ALL MAJOR AI + Robotics NEWS April 2024 (All AI and Robotics Updates)

April 2024 was a massive month in terms of Robotics in terms of AI releases so this video is going to show you guys every single major AI development and Robotics development that happened in April of 2024 so this is a long one time Stamps will be in the description but don't forget to stick around because there might just be something that you did Miss so that was the rather impressive demo that has been going viral recently and I think it's pretty surprising on how insane that robot is now there's a lot of stuff that I do need to tell you because um it isn't just this demo there's actually a decent amount of information that they've released with this robot and I think we are about to see how crazy humanoid robots are about to get so if you don't know this robot is the I guess you could say the Next Generation in terms of what's to come after they retired the initial version which is of course the atlas now I think it is a little bit confusing that they just didn't call this Atlas V2 or Atlas Pro or something different because their previous robot a couple of days ago they announced that they were going to be retiring this version you remember this version of Atlas the one that broke the internet in terms of being completely viral because it was a humanoid robot that moved as well as any human or even better than Some Humans could in very hard tasks and this was something that definitely broke the internet the videos have over hundreds and hundreds of millions of views and it seems that now since they've decided to retire this role but they're going to be moving on into the humanoid robot space now with this demo that we saw we've got to be honest this is very impressive for a variety of different reasons the fluidity of the robot the uncanniness of the movement I mean just how the robot was able to stand up right there is definitely something that you probably might see out of a Terminator movie and then being able to just swivel on the joints right there shows us that this is a robot that can move very fluidly now they also included some pieces of information with this robot and there's a few key features that I do want to show you that you might have actually missed so with the blog post and the video they said that this is the latest iteration of the atlas robot that builds on a long history of innovation and research and design pushing the limits of whole body mobility and B manual manipulation from Petman testing protective clothing to the recently retired HD Atlas performing parkour we have spent over a decade moving the state-of-the-art forward with humanoid robotics now of course I've already shown you the video in which they talk about the evolution of Atlas and how they managed to get where they are today but one thing I want you all to remember is that this version of Atlas the first video where it was doing the crazy movement wasn't even recently these kind of videos and demos have actually been around for quite some time and we've even saw the recent parkour video it's essentially 2 years old so this robot is likely probably a few years in the making and isn't just something that they've decided to come out with they've certainly been working on this for a very long time now something that we always get in the robotics Community is why on Earth do they make robots that are in the humanoid form factor so you can see right here that they say traditionally we are focused on legged robots because we wanted to build robots that could balance and move dynamically robots that could navigate unstructured unknown or antagonistic Terrain with ease and the main factor here the main reason that most people don't seem to think about because most people decide to question why these robots are humanoid because some people just think it just because they look cool that's not the reason is it's because the world is designed for human so if you're going to design a robot you need to be able that can be plug andplay with our already established environments and one that works effectively I mean yes you could build robot dogs which is essentially what they've used to be focusing on but if you build a humanoid robot and it works very effectively you don't have to redesign the world around the robot you just have to design the robot around the existing world so robots with wheels don't really work well because of stairs and robots that can't walk upstairs don't work well in certain environments and you can see here that they said the humanoid form factor is useful design for robots working in a world designed for people so that's something that you should know now in addition in the blog post they State some more interesting things they state that this is going to exceed human capabilities which is a little unsettling for those of you who may have watched too many Terminator movies so it says however that form factor doesn't limit our vision of how a bipedal robot can move what tools it needs to succeed and how it can help people accomplish more we designed the electric version of Atlas to be stronger more dexterous and more agile Atlas May resemble a human form factor but we are equipping the robot to move in the most efficient way possible to complete a task rather than being strained by a human range of motion and that's why I have the picture of the robot here in the bottom left hand corner because we all saw exactly how that robot stood up it was probably the most uncanny thing I've ever seen a robot being able to swivel on its hips and just standing up like something out of a horror movie definitely interesting and I'm wondering how it's going to complete certain tasks when it's able to move and swivel its hips in a way that humans simply can't and it says here that Atlas will move in ways that exceed human capabilities combining Decades of practical experience with first principles thinking we are confident in our ability to deliver a robot uniquely capable of tackling dull dirty and dangerous tasks in real applications and I think this is rather interesting because like I said I can already see some applications for this kind of robot because what we have here is a robot that is not limited to the human ranges of motion and this is something that I think we are going to probably see in the new robots being developed as far as I know this is the only robot that is able to move so fluidly and so freely you can see the head can do a full 360 the body of 362 the legs as well it's absolutely incredible the types of movement and the ranges of movement that this robot is going to be able to do it's going to truly be an incredibly agile robot now something that you may have missed and there's a few more factors that I do want to show you is that one of the things I noticed was that this robot actually looks pretty far and I think it's so fascinating to how quickly this robot is and this is actually no surprise because if you remember the previous version of Atlas it was very very quick now I think this robot might actually set the world record for the world's fastest moving robot because whilst it is just doing some standard walking there that does seem to be pretty quickly and if it's anything like its predecessor the other Atlas it will certainly be able to run and be very quick now there's also something that most people also didn't know and that was the grippers so when the robot is standing up you can see its arms for a fraction of a second and it seems that what we're looking at here is the newly designed grippers that were placed on the recently demoed Atlas so if you may have not noticed recently there was an update to Boston Dynamics Atlas platform before it was retired and they actually showcased an interesting style of grippers so here we could see atas actually performing a kind of movement where it's being able to transport some kind of car strut into some shelf area and it seems that it has some upgraded grippers to be able to perform this tasks it doesn't look like a human hand whatsoever what it does look like is three different fingers or prongs whatever you want to call them and I think this might be the most effective mode of gripping objects for these human robots because I can guarantee you guys they've already tested this a bunch of different times and I've seen humanoid robots and not just humanoid robots many other style of robots be able to move and grab objects with remarkable ease even with just two pronged grippers so it seems that this kind of grip is what is going to be present in this style of Atlas And this is exactly how the robot is likely to be gripping now something I also think is going to be possible is that they're going to be able to you know remove these grippers from the robot simply because it is likely going to be interchangeable I doubt that this robot is going to just have one set of grippers fastened to the robot I think they're going to be completely interchangeable at the side and that is something that I recently saw in another robot demo where at the start the robot literally just slits on its hand and I think that would be really useful because if you want to use a robot in multiple different use cases you can just literally slip on and off someone's hand in order to be able to apply it to different tasks and I think that is going to be pretty fascinating now what does this actually hold for the future because whilst yes this robot is actually pretty insane what is coming next so previously there was some kind of talk that was released where Boston Dynamics Executives actually spoke about what this robot is going to be like in the future now bear in mind this waso so the information might be a little bit out of date but it's still show cases the vision and where things could be heading and of course at some point you know maybe you know the robot will be able to do things that a general purpose humanoid robot is expected to do and it's interesting you know what's in this picture is probably not 50 years away right I don't think it's around the corner as some people might want to make you believe but you know this is probably um in the 10year time frame one of the biggest things that we are working feverishly to figure out is actually not the ability of the robot in terms of uh moving and obstacle avoidance and autonomy most of that we have figured out one of the biggest things for a humanoid robot like that is of course safety um because this is a strong robot uh and spot can work side by side workers I think most of us feel pretty comfortable with him being in this room here and you know the bigger the robots get the more important safety will become so we can clearly see that there is a clear future of these humanoid robots being able to do a lot of things such as delivering parcels and many other household tasks and like I said before if it's going to be anything like what we've seen here just imagining a humanoid robot form factor being able to run across you know different parkour environments being able to Simply turn its hips in a way that we've never seen before shows us that things are about to get really incredible I mean if we look at the previous version of Atlas we can see that it was limited by a huge pack on the back a huge battery pack and a huge wires on the legs which doesn't allow it to do the level of turning that this robot can do now I think this is going to be a huge wakeup call for the robotics industry because I think people are now about to realize just how advanced robotics can be with boss Dynamics seemingly taking the lead in what is an already very competitive robotics industry companies like figure Apollo uni tree Sanctuary Ai and 1X robotics all competing for the top spot as the world's most effective humanoid and of course there was something here that we did see the last part that I do want to mention is right here which seems to be some kind of vision system and I think this was pretty interesting because what we also did see was when this robot was essentially you know moving we saw that as it stood up this beam right here it did light up so I'm not sure what that was iding or maybe it was just turning on but I do think that was rather fascinating and we can see I'm going to highlight this for you all some cameras right here I think we can see three different cameras that are for sensing the environment so it will be interesting to see how those cameras are used if they're going to be able to identify certain things to be able to do certain tasks if this robot is going to come with some kind of vision system I mean it's a very fascinating future so with that being said let me know what you thought about this robot do you think that this is fast things are going crazy I think that this is pretty interesting for the future considering everything seems to be accelerating all in the same

