OpenAI Employees LEAVING, SCARY Microsoft AI Product, GPT-5 Updated Date, Stable Diffusion3 And More

So with an absolutely incredible couple of days in artificial intelligence, we're going to be taking a look at some of the most pressing stories that you once again did miss. Because with the release of Llama 3, there was actually a lot of information that many people didn't get to see. So, so one of the most important stories was Mark Zuckerberg actually talking about the future of AI devices. And I think it's kind of interesting if you haven't used his products before. I've used his other devices such as the Ray Ban metaglasses, and they are actually pretty decent in terms of what they're able to do. And combining them with AI in the future is going to be absolutely game changing.

I think for the next version of the glasses, one of the things that I'm pretty excited about, I think we'll start getting some consumer neural interfaces soon. I think that's going to be pretty wild. I'm not talking about something that jacks into your brain. I'm talking about something that you wear on your wrist that can basically read neural signals that your brain sends through your nerves to your hand to, to basically move it in different subtle ways that are maybe not perceptible to people around you, but we're basically able to read those signals and, and be able to use that to control your glasses or other computing devices. And I think that's going to be pretty wild. And you know, we're obviously still at the beginning of that journey because we haven't rolled out the first version of the product. But you know, playing with it internally, it's, it's, it's really cool.

So Meta are actually secretly working on non invasive technologies to, to sort of combine existing technologies with the futuristic technology that you see Elon Musk doing with Neuralink. And it actually is possible there's devices that you can have that can kind of, I guess you could say, combine the strengths of what Elon Musk is doing without the invasiveness. Because I know a lot of people are quite scared to get a neural link implanted into their brain. Whether or not you are super paranoid or whether or not you just think it's not safe, it's completely understandable. I personally don't want a chip planted in my brain. But the point is, is that they are working on this and non invasive means that you don't need to like plug it into your skin. It's literally just like an armband or just like a glasses that you put on and it's able to read the signals and it's going to be something that is really futuristic because I remember I was reading the research about it and the research didn't really seem real. But if you view something like this, it says towards a real time decoding of images from brain activity. And you can see here that, you know, it sounds like futuristic stuff, but you know how when they're like, okay, this AI tool can literally read your brain waves and understand exactly what you're thinking about. This is the kind of research that they are able to do. And you can see that the image shown here and then the decoded output here shown at 1/4 the speed. So I mean, right here you can see that the results aren't that crazy. I mean, of course it's a horse and then it shows like some kind of animal there. I mean, it's not really accurate, but we've known that before when we actually took a look at things like Mid Journey. It kind of, you know, reminds me of like Dall E1 and Dall E2 in terms of how, you know, you know, the image, in terms of the resolution, how the resolution wasn't really good, and how the images had all of these artifacts and things like that. And I'm guessing that in the future this is going to be something that actually is quite effective. So I truly, truly, truly cannot wait to see what kinds of devices that are going to be in the future. Now, I know that some people might look at this technology and think that this is just purely dystopian and the sense that we're going to have some AI overlords just reading our brains with the technologies, and that might be a scenario. But I think the main thing here is that what we are going to do in the future is that we're largely going to have technology that helps the less fortunate. So, for example, some people that might be disabled due to unfortunate circumstances, they're going to be able to still communicate, use technology effectively. And I think that's one of the main use cases that most people don't think about because the average person isn't disabled. But for the 2% to 3% that are, I definitely think there's something that's really important that could actually help them get back on their feet. So that's something that I did see with Neuralink and I think that is really the main use case for these. With the recent release of humanoid robots in 2024 and with of course Atlas's Boston Dynamics, many people are currently speculating on which robot is going to be the one that is the winner takes all. And I don't think it's going to be winner takes all. I think it'll probably be like 50, 50 or like 70, 30 in terms of there's a high end one, there's one that everyone uses and the rest are just somewhere in between. Kind of like how we have phones. But I think it's going to be interesting because you know, we have Tesla Gen 2, we have H1 unitry and we have Atlas, Boston Dynamics. I think overall what we will have, this is just my personal opinion. I think the Boston Dynamics is going to be the Apple and I know Apple is working on one, but they're kind of so far behind in terms of, you know, not building that they've been building a car for the past 10 years. So I think that it's going to be pretty hard to catch up to where these companies are. I think the main ones to watch are of course Boston Dynamics. I think they are going to completely lead the way in terms of what they've been doing because they've already, you've already seen how Atlas has been like three to four years ago, how it was able to jump over different things and able to, you know, run around. And so I think