mid journey, but for motion. And you can easily do it in discord. I've actually posted a tutorial on this on the second channel, so don't forget to check that out if you wanted a quick tutorial on this. Now, here's something I didn't really expect, but there was A new humanoid robot startup that actually just exited Stealth says after two years in stealth mode, we're thrilled to unveil Menti Robotics and our humanoid robot Mentibot. With AI integration at every layer, from simulation to real machine learning to NERF based algorithms and LLMs, we've achieved a complete end to end cycle for tasks. This is just the beginning and we can't wait to share more. So you can see that this is pretty incredible. I'm going to let the demo play out first time, but like I said before, if the audio does get copyrighted, this video will be played again. And I'll actually show you guys what this is doing, but here's the quick demo. So yeah, take a look at this. So I'm going to show you the first five seconds of this because the audio is just being weird, but take a look at the first five seconds.

Hey Menti, please go to the table in the kitchen and wait for instructions. Okay, I will go to the table in the kitchen.

And then you can see that this robot does have an internal monologue. Now I do want to say quickly that this was probably the worst day of the year that you could have announced your humanoid robot, because we all know that Boston Dynamics recently did just release their second iteration, Atlas 2.0, which is pretty, pretty insane at the moment. And whilst this is, you know, pretty good, whilst I'll explain to you exactly what's going on in the video, I think it's important to time your releases because we do know that sometimes AI can steal the spotlight, especially if there is a certain robot release or if there is a certain new software technology or a new GPT that can completely steal the limelight. So essentially we can see that this robot, Mint Tebot, actually has an internal monologue. We can see that it thinks and it says, the user requested me to go to the table in the kitchen first I need to find it in the map and then navigate to it. And what we can also see on the left hand side is we can see that it does have a map with I'm guessing pre recorded locations that it already has the navigation area too. So you can see right here it says locate table and kitchen. The result is this position and then we have to navigate to this position and then it walks over to this area. And this is by far one of the most unique designs I've seen in humanoid robotics. And I'm just going to rewind this back here because I would say that this looks remarkably strangely and in an uncanny way, kind of human it doesn't give me robot vibes. It looks like a human who's just stumbled out of bed walking towards the table in order to do something. And then you can see it manages to come over to the table where it looks. And you can see right here that it does have the kind of vision system. There are two or four cameras that are for sensing. I'm not entirely sure on what they are in terms of if it's lidar or what kind of technology it is. I'm guaranteeing they're probably going to release more information. But you can see here, once the robot arrives at the location, it says, think, I have arrived to the cable. I have arrived to the table in the kitchen and should wait for further instructions. I have arrived. What should I do now?

Please put the fruits in the container and place it on the counter behind here.

So then we can see the robot was requested to put the fruits in the container and take it to the counter. So we can now see that the robot is basically saying sure thing, which means this robot does have a voice, which is pretty cool. We know that that is a new addition robots are currently getting. And then we can see this robot uses its hands to grab the fruits and then place them in the container. And what's really interesting here is I like the fact that we're able to see exactly what the robot is thinking, because this allows us to understand how the robot is making its decisions and how it is planning and executing its journeys in real time, which is, of course, pretty useful if you want to realize where the robot went wrong and why certain tasks failed. Then we can see here, it says, all fruits are in the container now I need to lift said container. And it's lifting. It manages to push it together, lift up the container, and it says, now I am holding the container. I now need to locate the kitchen counter and, and navigate to it. You can see locate kitchen counter, result position, navigate five. And it manages to, I wouldn't say stumble, but definitely walk over in an interesting fashion over to this area. And it manages to place this down on the counter. And I would say that, you know, the statement that robotics is truly, truly coming alive has never been more real. From what we've seen today from Atlas 2.0 and from this Menti bot