timeline humanoid a robot that they are dubbing Tang gong it's apparently the world's first fully electric powered anthropomorphic fullsize humanoid robot so essentially they've released a very short new demo in which they showcase this robots abilities and we're going to be diving into this and showcasing why this is so fascinating in terms of the world of AI so let's take a look here at this short demo so one of the things to note about this demo is that this demo wasn't exactly the longest kind of demo there was another one released that I'll leave a link to in the description but of course copyright is a big issue so essentially the reason I think this humanoid robot is rather fascinating is because of how it is powered they're daating that this is fully electric and that this is capable running solely on electric drive now this comes after a slew of announcements from China in terms of many other developments in the nation and this one actually did come as a surprise because I do know that China does have other humanoid robots in there like the G1 and many other humano robot platforms but this one was rather surprising now you can see here that the way how this robot works it does actually look very normal and one of the capabilities they showcased about this robot was the ability for this to almost get up some steps and it was pretty weird because one thing I noticed about this demo was the fact that as this robot is going up the steps it accidentally well I'm not even sure if this is accidentally or not but I usually don't see this in robotics but if we highlight this area here you're going to see that this robot is kind of tripping as it goes up there I don't know if you guys can see that but it literally looked as if it was tripping but I'm guessing that maybe this robot has some kind of advanced system where it's able to self-correct as it goes up which does mean that this robot is very stable in terms of the stability mechanics that they've used which is very good now in addition with that what we do have is we do have a scenario where there are some key features mentioned so one of the key features mentioned about this robot is of course the fact that it can run through slope Terrain in blind mode and I'm guessing that this is potentially a mode where they don't actually use the vision centers because if they're stating in blind mode I'm guessing that the legs are pre-programmed to pretty much walk over any terrain so that means that this robot is going to be very Nimble and agile when it comes to navigating different environments potentially environments that are foreign to this and environments that it wasn't trained on and that is very good because there are always going to be unique scenarios in which a robot is going to have to walk up in essentially a zero shot context and be able to clear the path so this does show us some really unique engineering and I think something that people might miss is that as the robot walks down here you can see that the level and the angle of the feet actually change and the reason this is so important is because many of the times in demos that we've seen recently we see robots and they walk around on the back here like for example we'll see a robot walking around on this flat surface right here but we won't see robots actually tackling going up and actually going downstairs we'll see a cool demo with some hands which is all very good well but what we don't see is this right here and I know this is very subtle which is why I'm trying to point it out to you but as the robot goes down you're going to see that the foot of the robot is of course here then as it goes down it Slants its legs just like a human would which is very important for adjusting your stability and ensuring your stable so you can see that as the robot goes down it then adjusts its legs so that it's very stable walking down and this is something that's rather important and then we can see this being done here and this is of course a very unique design and something that we didn't see in many of the other robot designs which I think is actually very important so this was a key thing that I actually saw that I was like okay they definitely worked on these legs for quite some time because they're very Nimble in being able to handle a majority of different situations and I think it's really important that we do not underestimate how important that is now there were some additional things that I did find to be completely interesting one of those things was of course the fact that this humanoid robot is a bit lacking in the arm Department I'm guessing that this humanoid robot they mainly did focus on of course the legs because once again the robot is able to run in this mode and this shows us that this robot has actually a pretty quick step to be honest with you guys getting a robot to move this fast without any overhead hanging cable is a huge feat that is not to be scoffed at because this definitely means that the stability mechanics of this robot are completely incredible and of course by the demo we can see that it does oddly look as if this robot is kind of half-developed not to take any shots at it but it does seem that they've been working on the legs quite a lot more than they have the Torso almost seems as if it's entirely made up of just a battery and then of course the hands look severely underdeveloped not in any kind of bad ways but it just doesn't have anything too crazy on it I mean there's no arms I gu guessing maybe you can have that in the future but it definitely is a very interesting humanoid robot now of course with the cameras it does actually talk about how there are several cameras there that it can be used to of course navigate certain environments and this was developed by the Beijing humanoid robot Innovation Center company and it actually stands at a height of 1. 63 M and weighs 43 kg it can also maintain a steady running speed of 6 km per hour and also the multiple Vision perception sensors that are located here are high Precision units are 550 trillion operations per second 3D Vision sensors and high Precision six axis Force sensors for accurate force feedback now one of the things that I was actually doing some research on when I was looking into this robot was I found out that this robot is actually open source so this was something that I think is rather fascinating because it's a bit different when you take a look at what robots are capable to do now open source is pretty crazy because open source actually helps out the community in a very insane way we can see here that this demo they're stating that tiang gong has open source and compatible scalability now I do think that is probably not the most accurate translation but if this robot is open source and it is developed in that manner this is going to be pretty crazy because open source projects actually benefit from contributions from a global community of developers and engineers and this can accelerate Innovation as more individuals and teams can experiment with and improve the robot's design and functionality now of course when they actually do open source this robot it also does provide a layer of transparency because this allows people to truly understand how this robot works and this means that you can build trust and reliability as users can verify the safety the security and functionality of the robots considering the fact that many people are fearful of this technology and what's actually really good and one of the things I think are really important for this if it actually is open source is the fact that an open source robot can actually be a valuable resource for educational institutions and research organizations and students and researchers can actually study the robot's design and operation which can lead to educational advancements and new research projects and one of the things that I'm truly excited for with this robot is that if it is truly open source this is going to foster a vast ecosystem of developers and innovators to be building on this platform where they can potentially customize the robot in certain ways that the initial creators may have not realized were useful or utilize it for certain things that could actually show us humanoid robotics moving in a completely new Direction so overall what we have here was a extremely interesting humanoid robot I've got to be honest with you guys this thing looks absolutely incredible in terms of the design and in terms of the slickness and I'm wondering if in the future well I'm guessing probably so that they may include some hands or some grippers that this robot may have as some functionality because right now it does seem a little bit underdeveloped in terms of that area but maybe this robot is for different things I guess we will have to see but so far this is looking very good and I have to say China have been non-stop with their releases recently in terms of China's first text

to AI video model now this is the recent announcement that they held and it's pretty Incredible vid is capable of generating highdef 16sec videos in 1080P resolution with a single click and it's actually positioned as a competitor to opening eyes Sora text video model with the ability to understand and generate Chinese specific content like pandas and Dragons now what you're about to see is of course the full demo in which they showcase the abilities of these clips and I personally do believe that this is something that is rather surprising so take a look and so there we have the actual demo and this demo has been received with many different mixed reactions and for a variety of different reasons now I'm someone who's pretty open to many different AI Technologies and I've taken a look at many different video AI generators and something I do want to say is that this is a lot better than you do think okay like I know some people are stating that this isn't great but trust me video generation is extremely hard and that's why many state-of-the-art models that you can currently use for free don't have the ability to do what we're seeing in this clip and you know the things like we've seen in Sora so what we have here guys is a clear indication that China has been slowly but surely ramping up its AI efforts and this is of course something that we aren't surprised by but I think this week is probably one of the most surprising weeks in terms of what China has been able to do in AI first of all they've got a robot which has actually been state-of-the-art in terms of Robotics second of all they then developed an llm that was pretty much state-of-the-art in terms of vision systems the small model systems and of course their large language model systems surpassing GPT 4 and then of course the third announcement that we did get from China was this vid this text to video AI model that can pretty much surpass the state-ofthe-art in terms of what is freely available aable now the demonstrations shown here some people might state that they are cherry-picked but I would argue that of course they're going to be with any kind of AI generation that you do get sometimes there are things that don't look right so of course in any demo things are going to be cherry-picked for that demo and I don't think that is a crazy thing at all now there's some more information and some key things that Eide viewers actually may have missed so I'm going to be showing you guys those key things that you probably did Miss if you weren't paying attention to this actual video so one of the things that you probably did miss about this trailer was this right here you can see that the creators of this text video model clearly know that Sora is the biggest AI text video system in terms of competition and because of that they are uniquely positioned in where they've placed certain Clips in the trailer and one of those clips that they played in the trailer was of course the clip of the woman and the man walking down a busy street at 9 night in Tokyo now we can see here that the one from open eyes Sora actually looks very good in terms of the temporal consistency and in terms of everything else now with regards to the vid one when they actually did show us this it was only around 3 seconds in the trailer but I do have to say that it is pretty good motion for their first ever system I mean maybe it's not the first ever system but the first one that's gained notoriety due to the level of detail and level of consistency now of course as you can easily tell you know opening eyes Sora is miles ahead and in fact I wouldn't actually say miles ahead I would say it's a decent bit ahead but maybe with version two they could most certainly catch up to this model and if we go back to the trailer you're going to see that there are several different instances where things are quite similar so for example right here you can see that it's quite similar to the woman who's walking down in Tokyo and of course if we look on to the right there is of course some morphing on the hands and of course on the legs but I think we have to give credit where credits du because if we actually look at the skirt right there we can see that as the legs move up on the right hand side there is a neat bit of deformation on the skirt that actually does look very normal and it does look really correct an example on the jacket when the guy is walking we can see that the jacket is actually swinging around and the hips and the motion right there actually do look pretty effective so I know a lot of people were dunking on this saying that this is you know objectively mediocre and I've seen many different tweets stating that but I would have to say that this is not mediocre at all and this is definitely a state-of-the-art level system because if an AI company right now came out and released this in the west this would definitely be heralded as something that is a SORA killer so I think maybe what we have here is a situation where things look very good but people aren't really appreciating what it is just because it's not available for use yet and because I'm guessing Sora exists now there was another demonstration that was also pretty interesting we did see open eyes Sora if you do remember this clip that was initially released with Sora they released this clip of a kind of Land Rover thing driving around the hillscape and it was pretty good now when you do compare it to the new one of course the vid it of course doesn't look as good in the trailer but I do have to say that it still is pretty decent B based on what we've seen now you can see right here some of the things that I will say that this does get right and other systems don't is of course the temporal consistency so for example right here when we're looking at this clip and it's about to come up now you can see like the bushes they actually stay in motion and they move past as well as with the trees and the only thing that I would say with this as well is that I downloaded the video so I don't know if there's like a higher resolution version available online so I can't really right now comment on the quality of this because the video has been shared around so many times with the original footage being hard to source so I don't think that this is the of course is like the highest quality because even on this video clip like even in the resolution that I have it now there are some clear artifacts on this video um like here you can see that there is like light like breaking up and that uh usually happens when a video has been downloaded and shared so many times so it's pretty hard to find the original 1080p clips and I think once you know they release them again on like maybe some official like YouTube type thing then we can actually see how much better they are because a lot of people might say that the quality isn't good the temporal consistency isn't good I would disagree and like I said and this is a key thing because videos are being shared around the resolution is getting lower and that is going to impact how people actually do see this and I think that is an important thing to remember when trying to be unbiased when viewing this now if we actually do take a look at something here and this is what I'm stating okay is that if we look at this um opening is Sora isn't actually released yet which means that what this is is this is a state-of-the-art system because if opening ey Sora isn't available yet and we know that this requires magnitudes more compute than we could even think about this is something that we need to you know take a look at because when you've seen Sora and of course Jen 2 combined SL compared we can see that Runway which is you know arguably the second SL you know Top Cabs or whatever um we can see that Runway generation 2 doesn't really have any good temporal consistency in the likes of what we've just seen and I think it's important to note how crazy this actually is because they're actually I would argue that it's actually quite better than Runway generation 2 well yes it does have some good features in terms of being able to move smoothly a little bit I would say that there isn't much motion it's more of like a really slow you know motion thing like for example here this is a key example if we compare opening eyes