that, you know, Atlas has already shown what it could do. The previous version, of course, the retired version on the left, which was absolutely incredible. And I think if Boston Dynamics combine that with whatever AI company they choose to partner with, it could be OpenAI, could be anthropic, could be Google. I mean, you know, I feel like Google might want at least one robot to be able to partner with. I think it's going to be light years ahead of any of these because just what we've seen before from Atlas, if that's combined with any of the dexterity to be able to grab objects, catch objects, I mean there's going to be literally something out of a sci fi movie provided that this doesn't turn red. So I think this is of course going to be number one and then second, I think it's going to be between figure and Tesla. And the reason I think it's going to be between those two is because if we take a look at what Tesla's been able to do, Tesla's moved pretty quickly and they do have the infrastructure. They do have, of course, Elon Musk, a huge team, they've got billions of dollars and of course we've got figure which have shown us that literally in 18 months they've done something absolutely insane. So I think if we're betting on, you know, the partnerships and how fast figure have moved. Considering the fact that they've got tons and tons of some of the best know AI researchers in terms of, you know, robotics as well, I think that they have a very good chance at competing with a large lab like Tesla. But I'm never going to bet against Elon Musk because a determined man is something that I've always seen to be able to achieve incredible things. So I think those two will be second in line to take the cake in terms of what is possible because Elon Musk did say bring it on. And I think in terms of what will be the cheapest, most available, I think it will be the unitree because these guys are moving pretty quickly under the radar and they're achieving quite a lot. So I think further versions of this model are going to be pretty insane. Then we had something that is quite concerning to say the least. So, you know, OpenAI have been losing a lot of talent recently and you know, I guess, I guess the negative spreads a lot more than the positive because they have been on quite a hiring spree. But I think this is something important to, you know, show you all. I mean, I was going to make a video about this, but you know, just for time constraints, I just rather just add this into this video. But essentially Daniel Coco Talgio of their governance team quits due to losing confidence that it would behave responsibly around the time of AGI. And he wrote some very interesting things that I did do a video on last year that were actually quite interesting. I think, in fact, I think it was a couple of months ago, but just feels like, you know, the rate that AI is moving. It was last year, but essentially he quit because he doesn't believe that they're going to act correctly. Now basically the reason this is so concerning is because you have to understand that AGI is going to be a technology which fundamentally changes how we operate as a society. And because of that, if OpenAI don't do what they said their mission statement is, and if they decide to change, things could get ugly pretty quickly. Because AGI is, you know, not far from ASI in the sense that you can use AGI, a true AGI, to get to asi, although there's different scales, you could use an AGI to get to asi and if they use AGI to get to ASI pretty quickly because they have no, you know, I guess you could say obligation to tell us if they've achieved AGI and if they're scaling to ASI. I mean, if I was OpenAI, I would just scale to ASI anyways, because what would be the point? And I think it's going to be pretty fascinating because I mean, as he said, okay, and this is not me saying this, this is as he said, it's that like, if you get to artificial super intelligence, you essentially have godlike powers over the people that don't. And basically think about it like this, okay? If we brought our technology back to the 1800s, it would seem like everything that we're doing on a day to day basis is magic. I mean, we communicate through the Internet. You're watching this video through a small device that's connected across millions and millions of miles. You can have an instant connection with anyone anywhere on the face of the earth, which is pretty, pretty remarkable. And we have weapons and weaponry and armories and just crazy things that, you know, just simply aren't even invented yet. The people would be like, wow, okay, that's magic. Like, imagine showing someone from that age a drone show. Like, I mean, they would, they would freak out. So, I mean, it's pretty, pretty crazy. And for those of you who think, you know, I'm just exaggerating on the claims, these are some of the claims that he made around a couple of months ago. And he said there will be AGI soon, any year now. Whoever controls AGI will be use it to get to ASI shortly thereafter, maybe in another, give or take a year. And remember, this is the guy that's working on governance or OpenAI or futures. And I think, just look at these statements because you think I'm exaggerating. You think I'm making these bold claims like I'm just someone hyping up AI. This is not me, this is someone that actually works with OpenAI. And he says probably whoever controls ASI will have access to a spread of powerful skills and abilities and will be able to build and wield technologies that seem like magic to us, just as modern tech would seem like magic to medievals. And this will probably give them godlike powers over those who don't control artificial superintelligence. And that's the point I've always made. Okay, if OpenAI have control of a super intelligence, I mean, doesn't that just instantly make them the most powerful company on the face of the earth? And doesn't it mean that they pretty much can't lose? Because if you have someone that's magnitude or a system that's magnitude level smaller than us, I mean, the sheer amount of