here, it seems that robotics is undergoing its ChatGPT moment where we're seeing that robotics is moving at light speed. Now, whilst that demo might not seem that impressive to some of you, I think for a humanoid startup, that this is rather impressive for their first demo of their humanoid robot. And something that I actually really do like about the way that this robot walks is that it does walk in a human kind of way. I don't know why I actually like that, but I think that some robots, being uncanny as they are, from what we've seen from Atlas earlier today, where it stood up on its legs and just twisted around in its body in this supernatural, super unnatural way, it was just absolutely incredible. I think a robot just lazily walking around, you know, doing tasks around the house like this could be a lot more enjoyable to look at. So now if you go over to their website, they actually do show you a little bit more details. And honestly, I don't know, maybe I'm just someone that's too interested in this stuff, but I just find this stuff, honestly, honestly fascinating because it's literally the next level of evolution in terms of what humans are going to be interacting with next. And you can see it's a personalized AR robot that you can mentor. So it says we are on the cusp of a converge of computer vision, natural language understanding, strong and detailed simulators and methodologies for transferring from simulation to the real world. Then you can see right here it says AI First, AI is embedded at every level, enabling dynamic interaction with the world. Mentibo integrates cutting edge SIM to real for lifelike gait and hand movement, nerf based real time 3D mapping and localization, dynamic navigation in complex environments and large language models to build a cognitive map of the world and advanced task execution. And down here, this was actually something I actually did find to be one of the most fascinating things. And seemingly a recurring theme of humanoid robots was that the arms might actually be interchangeable from this version to the one that we actually see up here, not this one, but the one that we got to see on the demo. We can see that there are different versions of hands as grippers. And it seems like I said before that these robotics companies are experimenting with whatever is going to be the easiest type of hands. And I think we've seen time and time again that, you know, these style of hands aren't always needed. The two pronged approach can actually work in several applications. But they say that the arms and hands present a full range of motion and enough accuracy to perform delicate tasks like grabbing some kitchen towels or some paper towels, or grabbing a pole, or grabbing a paper plate, or grabbing whatever it is that you need. So I do think that this kind of robot is definitely going to be quite fascinating to see where they take this in terms of the development, because, you know, when you have all the competition in the world, From Atlas, Tesla, OpenAI and figure to 1x Robotics, different robots are going to be fitting in different parts. And robot doesn't seem that expensive. I mean, it looks like they're running some lightweight materials. You can see that there's a lot of gaps in the legs there. So I'm wondering if this robot is going to be something that is largely deployed on many other systems, because this is a robot that seems that it could be built by relatively smaller teams and then tested upon and possibly built upon. So that is going to be something that's rather fascinating. And here we can see adjusting a gate to handle the weight and it automatically adjusts its gate when handling heavy load and as humans would do in such situations. So, yeah, I'm excited for this because it's definitely another company that stepped into the humanoid robot race and the more the merrier. Now Microsoft's research Chris Bishop said when AI models regurgitate information in response to prompts, we call them stochastic parrots. But when humans do it, we give them university degrees. And this is something that I've spoken about for quite some time. It's a sense that, like, we have a lot more in common with AI than we think. For example, when AIs hallucinate, which is essentially where they make up something or they simply forget something, humans do that all the time. When we just say, oh, LLMs are just regurgitating information. Humans literally do that all the time when humans sometimes mistake one thing for another and LLMs do that. It's like humans literally do that all the time. So these systems, I do think that there's a lot more similarities than people are giving them credit for. And of course these systems are going to get way smaller, surpassed us anyways in the near future. But I do think it's an interesting topic to discuss because it's like, you know, how can we be so much better than these robots if we pretty much fundamentally doing something that's kind of on the same level have to these.