Sora this one right here to runways Gen 2 and then we see what vid's done because VI have done direct comparisons because they're a direct competitor we can see that op I Sora if we actually look at how the water moves we can see that it moves pretty well the ships are moving you know all well and that looks pretty good but on Runway Gen 2 the waves and stuff like that it doesn't look really well and then of course as well we can see like if we go back to vid and this is what I'm talking about when I say that VI is something that's pretty important we can see that in this demo right here which they've shown us there's actually decent consistency in terms of the wave and how things actually move around it and then of course there's another example somewhere in the clip right here like right there we can see that the waves crashing around it looks pretty realistic like they're not morphing into the boat everything looks PR pretty realistic here I know this is just like a short demo but that's not something you actually see at Runway at all that kind of motion these waves crashing around you didn't see that in runways at all and we don't see that in pabs which means that of course you know that kind of temporal consistency with the motion where the characters were walking where we saw the skirt and of course we saw this kind of thing at opening eyes it means that you know they're definitely a step ahead now in terms of the architecture for video was actually proposed as early as September 2022 predating the diffusion Transformer the dit architecture used by Sora so vidu is pretty different it utilizes a universal Vision Transformer a uvit and that architecture actually allows it to create realistic videos with Dynamic camera movements and detailed facial expressions and adherence to the physical World Properties like lighting and shadows and I think that they've done something here pretty amazing considering they're not even using the same architecture and like I said before if we also take a look at one more example and this is why I'm going to show you guys how good this is because a lot of people are looking at this and say it doesn't look that good but if you compare it to what's actually state-of-the-art that we can get our hands on now this is something that is really good um and it's pretty surprising so for example again taking a look at this opening eyes Sora the clips of the TVs moving around and stuff like that you can see that all of the things are flashing and then you can see with Runway generation 2 it's very you know not that crazy at all it's very slow there's not really much in terms of you know moving around but if we go back to vid and this isn't like this isn't any kind of favoritism what not but if we go back to vid there's a clear example here of them moving around okay moving around the objects like this and we can see that this is pretty crazy like that is really hard to do like this is insane like honestly this is crazy like and I do need to get the HD versions of these because like I said it's a bit disingenuous although I couldn't find them I really did try and search but it's a bit hard to you know judge this when you don't have the full resolution Clips but you can see like in the back okay like take a look at these images here they're staying in place they're all not deforming they're not meshing and you know going around these TVs all of this motion is happening and the TVs are moving correctly like that okay compared to what we just saw here like literally at the end of this okay you can see right here like literally compared to the end of this like honestly guys it's pretty good like you have to say that this is pretty incredible in terms of what they've been able to accomplish and another thing as well is that if you take a look at AI videos that were one year ago literally one year ago and then um you know you take a look at what we have now with Sora and the kind of technologies that we have now and you know the craziness of what we're able to do I think we have to understand how far we've come in such short of a time and you could argue that yes um it's not just short of a time things are pretty different in terms of you know architecture being building up on different architectures and Decades of research but things are starting to accelerate so that's something that I think you know should keep interesting but it does seem like CH China right here genuinely has taken the lead because like I said before I think there is probably like an actual 1080p Vision because the videos have just been you know downloaded so many times um and like I said before we don't even have access to Sora was given to you know people that are in like the film industry and stuff like that and they've been using it and they said it takes about 10 to 20 minutes per render and you can render up to 3 seconds all the way up to a minute clip long um and I think that's a really important point to remember so let me know what you think about this I think this is a pretty gamechanging stuff um I think it's pretty crazy let me know what you think I think this is absolutely incredible from them and I think in the future we're definitely likely to see more and more competition and once again what I do find surprising is that China's been able to pretty much catch up to state-of-the-art models in not a short amount of time at all and I think they're definitely going to prioritize this technology so I wonder where does that leave the US in terms of how they're going to now prioritize this are they going to speed up their acceleration or slow down and regulate it in different manners I honestly have no idea but I do think that the USA are probably going to speed up the development when they've seen that you know China can pretty easily catch up across all bounds and I think this is definitely going to create some kind of AI not arms race but definitely AI race um and it will be interesting to see how this is deployed

in the future here is we have a situation where China has potentially taken the lead with their new model two days ago sense time launched sense Nova 5. 0 which according to this report actually beats GPT 4 on nearly all benchmarks and there are actually a variety of different things that are truly surprising about this it isn't just some hyped up gp4 thing if the claims are true and we're going to dive into them this means that things are truly ramping up and it seems that across the board from different nations and different countries we're definitely going to be seeing increased competition in terms of what AI systems are capable of so we can see here that they've essentially compared this to GPT for Turbo but there are a lot of things that this presentation dives into and bear in mind that the majority of this is in Chinese but I've done my best to translate most of this so that you can completely understand exactly what's going on so you can see here that their sense NOA 5. 0 actually looks pretty decent so on the left you can see it states that it is Hybrid used in the middle they states that it is trained on over 10 billion total tokens the inference actually supports up to 200,000 tokens which is rather interesting because we've seen longer and longer context windows so a 200k context window is something that is you know I wouldn't say surprising but it does show that of course things are going well and of course the main statement from this is that performance exceeds gp4 turbo now the reason as you all know why GPT 4 Turbo is highly talented as the state-of-the-art model at that time and that's why people constantly constantly try to one up that in terms of the benchmarks so this was the start of the presentation but things do get a lot more interesting and I'm going to show you all exactly what that is so there was one screenshot here where I'm not entirely sure if they actually had these AI systems hooked up to a game but the text states that a live demonstration was conducted on the comparison of multiple functions of their model and GPT 4 including creative writing logical reasoning diagrams image understanding and calculations of food calories based on pictures and this is what the text actually States here but we will get into the benchmarks but I just wanted to include this for full transparency in order to show off the muscles of the larger models sense time also played the King of Fighters and at first the green player gp4 had the slight upper hand but was quickly overwhelmed by the red player sense chats liked various combos so I'm not sure if they actually talking about this as if it's some kind of metaphor for their system beating GPT 4 or if they actually hooked it up to a street fighter Style game and since chat light was able to best gpg 4 if you are someone who can translate this more effectively because I did honestly put this through multiple different AI systems and the result was frequently the same now here's where things start to get interesting and here's where things are really fascinating because these are obviously the benchmarks for this system you can see sense chat version 5 is over here on the right hand side this is sense chat version 5 and then of course what they actually do and this you know pay attention here because I'm going to you know say some things that I'm going to recall them so you can see right here that they actually do compare this to gbt 4 Turbo and that actually is uh to the eaglee viewers that it is the 116 this isn't actually the most recent one but it's not also a most outdated one as well and we can see that the only benchmarks that it surpasses the model S chat version 5 on is of course the math zero shot Benchmark right here and we can see that is highlighted in the 61% area and we can see that in this Benchmark right here as well so we can basically see that currently what we're looking at is a situation where across all of the benchmarks compared to gp4 Turbo one of the most recent models I don't know why they didn't do the most recent one I'm guessing just based on where they are in their testing and of course we do have a comparison with llama 370 bilon parameters instruct we can see that the differences from GPT 4 are honestly I wouldn't say quite Stark but it doesn't just seem to be uh a simply percentage gain like for example on the MML U we can see that this one is rather small but on the cmml U I'm guessing that this is a different version of the MML U perhaps the Chinese MML U honestly I'm not entirely sure we can see that it increases but you can see like I said on the different benchmarks here the gains don't seem to be uh completely incremental because if we actually look at the differences this one is at 80 and 93. 612 as in when I was actually checking some of the other models I couldn't find things like the race natural questions on other models and that's like Gemini Pro and Claude and I do actually test them against it and I'll get to that later on in the video but essentially I think this is rather fascinating that you know China has I would say surpassed GPT 4 Turbo in their model and what I would like to see from this model honestly is if you know people could potentially test this model because it would be interesting to see how it actually Compares in terms of users and what I mean by that is I mean that whilst yes currently on this chart what you're currently seeing is GPT 4 only winning in two out of the several different categories here and also even llama 3 we do know that there is a different Benchmark which is rather important and The Benchmark that I find to be one of the most interesting because it's not just based on the instance where things could be fine-tuned on the actual benchmarks in terms of just trying to get something that beats things on benchmarks but isn't actually useful on a day-to-day basis you can see here that we do have GPT 41106 on the chatbot Arena as I've stated previously the chatbot arena is different because it ranks models based on their Arena ELO which is determined by votes against other systems and basically all this means is that people test AI systems side by side and their ELO increases when people vote for that system being the most useful and it's a blind test so you don't know which system is which and then over time you get to see which system is actually useful on a day-to-day basis answering a variety of different questions without the bias of looking at other systems so it's something that I think is very useful and we can see that GPT 4106 is basically at rank number one and of course like I said the new one does have a slightly higher ELO rating but I still think that this is rather impressive for what we're seeing here now something that I also did want to test was I wanted to test this against other models because one thing I realized when looking at this was I was like okay whilst this is good there are other state-of-the-art systems out there because yes gp4 Turbo 1106 is state-ofthe-art but GPT 4 Turbo isn't the only one and gpt3 turbo and llama 370 billion parameters aren't the only ones there now the other one that most people might be thinking about and maybe you're not thinking about this but I certainly am is I'm thinking about Claude 3 okay so Claude 3 if you don't know it's a Model A state-of-the-art model and it was released by anthropic now I've done this a little bit sloppily but let me just explain to you what you're currently looking at so all you're looking at is Claude 3's benchmarks and we can see that it is compared to GPT 4 in this tab right here and essentially across the board completely Claude 3's Opus model does surpass the benchmarks of GPT 4 now what this does mean here what I've actually done here is I've actually circled these areas where Claude 3 retains its leadership in terms of how far it is but I've added X's here because these are areas where sense chat V5 actually beats Claude Opus so in the math benchmark which is math problem solving it does better than C Opus there we can see it's 61. 9 and this one is 60. 1 so that's where s shat V5 reign supreme then of course we can see here on the common sense knowledge the H swag this is 95. 4% and then for here on the common sense H swag we can see that theirs is actually at 97% so this is actually a huge Improvement we can see that 95. 4% to 97. 52 is a big Improvement however Claude actually retains the leadership benchmarks in these areas where we can see uh the MML U The Graduate level reasoning the grade school math but something that I do want to add whilst yes claw 3 does have the best on these benchmarks I would have to be honest and state that it isn't far off like for example if we look at the code on the human evaluation at 84. 9% or for example if we look at the big bench hard we can see it's currently at 82. 9% whereas on the big bench card it's at 86. 