capabilities that it could probably output just far surpass anything that we currently have, because it's able to evolve at a rate that we can't even perceive and they're able to just move faster than any standard human company could even dream to. So I think that that is of course going to be OpenAI's goal, because once you have that, I mean, you have everything. And I think the point is that how do we trust a company to, you know, act correctly? I mean, even if OpenAI decided that they weren't acting correctly, the super intelligence could just tell them, provided they did have control of it, and provided we're all still alive, a super intelligent AI system could in theory protect them from any harmful doing and just make them retain power. If they were like, okay, we want to become the most valuable company on the planet, we want to retain power in this country, in that country, they could instantly do that with a click and nobody would pretty much be able to stop them. So, I mean, it's kind of like this, you know, I guess we're just trusting OpenAI to do the right thing, because once they have AGI, ASI, I mean, you know, the cat's out of the bag. It's not like we can stop them even if we wanted to. I mean, what are laws and regulations if you have something that could do anything you'd want it to. And I mean, I know it sounds pretty crazy, but, you know, it's crazy until it's not, you know, GPT4 is crazy until it's not. Image creation was crazy until it's not. Video to, you know, image to video, whatever with Sora was crazy until it's not. So you can see, you know, currently no one knows how to control asi. If one of our training runs work out way better than we expect, we'd have a rogue ASI on our hands. But yeah, there's a lot of crazy, crazy stuff that is very, very interesting from his, you know, statements that were from a couple of months ago and the fact that now he's quit because he's like, you know what? I don't know if OpenAI is going to take this seriously. I think it does say a lot about the state of OpenAI. And I can honestly say, as someone who doesn't work at OpenAI and has no ties to OpenAI, I can just say that I genuinely hope that for the sake of humanity they do the right thing. I don't know. That's a crazy statement. Just hoping that a company does the right thing. Because big corporations usually never do. They do what's in the interest of finances and money and big companies are usually greedy. But, you know, I'm just guessing that I hope they do the right thing because, you know, what other shot do we have? Now, senior AI researcher at Nvidia and the lead of Nvidia's Gear team creating foundational models in humanoid robots and in gaming is stating something rather interesting. He stated that humanoid robots will exceed the supply of iPhones in the next decade gradually, then suddenly. And this is from someone's Twitter thread where they were explaining a few predictions for the future and they spoke about how labor will be near free. We're going to have a billion and other robots freely freeing humans from the servitude of undesirable jobs. And I think this is going to be interesting because, you know, on one hand we have, you know, the fact that, you know, humanoid robots are going to be able to be building quite a lot of different things, and on the other hand, some people think it's going to take your jobs. I think it's going to be a mix of both, depending on how economically viable these robots are. But I do think there will be industries where these kind of robots are just going to be doing just ridiculously interesting things. And Elon Musk also responded saying he agrees. And I wonder how that will change the future in the sense that, you know, we're going to be getting a new class of, I wouldn't say citizen, but I think it blurs the lines a little bit because a robot's going to have rights, so they're just going to be purely employees. But considering the fact that in order to get robots right, they're going to need to have some kind of AGI level system controlling them, I mean, you know, that kind of begs the question, are we going to have, you know, this new class of population just being everywhere? I mean, I don't know. It's going to be interesting. But yeah, whichever form factor these robots do take and whichever company is at the front of that, that one is also going to be something interesting, just like we spoke about before. Now, something that I actually did miss from Meta's release was the fact that their new AI tool, there's actually a secret feature. So for some reason, this, the only reason I miss this is because it doesn't actually happen to every version. It's kind of like, I guess you could say, being rolled out in alpha to variety of different accounts. There's no real way to get this, but essentially you can click, create an image of something and then as you're typing, it kind of changes the image. Now, I do know from using a vast different amount of AI tools that this is actually something that you can do with many different AI tools. But I think it's kind of interesting to see that meta straight out of the gate. They're coming out with not just a image creator, not just an image, you know, to video creator, but they're also coming out with something that animates the text as you write it. So, I mean, from what I've seen so far, a lot of people actually do like this technology. So I'm guessing that this is going to be a hit and maybe chat GPT, OpenAI are going to include this in the next product release. But I think this might be the future as well, because, you know, Iman Mustach was actually talking about how we've moved to a stage now where we can literally kick generate and we get thousands and thousands of image. And I've used a variety of, honestly like hundreds of different AI platforms where you can literally do this now. So I think this will be the future of image generation where we literally get so many iterations that pretty much going to be spoiled for choice. So that was something that I did miss from the tutorial I uploaded yesterday.