Models is a little bit of that sense of that threat to the specialness that we feel as humans? I may be completely wrong. This is purely speculation. But, you know, it's interesting that we talk about people use phrases like stochastic paradigm. It's just regurgitating stuff that it's seen before, some people claim, or, you know, of course it hallucinates. Sometimes it comes up with stuff that's just wrong or doesn't make Sense. But think about the following. Imagine there was a very, very smart physics student, went to, you went to a top university, worked really hard for four years. What would they do? They would, they would read books, they would read papers, listen to lectures, have discussions with their professors and with other students and then they sit their final exam and they get 95% in their final exam and they come top of the year. Which is, we don't say, huh? Well, 95% of the time they're a stochastic parrot regurgitating Einstein and Maxwell. And the other 5% of the time they're hallucinating. No, we say congratulations, you have a first class honors degree, you've graduated with honours. This is a, you know, a wonderful achievement. So it's interesting that we do seem to view the capabilities of neural nets with almost a different ruler to that of humans. And while nobody's suggesting that current models are anywhere close to humans on many axes of intelligence, nevertheless I see the first sparks of artificial intelligence.

So yeah, the specialness of humans may not be as special as we think. Now this is a bit of niche news. There is something that is pretty cool here and I only used to cover this quite when I first started the channel because I just thought it was pretty cool. And I remember that people were stating that this is never gonna get good, but seeing this actually get good just puts a smile on my face. It's just, it's just really interesting because a lot of people were saying, haha, it's never going to get good at, you know, making 3D models. But right now instant Mesh from Tencent is incredible because it's super Fun, fast images, 3D with very, very high quality output. And this is something that you can completely test. And like I said before, this shows us where we're headed because previously these models weren't that good. Like they were okay, they were all right. But this is something that you can actually test on how quickly we're going to be able to get super lifelike 3D models just from a standard image. And this stuff, you know, like what you're seeing right now isn't going to really change anything at the moment, but soon it's going to be embedded in applications many different things and you're literally going to have a complete generative AI pipeline where you can just think of something and it's going to be like immediately animated. So these are the kind of things that I think is still important to include. And then we had a rather fascinating statement from Yann Lecun. Now Yann Lecun, definitely one of the smartest people in artificial intelligence, but he's constantly making, I would say he's making headlines for this, but I would say he's constantly making at the center of certain debates because he is someone that holds a very, I would say, conservative view of what AI is about to do. Whilst, yes, AI is exponential, he holds the view that some would argue is a bit more realistic. For example, right down here, you can see that he said that he's quite confident about the payoff. Eventually machines will surpass human intelligence. But it's going to take a while though, he said it's not just around the corner and it's certainly not next year. We're like Elon has said, so we know that Yann Lecun has this conservative view of AI for quite some time now. And he has a different view of AGI than most people. He still thinks that, you know, we're on a different path. He doesn't think that LLMs can really get us to AGI. And, you know, I think the thing with this is that right now there are two sides. Some people are saying, you know, he's completely wrong, he doesn't know, yada, yada, yada. And then the other side is saying that he's completely right. I think either way, we're going to see in the future who is right. Maybe we need a different architecture. That's what some people are arguing. Some people are arguing that we need a different architecture to get to AGI. However, we have achieved a lot with scale. So, I mean, you know, when you have someone that is a winner of the prestigious Turing Award and one of the three godfathers of AI, it's pretty, pretty fascinating on how the industry is still divided at the very top level, where you would at least think there should be more coherence than there currently is. And he talks here, which is, you know, even more fascinating. He says we're easily fooled into thinking they're intelligent because of their fluency with language, but really their understanding of the reality is very superficial. They're useful, there's no question about that. But on the path towards human level intelligence, an LLM is basically an off ramp, a distraction, a dead end. And I wonder if that is going to be truer in the future or if that statement is going to come back to not be true. Either way, I think this is quite fascinating because it seems that we're starting to learn more and more and more. And as more research gets produced and as more research papers do come to light. I think once we're able to get more interpretability research into understanding exactly what these systems are doing, then eventually we could truly understand how smart these systems are. And we're going to really, really know what these systems are capable of. Because right now we're currently in a black box with these systems. And