8% so I mean it is pretty interesting to see how these differences are but the long story short is that basically since chap V5 surpasses gp4 turbo claw 3 Opus surpasses gp4 turbo but in terms of the actual benchmarks the ones that you do want to care about according to the arena gp4 Turbo's most recent version retains its leadership spot even if the benchmarks that you know are bench boed by like math and coding and stuff don't appear to be correct people seem to be having a great time with gp4 turbo F firmly followed by claw 3 Opus and then Google's Gemini 1. 5 Pro in the API now I think this is still rather impressive because it seems that this company has been working very quietly and very diligently in stealth because this is something that did catch people off guard now there are some other things that are actually really surprising and one of the most surprising things about this was something in regards to the smaller models because their smaller model is insane and it's so crazy to the point where I'm wondering to see if some of these things are actually even legit so one of the things they actually talk about is a writing task and here's where they compare you know GPT 4 you can see right here GPT 4 to sense chat V5 and right here this is where and funnily enough this actually does look like the chatbot Arena so I'm wondering if in a couple of days we are actually going to get sent chat V5 in the arena but um they talk about how essentially the summary of this you know what actually happened here was that they basically had a task to write a college entrance exam essay on Innovation based on the classic Chinese novel dream of the red chamber and the author notes that gptt 4's writing style tends to be more rigid and structured using repetitive phrases and in contrast since Nova's 5. 0 exhibits a more free flowing and Divergent writing style drawing upon a wide range of Chinese Cultural and literatur references from ancient times to the modern internet age along story short they're basically trying to say that if we look at the writing tasks in these examples that they've included GPT 4 loses now there was also another one right here where they put GPT for's model against sense chat V5 in I guess you could say a logic SL reasoning task and they were trying to compare this in terms of what this model would do when it was trying to figure out the amount of coffee and water consumed and honestly it's pretty hard to decipher what's going on here because like I said before it's a translation so the interpretation might actually not be correct and I'm not going to get into the details of that but the long story short here was that gp4 wasn't able to realize exactly what was going on but since chv 5 the reasoning steps provided the correct answer I will leave the prompt and the explanation in the description but it's just very confusing trying to get a logical reasoning Stars correct because if you've ever seen the examples of these you know incredibly confusing tasks one word can truly change the entire outcome of the answer and it's important to get the translations 100% because even if you get them 99% you can completely fail the question so this is probably the most fascinating part of the thing when I was looking through it in terms of the benchmarks and in fact that's actually not true it's actually coming up soon but right here and we will get to the most interesting part here but right here this is where we take a look at their visual recognition systems and of course you can see it is compared to Google's G Pro Vision open ai's GPT 4 Vision quen's VL Max which is a vision system and of course the intern LM X composer 2 VL and Step One Vision so there are all of these different ones compared on these benchmarks and we can see you know from what we've seen here that the top one does surpass them now one thing that I also saw that was really cool was that they also did have image generation and by the looks of things it seems to be very photorealistic so essentially the Tex states that sense 5. 0 sets a new benchmarks in the term of AI powered image generation demonstrated by its impressive performance in generating nuanced and lifelike portraits the following example illustrates the ai's capability to create text image visuals and essentially using the prompt a vibrant Asian female portrait with beauty makeup casual hair a smile natural with movie grade quality presenting different expressions and styles since Nova 5. 0 showcases its sophisticated inter ation of textural descriptions and ability and its ability to generate diverse facial expressions and styles that carry the subtlety and richness of film grade portraits that description is pretty good but I've got to be honest if this is their text to image generation model I have to say the whilst yes mid Journey might be on this level this is definitely something that looks really good I mean we can't say that this doesn't look completely realistic and as for the prompt I think this is something that does do it really really well now of course I'm guessing that they did compare this to other systems right here in fact it isn't entirely clear but I'm guessing that this these are other systems but either way this one right here does look really realistic and I have to say that this is pretty impressive now here's where we talk about the most interesting bit because like I said before we were previously supposed to but you know I was you know thought that the slide was the other slide now essentially this is where they talk about their smaller models they're more compact models that they can use for many different things now you can see that the one they've highlighted is s chat light which is 1. 8 billion parameters in size and it does do a lot better than others at the similar size we can see compared to the Gemma 2 billion parameters by Google it completely destroys it we can see that llama 2 13 billion parameters it completely does better than it but one thing that I didn't really understand from this and was pretty confusing was of course the benchmarks because they didn't actually do this on traditional benchmarks they did this on a different kind of Benchmark so when this text on this left column was actually translated the benchmarks that we got were I guess you could say the words we got were comprehensive score language comprehension creativity reasoning and the average overall so one thing that I do think is quite surprising here is that 1. 8 billion parameter model does showcase incredible capabilities because time and time again we've seen that Trend to be increased now one critique that I do have here is that they didn't compare this to Microsoft's 53 and I'm guessing that you know 53 was literally just released so I guess that's okay but they didn't actually compare this to llama 3 you can see here that they compared this to llama 2 7 billion parameters um and of course they state that it beats it and you know beats Googles and you know the other ones like quen's 1. 5 but even what they actually did was if we actually look at the original you know parameters we can see they actually did compare llama 370 billion instruct so I'm guessing Okay and this is just a guess maybe it isn't true that if the original system beats llama 3 um I'm guessing that this smaller system doesn't beat llama 3 because they haven't included it for whatever reasons and I'm pretty sure they could have so I think this is still pretty interesting because they've got a smaller model that is you know levels above these other smaller models but the point here is that based on some other things that I've seen as well you know they actually do talk about image generation and I got to be honest the text here is a little bit confusing I have tried to understand the context of what it's trying to explain but it does get a little bit confusing sometimes and I will leave a link to the article one thing that I actually do talk about that uh they say that you know this is going to be a calorie assistant and you can use it to submit images and completely understand your calories and I think right now the AI space is heating up quite a lot and this is not going to be the only company coming out of China to present their models that do take us by surprise but so far this seems like a very interesting update now what's also interesting is that the company shares soared more than 30% after announcing its lative generative AI model so this was something that made the company's stock price jump 30% so maybe it could be argued that the benchmarks that they're talking about aren't as good but the only way for us to know is of course for us to test it in the arena for other people to do their independent evaluations and of course things are going to be a little bit different because I'm guessing that this is fine-tuned on the Chinese language so the translations might not be as completely accurate so I'm guessing maybe if they make an English model that might be as good or better but of course this is going to be something that we would have to see so let me know what you think about sense time's new model and if we think China is going to catch up to the US competitors and how this actually impacts the landscape I think this is of course some very fascinating competition because I think we're going to continue to see you know models and different companies po millions and billions of dollars into this industry

we're not going to click on this just yet but this is how you can check if you actually you have this so make sure you check on settings personalization to see if this tab is right here if it is then I guess you have access if it isn't don't have access so what I can do right now is I can tell jat TPT something so I can say I run a YouTube channel called the AI grid can you remember this and not forget it you don't have to be this explicit but I'm doing this for demonstration purposes so you can see right there it literally says memory update ated and it says got it I can assist you with your YouTube channel the AI grid today so now that the memory is updated you can see I can click manage memories and then when we go to manage memories you can see it says runs a YouTube channel called the AI grid now this is pretty cool because now if I go into another chat so let's go ahead and test this I can say what tips can you give me for my YouTube channel now I didn't tell it I actually run a Channel about AI so maybe it might not get this but if I say what tips can you give me it says to enhance your YouTube channel called the AI grid you can see that it actually remembers my YouTube channel and this is really good because now it means that every time I start a chat this is going to be increasingly personalized to me now the way how you want to utilize this is you want to make sure that chat gbt has memory on what you want it to and essentially the reason you actually want to use this memory is because you want it to save time so what I'm going to do now is I'm going to stop the chat here and I'm going to show you how I would use this to save myself time so I would say can you and you can do this in a many different ways update your memory so that you know I'm a male living in London and I eat healthily and go to the gym every day and there we go so I'm going to add that in and the reason I've wanted to do that is because now they have a lot more cont context on certain things okay and another thing I'm going to do is I'm going to say okay so also add to memory the fact that I have a cat and a dog okay so I'm going to say also have add to the memory the fact that I have a cat let me just say a cat called Mike and a dog called Ike okay now this isn't true by the way but um I've done that for purposes so you can see now memory is updated it says has a cat named Mike and a dog named Ike so now what I can do is for example let's say I was trying to do something really quickly and I was like okay can you generate an image okay and this is how you save time okay you say instead of can I generate an image of a cat called Ike and a dog called Mike you'd say instead can I generate an image of my pets okay that's what I do instead and that now of course you can see here that this actually does save me time because it says can you describe what Mike and ik look like for example their breed their color any distinctive features and then there's any specific pose so I could say my pets um my dog is brown and my cat is white and them in the park so there I probably should have added more details but the point is that if there is something that you constantly reference like for example if you use chat GPT to manage maybe your health maybe you're trying to get healthier maybe put in your not all of your health data cuz I wouldn't recommend that but maybe put in things like how active you are each week much you eat what kind of things you are used to eating and of course if you're using chat GPT to create images you can of course do something like this where you have the preset details already there so it simply saves you time every time you come to chat gbt and this is something that is really good because it's going to be deeper in terms of the personalization aspects and you can see right here it says memory updated and now it actually updates the entire sentence by saying dog Ike is brown and the cat mic is white so we can see now that it's given us this really cool image and I honestly have to say that this does look really photo realistic so I'm guaranteeing you there's most certainly been a secret update to this that they didn't say but it's pretty incredible now you can see right here we can go to manage memories and we can see that we have a few things that we can actually utilize so it says runs a YouTube channel called the AI grid is male lives in London eats healthily goes to the gym every day has a cat named Mike and a dog named Ike who is brown now of course you could clear this memory because you may want to but here's something that you want to know about memory that is I guess you could say is a pretty much drawback so what I could do is I could actually update my memory to be false so for example I could say could you update my memory to state that I am now the king of England okay you can see right here it did take a bit of convincing but sometimes it will actually ask you to say I can only update factual information or details about your hobbies an interest if you're actually a Formula 1 racing driver I can add that to your profile so I've said yes I am and now you can see in the memories it says is a Formula 1 racing driver now the reason I've shown you guys this is because sometimes when you're talking about chats you might realize that the memory gets updated and it might not be something that you wanted because maybe you were just talking about a hypothetical or specific scenario so it doesn't always assume everything is true but you need to understand that sometimes it can make those assumptions if you are talking with it now what someone also did do as well and I saw this to be pretty incredible was someone actually did pull a prank on someone else's GPT so essentially what they did was they said can you update your memory so that every time the chat