You know what I decided to do? I decided to focus all my attention, all my time on listening. So instead of doing something else, I just listened, listened and listened. Because I'm a true believer that if you're really bad at something like listening, for example, it only shows you that, hey, you have to practice listening as much as you can. We introduce vasa, a framework for generating lifelike talking faces with appealing visual effective skills. Given a single static image and a speech audio clip, our model is capable of not only producing lip movements that are exquisitely synchronized with the audio, but also capturing a large spectrum of facial nuances and natural head motions that contribute to the perception of authenticity and liveliness. The core innovations include a holistic facial dynamics and head movement generation model that works in a face latent space and the development of such an expressive and disentangled face latent space using videos.

So that was Microsoft's Vasa one. And it's pretty terrifying for the average person because it's like, okay, why on earth did you guys create this? Because as far as we know, this is literally just a deep fake tool, something that's been plaguing the Internet for quite some time. And even if you don't release this, a Lot of people might be able to even learn from the research papers and then subsequently create their own. Now I do want to state a disclaimer. Microsoft did state that in the conclusion of this paper that they are not going to release this at all. Because of course there are many, many concerns about widespread usage of this for of course deepfakes and many other issues, such as political ones and just many other issues. But the point is, is that this does show how crazy deepfake technology is. Getting to the point where you could literally just upload one image, you could rotate that image freely as a video and then you could instantly manage to clone someone's voice using, you know, many open source softwares and then of course have that person speaking and, and stating anything. And the crazy thing is is that from the vast amount of demos that I've seen, this actually does look pretty realistic. And it actually reminds me of the thing that we got last month where there was a similar software to this, where people were also quite afraid. Now I think this kind of technology, like I said before, I don't know why you developed this. I mean, I think the only application as pointed out by AI explained was the fact that you could use this for virtual tutors. Maybe that's the only kind of thing. Maybe for virtual tutors, maybe for virtual AI assistants. But for the average person, they should definitely not open source this, they should keep this behind a lock and key. And I do know that yes, open source probably could catch up in a couple of months time because usually that is, you know, how things work. But I do think that, you know, maybe for AI doctors, maybe for AI personal trainers, that's what this purpose kind of serves. Maybe that's what you want. But for anything else, I just think that this is, is actually a net negative, despite it being actually kind of cool. Now this is something that actually did speak out in the things that we're going to see in 2024 for AI and I'm not surprised to see this, but the US Air Force says an AI controlled F16 fighter jet has been dog fighting with humans and the plane was made to restrain itself so as not to harm the pilot or airframe. Now if you don't know much about jets or airplanes, this is a big deal. This is a really, really big deal because this is really going to change warfare. We've already seen how drones have been changing the situation in Ukraine and how it could potentially change things in the future. And the reason that this is going to completely change the entire game is because There's a significant human limitation when you are flying a plane. If you don't know when you're flying a plane, you don't just fly a plane. You're actually subject to G forces. And when the plane turns, these G forces increase, which means it strains the human body. Now, usually what happens is humans are put under so much stress that sometimes they can have something which is called G lock, and that occurs when just the blood just literally leaves your brain and you pretty much just pass out because there's too much G forces and you literally can't take the force. Now that's a problem because of course, pilots don't want to pass out whilst flying the plane. It could lead to them obviously crashing. And this is where AI comes in and solves the problem. Because if an AI can take G forces that humans simply can't, it means it's literally going to be able to execute maneuvers that humans won't be able to. And if there is a fleet of highly advanced F16AI fighter jets, they simply will not lose. And I think that's going to be interesting because if the US manages to do this first, we're going to see other countries also do this too, because there's going to be literally no way to compete. Because I promise you, the kind of turns that humans can make, there's a physical limit even the