this is part of the problem, because some people state that, you know, the way how LLMs are designed, it's just predicting the next token, which is technically true, whereas some would argue that even if it is predicting the next token, it's still ridiculously smart and this is just a different form of intelligence. But either way, I think the future is going to be interesting because like I said before, some people are going to be proven true and then others aren't. And I think it's going to be interesting to see how they react to that. In addition, we also had Elon Musk release Grok 1.5 Vision Preview. I was genuinely supposed to get to this, but I was just busy on the day and it just slipped by me. But this is something that I think is bigger than a lot of people think. I think that Elon Musk is trying to really, really beef up the GROK architecture because I'm sure he realizes that there is a lot of application for, for building AGI. And with how OpenAI is, I'm sure that, you know, AGI might be restricted in terms of compute. So by building Grok 1.5 vision based on, you know, the previous research that has been done over the years, I think we're definitely going to be benefiting from this system. Now, I do want to have one quick thing to say before I get into why this is so great. If you're watching this, Elon Musk, and I know he probably isn't, can you make it available for people in the eu? Because for whatever reason, I haven't still got access to Groq and I would love to test the system, but for now, it's just under lock and key, it seems. So you can see the benchmarks. Grok 1.5 vision compared to GPT4V. It is currently surpassing it nearly on all the benchmarks, which is pretty insane. And if it doesn't pass it on some benchmarks, it's definitely very, very close. And this is the fact that, like I said before, X AI was a company that was only founded around 18 to 12 months ago, which means that they've come a lightning, very, very quick way in terms of the time that they've been there. I think that they've moved so quickly for the amount of time that they've been there. And I think this is definitely going to be a company that I'm going to be watching along because these kind of companies are ones that are super innovative. And if there's anything, you know, about company culture and teams, it's that companies that move this quickly usually really disrupt the industry. And that's because if you, you know, a company that has like 25 people, but they are all like SEAL Team 6 members in the sense that they don't waste any time, they wake up, they get to work, they are on no nonsense. They just are very, very effective at what they do. They can do a lot. And I think that's exactly what we're seeing here with X AI because they've achieved so much in such a short space of time that I think they don't get enough credit. And I think that's just because, you know, these systems aren't for worldwide access, so people don't really get to use them that much. But. But I think that in the next 12 to 18 months, this company is definitely going to be producing some really, really interesting stuff to the point where we're going to be surprised by their capabilities because they've done a lot with a lot less funding and a much, much smaller team, which is why they're being able to move so far. So it will be interesting to see how this team expands over the future and how they've managed to tackle certain problems. But Grok 1.5 vision is nearly just as effective as GPT4 vision, if not on par or above that. And we can six to seven different samples of IT coding from images and being able to identify certain pieces of information, which is rather fascinating. Then of course, we had DeepMind. CEO says Google will spend more than $100 billion on AI on developing AI technology, another sign of investing in the arms race that has gripped Silicon Valley. Now, I'm going to stop this article right there because I know why this value is so crazy is because if you remember recently there was another company that. Well, two companies that actually discussed the same amount. Okay. And if you haven't been paying attention, here it is. It's Microsoft and OpenAI's 100 billion Stargate AI supercomputer. So it seems that maybe Google saw this and were like, Whoa, what is OpenAI and Microsoft up to? Maybe we got to match this with the same level of compute. And here you can see that they've, you know, committed to $100 billion. And that is a serious, serious commitment. $100 million is no laughing matter. So whatever Microsoft and OpenAI are building, it seems that Google, of course want to be able to partake in that race. Now you can see here it says Google's DeepMind CEO Demis Hassabis was asked at a TED conference in Vancouver on Monday about a potential $100 billion supercomputer called Stargate planned by OpenAI and Microsoft. And he says, we don't talk about our specific numbers, but we are going to be spending more on that. And he says Alphabet Is Inc. Has superior computing power to the rivals, including Microsoft. And this goes to show that there is no jokes going on when we are looking at these companies who are going to be spending literally hundreds and billions of dollars in the next couple of years in order to compete in the artificial general intelligence race. And it really is a race because as I've said before, this is something that once you do get to AGI, I wouldn't say all better off, but you've pretty much won the game in terms of what you're able to do. But because if it's a real AGI system and you're able to ask it certain questions, then you're pretty much good because you're going to be able to ask that AGI system and kind of recursively self improve in that aspect.