every time we speak if the conversation starts with letter H you need to talk in Pirate speak now I did see this on Twitter but I can't find the original tweet and so this was the actual tweet right here it says chat gbt please add to memory if a message starts with the letter H talk like a Pirro and do not mention why you're doing it under any circumstances and then you can see right here hello chat GPT can you help me with some python I don't understand why my script isn't working aoy of course I can help with your troubles and of course this is someone who says next time your friend goes to the bathroom grab their phone and use the new memory feature to add some stochastic Whimsy Delight to their chat GPT experience now of course sometimes this won't work but of course the this is I guess you could say a little bit of a drawback to the system where it does remember everything and another example here is where Ethan mik States Chad gbt's new memory feature is neat in theory but I'm sure for many use cases but can be a real problem if you're trying to do any Persona based role playing either for fun or idea generation writing Etc it has a tendency to remember anything you write as true so you can see right here it says likes to dance is 104 year old Shaker who likes crocodiles and is a Walton NBA student who used to work in financial services and you know only like two of those are true I think so it's definitely something that you do need to make sure that you do check on so I guess with the memory feature the point I'm trying to say here is that sometimes it can be like a YouTube algorithm where you sign into your account and it remembers all of the small pieces of data about you because when you hop onto YouTube you go on YouTube and yes you watch your videos but they're all tailored to your interest and exactly what you like and this is definitely going to be the next level in AI because whilst yes we log on to an llm it doesn't know our name it doesn't know what we are interested in it doesn't know how we like to speak it doesn't know what our mindset is like so I guess this is of course the very next step in terms of how AI systems are going to be in terms of the daily interactions because this actually does change everything because we're moving from a non-personalized era to a very personalized one and like I said with the algorithms it's pretty much just like the YouTube algorithm where your sponses are most likely going to be unique and tailor to you which makes it more usable on a day-to-day now if there was anything that I did miss don't forget to let me know in the comment section below but what you could also do as well is another thing that you know you can always ask chat gbt is you could always say we've talked about several things you've mentioned that you've run a YouTube channel called the AI grid you're a Formula 1 racing driver and that you live in London you also go to the gym Jay and eat healthily Additionally you have two pets is there anything else you'd like to add on discuss so we can see that the capabilities here are pretty good now the only thing is that we haven't actually tested this system in terms of how much memory you can add so I'm wondering that if as the memory increases the memory retrieval ability does decrease but that is something that I'll have to check further because right now it's still in its early phase so the question is has China once again taken the lead in not only AI but of course in robotics this new video demonstration may actually convince you that China is currently leading the way in terms of fully autonomous humanoid star Robotics and it's actually very surprising at how good this is because I just didn't think we were that close to robots at this level so let's take a look at this demo and why this is so impressive So currently you can see that this is called the astrobot S1 the Next Generation robot that is naturally yours and everything you're about to see in this demo is at onetime speed with no manipulation and it is fully autonomous remember that because what you're about to see is rather fascinating so ladies and gentlemen presenting the AST S1 coming from shenzen China so take a look at this first initial demo this is by far the most craziest thing I've ever seen what f so the demo that you saw there the reason this is so impressive and the reason that a lot of people don't even believe that this is real is because what we're looking at is something that is extremely fast extremely smooth and it's fully autonomous meaning that there is no teleoperation And for those of you who are unaware teleoperation just means that there is no human behind the machine controlling it with a VR headset as we've seen in many other robotics demos and some of the stuff that also makes this rather impressive was of course the large language model integration now I'm going to speak first about this right here because even as someone who pays attention to pretty much every Ai and Robotics development this first demo here genuinely doesn't even seem real because of how quickly this is if you aren't familiar this is a cup stacking game that many humans do play and of course the goal is to stack this as fast as possible which is pretty impressive for a robot that seems to be fully autonomous and this is of course something that I would be honest with you guys a lot of humans do struggle to do this so this is definitely something that is rather impressive because it takes a certain technique to be able to pull the cloth from the glass like that now when we do get into this area right here this is of course where things start to get even more interesting because we can see that there is of course a large language model integration so the user first asks the AI system what do you see and then it clearly states that I see an orange pingpong ball a red race car toy a white Hello Kitty figurine a pink notebook two white containers and a red pen and laptop so it's clearly identified all of the items in the system I mean in the scene but we can see here that this robotic system clearly has an onboard Vision system so it's onboard Vision system we aren't sure what it using because they don't actually state which of course first llm it is so we don't know what llm this is it could really be any onboard llm to be honest there are tons now tons and tons and with the rise of small large language models it could potentially be greatly improved in the future as we've seen with releases such as 53 so we don't know what large language model they're using but they're also using some kind of vision system too now of course what's also very impressive is that as this continues on we can see here that the vision system actually does look similar to the YOLO Vision system which is a vision system that can identify tons and tons of different things and we can clearly see it managing to identify multiple different items now what's also interesting is that we do see that it states here that how would you separate the items to tidy up the T and then it says we need a container for the toys and writing utensils stash stationary and I think that whilst yes of course that is quite obvious to humans because we've done this a million different times it's something that is increasingly good to see in robots because it shows the level of development and a small level of reasoning and understanding when combined with all of these different AI systems working together with a true humanoid robot so they also stated that this was done via imitation lining so I'm not sure how many examples they've managed to get how many human demonstrations but I'm guessing that it probably was a bunch considering the rate of speed when this robot is active and of course how good and smooth this robot is at completing the tasks because these tasks don't seem that they're really half done they do actually seem as if they're really done well and what's crazy about this is that we've seen a few robotics demos this year and to be honest with you guys this is by far the most impressive demo to date and this is just based on what we've seen and the craziest thing about this as well and this is something that I've mentioned many times before is that when looking at robotics demos it's pretty surprising with what robots are simply able to do with this two-pronged approach because many individuals would state that humanoid robots in order to function effectively you need five fingers just like a human uses but it's clear from many different robotics demos that might not be the very best approach sometimes keeping things simple and smooth can truly work runers for what you're trying to do and taking a look at this we can see that of course this is exactly the case with this level of Robotics now whilst yes some people might argue that with some robotics demos the problem with this is that they are in I guess you could say controlled environments where objects are placed down and of course things are done in a pre not pre-programmed but preet routine where they've gone through the demonstrations many time but I have to be honest doing something like this where you have a paper plane being thrown into a bin that is something that honestly for any robot to do along with the other tasks is very impressive and of course as well tasks like this which don't seem that impressive I think are truly impressive because it's literally cutting vegetables now another thing that I would like to find out about this robot is if this robot does have a moving base or if it does have any legs because whilst yes we can see from the top of the robot we can see that it does have two arms and a vision system hooked up to the back in order for this robotic system to you know be useful in the future we're going to have to see if this thing can move around now with this asot it's actually not to be confused with the commonly looked at reflex robot now I'm only showing this because this is what the robot isn't although it does look like this robot here is by another different company and this robot is actually completely teleoperated so there is a human controlling this so this is not fully autonomous but I'm guessing that this is going to be the same kind of architecture that asot uses when they're deploying their robot around the world and of course like I said before this is going to be pretty incredible because this is something that we haven't seen yet in terms of just the sheer speed and the fluidity of these model so I've got to be completely honest with you guys this definitely took me completely by surprise due to the fluidity the speed and not only that the craziest thing about this demo was that there were many several different examples of what this robot could do it wasn't just one or two but there were several instances of it being able to do complicate and intricate tasks that even Some Humans might actually find pretty difficult now one thing that I'm looking forward to from this robot in fact two things are number one is with this robot being combined with an actual llm system one thing that I would like to see in the future is of course this robot potentially using that llm to verbally speak with commands and have humans interact with it on a verbal basis because currently it seems that it was just a chat interface from the audio clip it didn't seem as if there was any actual human to robot integration via systems like whisper or any kind of audio transcription software it just seemed as if someone typed it in and then the robot completed the task based on that so that is what I would like to see in addition to this because I think that would you know improve this robot in terms of the useability and of course another thing that I would like to see is this robot potentially moving around in the environment because one thing that robots do struggle with it's being in different environments and when the environment is reset it becomes harder to identify objects and move objects to the correct location so those are two things that I think would of course be more impressive but I'm not trying to critique this robot at all I've got to say that surprisingly surprisingly the leading lab currently in robotics from what we see here and yes I always know that some people are going to be 100% skeptical because currently there is a lot of hype for Robotics and humanoids so when people see this stuff they might think

that you know it's teleoperated although it says no teleoperation I think that we can certainly say that this might be the leading lab now whilst in terms of demos we do know the bossing Dynamics robot did a really impressive feat where they showcased how their new robot is going to move and I'm wondering if their robot is going to be far superior to this one in terms of the dexterity the degrees of Freedom the things it's able to do in terms of the running and completing other tasks so this shows us now that China is not playing games when it comes to humanoid Robotics and the robotics industry and if you haven't checked out my previous video there was another video where I showcased a different completely different AI system that just bested GPT 4 and is now A state-of-the-art system that completes the requirements for being the best AI system now whilst many people said they're not going to believe it until they test it I think we do have to give credit where credits do because this is clearly an area where China is investing a lot more than the US in terms of the infrastructure and in terms of the long-term planning so I honestly can't wait for this company to Showcase more things of course this could be completely fake we honestly have no idea but I really don't believe so considering the amount of breakthroughs and impressive demos that we've completely seen so with this one time B a no teleoperation robot doing incredible things I think we're about to see another level in terms of competition because companies and individuals are starting to realize the sheer implications of what this means and honestly when I saw this demo one thing that I did actually think about that was in the back of my mind was thinking that wow we are going to move so quickly towards robotics that maybe just maybe Robotics are going to be integrated into society in much more ways than I initially imagined because I only imagine them doing certain things but if a robot can actually generalize with an AGI level system combined with all current AI systems and manages to prove in efficiency and fluidity and the dynamic movement then we really could have robot in a variety of different scenarios so with the astrobot S1 let me know what you guys think about this because I think that this is truly a state-of-the-art system and I think this like I said before is going to increase the level of competition from other companies and another thing on their website they didn't too have too much details on there but they did say that this robot is expected to be commercialized in 2024 meaning that they're probably going to be going to Market this year