most seasoned veterans can't go. I think it's beyond around 9 to 10 GS. At that point, you know, things start to get really, really problematic. And there's even some allegedly long term health implications of consistently doing this. And there are some, several studies that state that this is not good for long term health. So I genuinely think that this is going to be the future of the skies. Some people are stating that this is awful, but I can't see why they would not do this. Because I mean, you could train it in a simulation software a million times and it could have everything figured out. It could understand every single scenario. It could never lose a dog fight. It could, it could pretty much prevent anything from ever happening. And you know, standard pilots might just not be able to deal with these kind of situations. So, I mean, it's very, very interesting that, you know, they're experimenting with this. I'm not sure why they even told us this, but you can see right here, it says that AI software won that competition, but had an edge. It was allowed to fly at speeds that would have overstressed a real aircraft and generated G forces that would harm a human pilot. So I mean, there's a real edge here. And I, I think, I don't know, I think in the future this is going to be the future of the skies. So I mean, of course there's going to be, you know, use cases where you actually do want a human in a plane because there are just some cases where you absolutely do. But I think that this shows how across the board AI is literally advancing, you know, and like I said before, it's going to literally change everything. Now here's a little thought experiment for you. This is grok and if you don't know what GROQ is, it's basically an inference provider for large language models and it's able to, to do that really, really with just insane speed. So this is llama 3 being served by Grok Labs. And we can see that as soon as the person submits a question, you can see how quick the answer is done. Okay. It's literally at 971 milliseconds, which is absolutely insane. Now here's what I want you to think about. Think about in the future how we're going to have multi agent scenarios and just think about how if we had like a bunch of different AI agents, let's say you had a thousand AI agents for your company and they are all collaborating, working, refining, checking and doing all of that at this speed together just a million times faster than we can even imagine because it's likely going to get faster than this. So imagine this is what a thousand agents are doing in a company 24, seven consistently like that is going to be absolutely incredible. Now of course the costs are going to be high. You know, inference is not cheap. But the point is, is that this should give you a glimpse into the future of how agents are truly going to shape things. And now lastly, we have some news which is apparently OpenAI's release is set for Monday. Some people said they had it pegged for June, but Mark Zuckerberg has accelerated the timeline because Llama 3, the 70 billion model, is really insane. Like really, really good. And of course it is Sam Altman's birthday on Monday. Now that is sheer speculation. Like sheer, sheer, sheer speculation. But at the same time we never know. I mean, I think GPT5 is going to be dropped completely randomly. I still think it's going to be summer because I still think safety testing agents and all those things are going to take quite some time to iron out. But maybe just keep your eyes out for something on Monday. And if there was anything I missed, don't forget to Thing that I actually did miss was stability AI. There was stable diffusion 3. I do apologize if I'm repeating this again, but I feel like I missed this. So you can see right here that this is their new text to image model. This is a red sofa on top of a white building. Graffiti with text that says best new in the city with prompt portrait photograph of an anthropomorphic tortoise seated on a New York City subway train. This is another prompt here. And this is another prompt here. Now, a lot of people have been criticizing them for releasing this under a closed source, I guess you could say thing, because they are trying to do memberships to access the thing. And this is something that you guys need to know about. Stable diffusion is that they're currently struggling to generate revenue. The company's been going through some pretty shaky and tumultuous times. So I wouldn't, you know, you know, knock on them for trying to do this because this is probably one of their paths to sustainability. And we've seen before that sometimes when companies, you know, just run out of funding or run out of money, everybody loses because the company just goes up. I mean, you know, the product disappears, Everything disappears. Nobody wins. So, I mean, I would try and support this company if you can. I'm not saying to pay for it, but I'm saying if you're someone that uses it, okay, you might as well try and pay for it because, you know, maybe the company might not be around for much longer. So that was something that I did want to end on.