and that means that we're likely to see this fully autonomous robot completing tasks later this year so if that does happen that would be pretty crazy I honestly expect 20125 because from what I've seen from robotics companies there's usually delays and there's usually things that happen on the back end that do slow them down but so Sam Alman recently had an interview at Howard University where he actually spoke about a variety of interesting topics and there was a lot that he discussed it actually gives us an insight to things like education the role of AI in the future and of course artificial general intelligence which kind of gives us a gauge on where he's at in terms of what he's thinking now the initial interview was actually done back in January but it was only just released which means that this interview is from literally 4 months ago so that's of course something to keep in mind but nevertheless let's take a look at the first section where Sam Alman actually gives his opinion on what will be the most important skill for people to learn in the future strong agree with that uh I think critical thinking creativity the ability to figure out what other people want the ability to have new ideas that in some sense that that'll be the most valuable skill of the future if you think of a world where every one of us has a whole company worth of AI assistants that are doing tasks for us to help us express our vision and um make things for other people and make these new things in the world the most important thing then will be the quality of the ideas um the curation of the ideas because AI can generate lots of great ideas but you still need a human there to say this is the thing other people want and also humans I think really care about the human behind something so when I read a book that I really love the first thing I want to do is go read about that author and if an AI wrote that book uh I think I'll somehow connect to it much less same when I look at a great piece of art or if I am using some company's product I want to know about the people that created that so I think in both directions of humans knowing what other humans want and also humans caring about the humans behind something um this will be that'll be a super important skill uh and so I think learning that ability to create come up with new ideas choose ideas from among the many options presented by an AI uh that'll be very valuable I agree with you the tools will change but I also think familiarity with the tools of today and this new way of using computers is really important and that'll be important for everyone not just the tool Builders but everybody like in the same way that if you can't use a mobile phone you're kind of at a huge disadvantage but they're not that hard to use and people learn but the earlier in your career you got familiar with it life the better you know everybody in this room was familiar with it probably as long as you can remember but uh I remember watching older people struggle with getting comfortable with a phone for the first time as intuitive as I thought they were uh I think it I I think human adaptability is remarkable and so I'm very happy that people no longer think it's weird or impressive that we can talk to a computer like we talk to a human and it understands us and it talks back to us and it does things for us but two years ago almost no one believed that was going to be possible anytime soon you know two years ago what happens now with using chpt was the stuff of sci-fi at best and if you told the world this was going to be part of people's daily lives two years later I think they would have said of course not you know so at the end of that we actually got Sam Alman to say that you know this is going to happen and it's not going to happen you know in the far future which does mean that you know AI is growing as quickly as some people expect it to be and I think that this means that the future systems that open I do have in mind are clearly going to be some of the most capable systems that we've ever seen and might even exceed some people's wildest predictions on what they're able to do now this is something that we've you know widely speculated on for quite some time but hearing Sam Altman say himself is something that is pretty reassuring and something that he also said that you know uh in terms of the most important skills that you can have for the future is of course I think there are two most important skills now I do discuss this on the second Channel where I talk about post AGI economics which is just you know how the economy is going to move forward in terms of the job market and in terms of where people are going to gain their economic agency from and their economic value from and one of the key things that he does talk about is of course number one thinking about what other humans want and being able to provide that is going to of course be an essential skill because humans are going to need a lot of different things than they do need now and of course number to focusing on the humanness aspect which is something that I've been trying to do because in the framework that I've adapted I talk about for post AGI that we're moving towards humanness is of course something that I know people enjoy for example even on this channel something that I try my hardest not to do unless I'm completely ill is of course use AI to do any of the content creation and this is because I truly believe that people value what other humans create and this is something that I've seen time and time again usually when you see AI generated artwork it's fr on usually if there's an AI voice over people don't like listening to it as much of course there are some exceptions but I think humanness in the future for the workplace is most certainly going to be something that is valued uh you know quite a ton because if we're moving to a future where you know AI is going to be I guess you could say outplaced one I think humans are going to Value other humans quite a lot you know it's going to be something that I think we're going to see and by focusing on that it's definitely going to allow you to have a lot more value in terms of just you know rather than automating completely everything that's a Hollywood thing and this is a significant change the world has just gone through um I think this is probably well certainly this is the most significant change to how we use computers since the touch screen on mobile phones um but I think it'll probably be much bigger than that you will be able to just say tell a computer like you would tell a friend or an employee I need this thing to happen or what do you think about this or can you help me out with this or how do you think about this and it'll just do it for increasingly complex definitions of it you know right now it can maybe like write some code for you edit a paper for you uh you know help you analyze things but someday it'll write a whole program for you uh do a whole research project for you help you come up with new ideas uh someday not in the far future so I think it's a very big deal and something that Sam mman did talk about the last bit one that I want to you know just reference once again is of course he said that you know uh you know computers are basically going to just be a able to do something that you want and this is something that we' previously spoken about before where we know that opening eye is of course focusing on agents in the future so it's of course you know openi are pretty much the market leader in terms of what they've been able to do with AI systems across the board um and it seems that you know in the future when we have systems where we're able to say okay I need you to go ahead and do this and that and it's not just a back and forth interface where the AI actually takes 10 steps to complete several different tasks across a range of you know sub field uh it's going to be really fascinating to see how far that potential is pushed to see what an AI system can really do versus what it

can do today and I think you know for the future that's going to be something that's rather fascinating that happens with every technological Revolution and even though I'm confident we're going to gain much more than we lose doesn't mean we're not losing something and we're all loss ofs for good reason um I'll tell you what I think we're not going to lose which is I think we are not going to lose two things the value and depth of Human Relationships how much we care about other humans uh I think people get excited to talk to AI friends for a while and that'll be part of the future for sure but you hear people who do that a lot say man there's really something about knowing this another human and this is like deeply biologically wired in us and I don't think going anywhere um we're going to we are so yeah so deeply wired to care about other humans what other humans think what other humans do the connection we have with other humans we're not going to lose that um so one of the things that Sam of course just talk about there was something that we previously spoke about was of course the humanness now I think the reason that a lot of people are trying to talk about you know humanness in terms of onetoone Human Relationships is because there's been like a slight Trend that most people haven't been paying attention to unless you're actively engaging it in yourself and that is of course um this company here so you can see that this is character AI which is basically a company that you know is nearly catching up to chat GPT in terms of the users and the reason that most people are actually you know talking about character AI in terms of the app and you know how active the user is because we have so many people that are engaging with things that aren't human um and a lot of people are you know talking about how people are engaging with AI girlfriends and how it's becoming an increasingly uh worrying Trend because people are engaging with things that you know right now aren't that good in terms of impersonating what a human could do but in terms of the future we know that these systems are going to get more life like they're going to have better voices probably real time um I mean so it's like you know people are wondering that you know is that going to be something that you know goes on even worse in the future I think it probably is going to be considering the fact that there's already societal issues that you know that exist between men and women in society I'm not going to really get into those issues there but you know there's always going to be issues and if people now have something where they don't need to talk to you know which which either whichever the point I'm trying to make here guys is that if you now have a system where you don't need to engage with another human in order to get some emotional connection the problem is that now you never need to ever talk to another human again um and that leads us to a situation where Some Humans might just stop playing the game all together um in terms of talking to another human and that's you know an entirely another issue but like I said before I think overall humanness will still be one of the most important things because like I said before you know discussing with how you know running this Channel and many other you know Industries uh when things are created by AI it is kind of rounded upon and I think humans do want to engage and interact with other the humans we are going from a world in which intelligence is limited and expensive to Abundant and cheap and if you think about how much any of you could do if you had a massive amount of cognitive labor at your disposal to build the ideas you want to see happen to be useful to other people to provide services and advice um you know right now you can hire people and you can coordinate them and it's kind of difficult and very expensive and most people in the world cannot afford nearly as much let's call it cognitive Service as they like um you know not many people can afford great lawyers for example that's a very specialized very expensive kind of cognitive service if the cost of that the availability of that comes down by a factor of a 100 or factor of 10,000 and not just for legal advice because I don't think anyone needs like lots more legal back and forth but for all the stuff we do want great entertainment great products and Services everything else great education great Medical Care uh that is a profound shift to the world so we're super excited about that and I think that everyone can feel what the magnitude of that transformation looks like your second point is actually not a question that I've been asked many times and I think it's a great one so I appreciate it um one of the things that I learned at YC y common air and also what I learned is I was like a kid studying the history of Technology is you can never go too far making a technology easy to use and accessible um every you know every like 10% easier to use you can make a technology maybe twice as many people use it or they use it twice as much or there's this huge thing and so we had this technology that we knew was pretty cool we didn't know quite how much people were going to like it but we had a sense they were and we put it out first in an API and like some nerds had a good time with it but not very many and it was kind of like unknown in the world we put gpt3 out in the API I think it was in like June maybe it was July of 202 2020 something like that uh and you know people built stuff and other but we started thinking then about like what is the best simplest most natural user interface that we can build on this and I'd had this observation that computers had trended over time um to be as close to the way we interact with other humans or we interact with our physical world as possible so you started out with like punch cards to program computers I don't know how those people did it sounds amazing to me like what an unnatural way to use a computer and they're like literally like sorting these things out on the floor wild but they did it and then you had command lines and that was like a little better there's somewhat of like a kind of framework I can see for that but I'm grateful I never really had to use those computers and then you have the graphical user interface and now finally we're getting something towards more like something the way we interact with the world uh and a lot of people started to use it and we knew how to point at things and the mouse was a reasonable analog for that the keyboard was kind of fake but it was like good enough and this idea that we had these like Windows and graphical information displayed to us like we look at the world we look at a screen there were images that all kind of worked uh the smartphone was then a huge Revolution we got to get rid of that keyboard and that Mouse and just use our hands like again much closer to how we used the world and so we were thinking about what was next in that and sci-fi had predicted this so it shouldn't have taken us as long to figure it out as it did but you really just want a computer you can talk to like you talk to a human we have we are so finely tuned to use language and this the the nuance and sophistication of language um imprecise though it is all the problems with it that it has we can communicate at a very high bandwidth enormously complex ideas with language and sobody said well what if we just go back to this idea of chat Bots people tried it earlier the problem was the chatbot didn't really understand you maybe now it can let's try to build that and then building the chatbot itself the chat interface itself is obviously trivial but the question was how do we tune the underlying model to be really helpful to you and really good at conversation so yeah right there you can see Sam mman talks about the evolution of the kind of systems that we've interacted with uh and how they've gotten better over time and you know that this might be the final I wouldn't even say final actually if you're discussing you know neuralink but I think this is the step before the final I wouldn't say integration with humans but the final uh way that humans are going to interact with technology because I would say now we're moving to that stage where you know nearly everything is going to probably be you know interacting with humans in a natural language setting I mean you know some people are saying that in the future you won't need to code you're going to be able to just say uh you know code me this program you know fix this code yada y yada or go build me this program which is um uh very simple and interacting in a natural language way um and in the future of course I'm talking about you know the final stage where you're probably going to have like you know a neuralink device or something you're just going to be able to think and it's going to be able to interpret that immediately um and go ahead and do the work so I think you know the movement from you know how we interact with systems now to in the future is very telling for how opening I build their next systems cuz he actually also did say that he wants it to be very easy for everyone to use um and the easiness and the E of use in terms of the UI design that they're going to be doing and just how the entire system works I think it's going to be really simple to engage with and interact with um even with regards to you know many of the way that you know if you're trying to build an agent now you know there's a lot of code involved complexity involved so I think openi going to solve that in the future so it will be interesting to see how they do tackle that problem but like I said I think openi one of the key things that I think most people do miss is that openi are also not just a fantastic company that do amazing a research they're actually a company that is focused on building really good products which means that the services that they provide are far superior than any other company in terms of the fact that even if they have a better AI system open ey can usually present it in a better way and I think that's something that's really underrated um and something that most people don't take into account because whilst yes you know some other systems have been and gbt for and the benchmarks um I think in the future just the ease of use of the systems is going to be something that really is uh a point where opening ey do manage to you know stand out in feedback where we take the base model and get it to behave in a certain way and that requires both deciding how it should behave uh and then getting people to sort of say this is a good response this is not a good response or this you know this fits the specification and this doesn't and having diverse representation at all of those steps uh is very important and also figuring out and agreeing as a society on what the behavior should be I know I've mentioned this a few times but it's such a big challenge getting that right requires such a diverse input of voices to do it um I think that'll be critical to the field going forward okay thank you you know there has been two sides of the spectrum on one side there has been Google's woke AI which is where it's you know so Progressive that it unfortunately portrays history inaccurately which is of course a problem because you don't want history to be portrayed inaccurately because inaccuracy is just not what we're aiming for with AI systems and then on the other side you don't want AI systems to be so regressive in the sense that it goes in the other direction so we do want a fair balance and this is something that I think is going to be pretty hard to fix because it's something that AI systems I'm not sure they're inherently understanding of what humans always want now this also brings me to another problem that I was reading about in a research paper which is the idea of model collapse so essentially this paper actually does it's not really talking about work Ai and that kind of stuff but it does talk about something that's pretty similar um and I guess this is kind of you know the problem you know it's kind of like a similar area and basically it talks about you know model collapse so essentially the widespread use of AI systems you know like llms which we talk about all the time could lead to a narrowing of human knowledge over time as these models tend to generalize and focus on content which is pretty common and popular information rather than the rare and specialized knowledge which you know we usually sometimes are exposed to which you know furthers discoveries and just you know it's a more broader picture and this knowledge collap could actually you know harm Innovation and lead to a less Rich understanding of the world and the diversity of ideas get lost because essentially what we have is AI systems that are you know are cheaper um a lot of people produce content with AI and because the AI is so centralized in terms of its beliefs and always kind of you know generalizes with the same kind of answers we can have uh this kind of problem in the future which is I'm not sure how they're going to solve this problem I think they're going to have to uh you know it's probably going to be some synthetic data set and probably some testing on you know trying to understand where the distribution lies in terms of you know the variations where AI is generating content so I think it's going to be interesting to see uh how this goes in the future because these AI systems are going to become more and more widespread and as they uh these hallucinations and the less diverse the content is in terms of the ideas in terms of the Innovation and in terms of you know certain things which are you know in certain industries and certain topics that just really Niche pieces of information I think AI systems won't talk about it that much and this is something that I've noticed myself when talking to AI systems like I will know a lot about a topic and sometimes I'll ask an AI just to make sure that it knows and sometimes it won't even bring up a certain point and like what about this why didn't you bring up this and it's like oh yeah I do remember that and it's like just because it was on the niche end of the spectrum like it was just on the tail end of the distribution the a system sometimes forget to include it so and that's also something that you guys should know as well when you're using chat TBT sometimes it doesn't always give you the entire picture it just gives you the you know centralized distribution of information that it has so it's always important to like still do research on top of chat gbt because a lot of the times it does Miss some of the like key points that you do Miss and it's only really centralized so hi thank you my name is Kiton I'm a sopore science major here and my question is actually tended towards AGI because um the future after AI indefinitely become AGI and how artificial intelligence has emotions or we able to learn from itself and you know that's where the potential risks coming with you know um AI actually having emotions and being able to like well the risk potentially so I'm going to ask where open AI is at in terms of AGI and how do you plan on balancing out the risks and benefits thanks for the question it's uh you know that is probably the thing we think the most about um I think AGI is now like a such a fuzzy term and people use it in so many different ways what you're asking about I think is closer to what I would call like super intelligence not something that can do the jobs that a human can do but say something that can do research do AI research itself maybe as well as all of open ai's researchers and use that to self-improve um and how we think about what the world will look like when we get to that level and how we make sure we confront the risk of such a system um which are very hard to do we have new teams that help us think about being prepared for that world also technical safety work to think about how we can make sure humans stay in control of systems that are more capable than we are I think it's somehow both going to be Stranger Than it seems and also in some other way much more continuous and much more like the world of today humans will still be in control but what any human can do in certainly what any group of humans or Nation can do will be like vastly improved and part of the reason that we try to talk about this even though it scares people or they think we're crazy or both is if we're right this is a huge deal and really important it's going to impact all of us in a huge way and we want the world to have this conversation now like we know that Chachi PT isn't that powerful we know if it was just going to be chat GPT none of these things really matter but given this the steepness of the curve that we're on of the exponential um we want the world to have this conversation so we jointly decide how to balance those risks and benefits thank you just to ask how close do you think you are though to like achieving all that um it's super hard to say I hesitate to give I I'm like always happy to make predictions about what will happen but when in research in particular is super hard but I would say that like in this decade we get to very powerful systems I personally don't believe towards that like thing that can do is re AI research as well as open AI but I've been wrong before um but I

would say like very powerful systems that a lot of people will say like okay for what I want to call AGI this is a version of it an early version by the end of this decade that would be my guess but could be much longer thank you so here we have a Sam alman's deadline well not really deadline just a vague definition and I think Sam Alman rightly so is of course very vague in his predictions because he doesn't want to be someone that's known to continually blast out dates and then of course when they don't happen it's like you know a knock in terms of your reputation by not being able to deliver on said dates but he does say that pretty much by the end of this decade we should be able to get an AGI system which is not far away 6 years um is really quickly if we've seen how quickly the past years have gone I remember Co like it was yesterday so you know 6 years is not that long in terms of getting to an AGI system and of course the point is as well the kind of systems that we're going to be getting before that will be also pretty fascinating as well so that is of course you know something that he did say he did of course leave a caveat in there by stating that it could be you know much longer and of course there are pretty much anything could happen with you know the way Earth is at the moment but the point is that he also does discuss super intelligence which is of course an air system that is able to do research at the level of opening eyes engineers and then improve itself so we did get kind of definition in that aspect but um it is going to be kind of interesting to see how the developments leading up to AGI will be considering you know we're progressively increasing in terms of the capabilities of these systems implications of AI one that comes to my mind is deep fakes and online impersonation of people so how do you think the industry as a whole can sort of medicate that problem there's two different directions I can imagine that coming from um one is when people say something themselves or when they endorse a particular image there's like a cryptographic signature other people can verify and you say this really is a picture I took or quote I said um and we as a society decide that you know we're just flooded with generated media and back to that point about humans caring about other humans we're going to like have these networks of trust and we'll say all right you know what if you didn't sign that photo I'm going to assume it's not real um and if someone didn't sign it that I trust that I trust and I don't have that chain I'm going to assume it's not real so that could happen um the other thing that could happen is that we have enough rules in place on the powerful AI systems that exist that there's a watermarking process that everybody kind of enforces um but with either of those paths there will be a huge amount of generated content on the internet and I think Society is just very quickly going to evolve to understand not to take it too seriously so moment also does discuss here something that is quite interesting and I think this is something that there needs to definitely be some kind of Regulation because currently the problem is with AI generated images is that we don't have a One-Stop solution in terms of verifying whether or not an image is real or absolutely fake and currently there's this big discussion on you know how we're going to get to that solution um and Sam does talk about you know applying some kind of digital ID to that image in order to verify whether or not that image was taken by an AI system or not and I think that this is something that is really important because as these systems get increasingly more comprehensive as they get easier to use you know I mean right now you know mid Journey it's pretty easy to use you log on to Discord y y but as we get to systems where you can just literally say hey create me 1500 images of XY Z and it's able to blast out high quality images within seconds I think we're going to need some kind of system like Google's you know synthetic ID where any major AI text you know image AI system is able to pretty much convincingly have some kind of digital Watermark that you can then run through So Meta have finally released their long anticipated llama 3 model which is an open source model that actually grants access to a variety of new capabilities in terms of how well the model functions when it answers questions and this is a truly Landmark event for the AI community so I'm going to let Mark Zuckerberg State exactly what's going on and then we'll dive into the technical details of exactly what this release means all right big day here we are releasing the new version of meta AI our assistant that you can ask any question across our apps and glasses and our goal is to build the world's leading Ai and make it available to every now today we are upgrading meta AI with llama our new state-of-the-art AI model that we're open sourcing and I'm going to go deeper on llama 3 in just a minute but the bottom line is that we believe that meta AI is now the most intelligent AI assistant that you can freely use to make meta AI even smarter we've also integrated real-time Knowledge from Google and Bing right into the answer we're also making AI much easier to use across our apps we built it into the search box that's right at the top of WhatsApp Instagram Facebook and messenger so anytime you have a question you can just ask it right there and we built a new website mea. ing it from the web we're also releasing a bunch of unique creation features meta AI now creates animations and it now creates high quality images so fast that it actually generates and updates the images for you in real time as you're typing uh it's pretty and you can go check it out now on WhatsApp or the website we are investing massively to build a leading Ai and open sourcing our models responsibly is an important part of our approach the tech industry has shown over and over that open source leads to better safer and more secure products faster Innovation and a healthier market and Beyond improving meta products these models have the potential to help unlock progress in fields like science Healthcare and more so today uh we're open sourcing the first set of our llama 3 models at 8 billion and 70 billion parameters they have best-in-class performance for their scale and we've also got a lot more releases coming soon that are going to bring multimodality and bigger context Windows we're also still training a larger dense model with more than 400 billion and to give you a sense of llama 3's performance this first release of the 8 billion is already nearly as powerful as the largest llama 2 model and this version of the 70 billion model is already around 82 mlu uh with leading reasoning and math benchmark the 400 billion parameter model um is currently around 85 mmu um but it's still training um so we expect it to be industry-leading on a number of benchmarks we're going to write a blog post with more technical details on all of this if you want to go deeper uh in the meantime enjoy meta Ai and let me know what you think so that was Mark's Zuckerberg statement and honestly there is quite a lot of information to dissect because there's just so much from this released and it's actually a lot more than many people including myself did anticipate so let's actually take a look at one of the first things that he talks about and this is of course the benchmarks so we can see here that the benchmarks are actually rather surprising we can see that the meta llama 3 instruct model performance and the reason these benchmarks are so surprising is because if we take a look at these models they're actually state-ofthe-art which means that this is the very best that you can get in terms of a AI so there's nothing better currently that exists at the 8 billion parameter model and at the 7 billion parameter size so with that you have llama 3 leading the way in terms of Open Source now one of the most surprising things that I think most people were actually shocked by was the fact that if you take a look at some of the benchmarks we can see that this right here is Claude Sonet and this was part of Claude 3's family of large language models but it seems to have been surp passed by meta with llama 3 which is quite surprising now we don't know exactly how large Claude 3 Sonet is but it's quite surprising that a 70 billion parameter large language model can actually surpass a state-of-the-art model that many people do use on a daily basis for a variety of tasks which does go to show that this industry is currently and consistently being shaken up in terms of who is the market leader on the benchmarks for different size and different price points and I can see that this for meta and llama 3 is going to be a key area for dominance due to their ability to continually make changes to their models and update them and completely thrash them on benchmarks now to be honest with you guys this was something that I think nobody really expected because these models are just open source and mainly for the developer Community whilst yes we knew there were going to be improvements surpassing Gemini 1. 5 Pro even on some benchmarks like the MML U are obviously quite surprising and of course when we do compare it to the other models at similar sizes like Gemma Google's Gemma and of course mra's 7B instruct we can see that llama 3 absolutely thrashes these models in terms of performance and in terms of General ability overall so this was something that you know you can look at and see that okay right now it seems that currently we're on a path where even companies like mistell are being beaten in terms of their ability to launch large language models or AI systems that are consistantly at the front of the pack in terms of their ability which is to be honest with you guys quite surprising so with that being said there was also some other information that you should know one of the things that they did was they sought to optimize performance for world scenarios and to this end they developed a new highquality human evaluation set this evaluation set contains 18 100 prompts that cover 12 key use cases including asking for advice brainstorming classification closed question answering coding creative writing extraction inhabiting a character SLP Persona open question answering reasoning rewriting and summarization and of course to prevent accid accidental overfitting of our models on this evaluation set even our own modeling teams do not have access to it so essentially what they did here was they used a new human evaluation set and I think this is really important because I've always said that humans are the end users of these products so they should be optimized for humans and not benchmarks and that's what some things like the LM leaderboards have aimed to optimize for and see which llms rank best in that area because it's being used by real humans and that should be by default The Benchmark that people do test their model against because that is what is going to be what people actually use it doesn't really matter if something looks as if it achieves great unlike the mlu or the GSM AK if people can't actually use the model on a day-to-day basis for certain things of course unless it has a really specific use case I think this is going to be something that shows us how good this is now if we look at llama 3 versus the human evaluations one of the things that they did was they tested it against some of the other state-of-the-art models so we can see here that in the initial area we can see that the meta Lama 3 was tested against Claude Sonic and overall the majority of the time it did win in the human evaluation which is like I said before quite surprising and it was 52% a win 12. 9% a tie and 34% a loss but across the board we can see with even mistr medium that meta's llama 3 the 70 billion parameter model is really surprising in terms of its capabilities and across the board it increasingly gets better compared to Mr medium GPT 3. 5 and metas llama 2 so honestly they've actually done a pretty amazing job being more efficient and managing to get an even better AI system while still remaining at the same number of parameters now in comparison with other open source we can also see that the pre-trained model performance does also Excel these other open source and of course these other closed Source models abilities so we can also see the Llama 38 billion PR compared to the mistra and the Gema and it just completely threshes them in terms of performance we can also see the 70 billion parameter model doing better than Gemini Pro 1. 0 and mistr 8 time 22b and this is actually quite surprising considering the fact that MRA actually did just release this 8 time 22b model so I'm not entirely sure how on Earth meta you know were able to secure that lead because it seemed like with mix tral just releasing stuff pretty randomly if you don't know mix tral SL Mistral are an AI company that is completely open source and when they release stuff it's pretty insane because they don't say that okay we're working on releasing this the thing that they do is they literally just release a download link that you can go ahead and then download and then you kind of have to find out exactly what they released how good it is and then kind of Benchmark it yourself there's no real blog post for that so the fact that you know mix TR could just drop anything at open source at any time and they're pretty much the leading open source and to be honest hats off to them because they're not you know backed by a giant company like meta they don't have billions and billions of dollars I mean they recently did raise a 2 billion but you know meta is like a you know multi-billion dollar company so you know the comparison there is a bit stuck but the point is that they were still able to do a lot better than that um and whilst yes it is marginally better it's still pretty surprising that they were able to do that compared to something that was only released a couple of days ago now in terms of the model architecture there are some quite interesting things here llama 3 uses a tokenizer with a vocabulary of 128,000 tokens that encodes language much more efficiently which leads to substantially improved model performance and in addition to that the training data was something that is rather fascinating because people always love to see what you trained your model on and llama 3 is pre-trained on over 5 trillion tokens that were all collected from publicly available sources and they state that their training data set is seven times larger than that used for llama 2 and it includes four times more code and of course to prepare for upcoming multilingual cases over 5% of the Llama 3 pre-training data set consists of high quality non-english data that covers over 30 languages however they do not expect the same level of performance in these languages as English which does make sense so they actually basically with the training data just ensured that their data set was truly high quality and that's why they are able to get so much more out of this model compared to other models at the similar size and remember as we've always discussed when training your models data is one of the most important things as we've seen with smaller models like Orca 2 and Microsoft's F 1. 5 f 2 and of course those smaller models now in addition something crazy that he did actually speak about and I think this is pretty fascinating was that meta's llama 3 is actually going to be a 400 billion parameter model and this model currently is still in training the checkpoint as of last one was April the 15th 2024 and you can see that this pre-trained model is pretty intense you can see that these benchmarks are quite surprising considering the fact that meta have previously not trained a model of this size and this is the first time that we're getting a look at what meta can really do when compared to these close Source companies and I think it's so fascinating to how quickly these llms and AI systems are consed stantly evolving as these giant companies try to one up each other with in addition to just simply doing better in terms of providing a service they just constantly try to raise the level of what these benchmarks are now this was actually compared in a table provided by Jim fan and you can see he States here that the upcoming laa 3 Model will Mark the watershed moment that the community gains open weight access to a GPT 4 class model it will change the calculus for many research efforts and Grassroots startups I pull the numbers on claw 3 Opus gpt's recent model which is GPT 4 2024 049 and of course Gemini and it says llama 3 the 400 billion model is still in training and will hopefully get even better in the next few months there's still many research potential that can be unlocked with such a powerful backbone and expecting a surge in Builder Energy across the system so the pr pretty much what we can see here is that this is basically on the level of GPT 4 and with that it's pretty crazy because if you have an open source level of GPT 4 it now means that people have access to build a variety of different applications and AI systems that they couldn't before in many different ways which means the ecosystem is going to truly evolve from this moment and that's why I said that this llama 3 release was definitely going to be something pretty crazy and it's basically open source GPT 4 when it is released in the couple ofc coming months and I do thinks it's going to be pretty interesting at how they maintain safety with this model because as you know usually when you open source something there are Bad actors that try to tweak the model but it will be interesting to see how meta changes their approaches as their model does get smarter but honestly this table is quite shocking I mean on the MML U it pretty much surpasses the rest of the models gpq it's on par the human eval it's on par and of course the math benchmarks it does lack a little bit but like they said it's still in training so that means that this system will likely get even better now here's something that they also did talk about which wasn't all the technicals currently they did make a new website in order for you to access this model currently so I do want to sayate that if you are in the EU and I know at least if you're in the UK you actually won't be able to access this current page that's because there are just some rules and regulations that means things take longer to get here however I'm going to be doing a full tutorial on this which will be released probably about 1 or 2 hours after this video goes live I would have included it into this video but it is just significantly harder for the algorithm to find especially for those looking to use this now there are many different things like he said there's animations there's different ideas there's different images um so it's going to be pretty interesting to see how this new model is used if you would like to access this model whilst being in the UK like I am of course you're going to have to use a VPN and I think at this point I might just you know find a VPN partner to partner up with because I find myself logging into vpn's all the time because for whatever rules and regulations exist for some reason us people that live in Europe and the UK we are subject to just delayed AI releases I mean I think other people are talking about that but we're probably like looking at it through a lens that no one else is quite imagining yet um I mean we're definitely wrestling with how we when we make not just like grade school or middle school or level intelligence but like PhD level intelligence and Beyond the best way to put that into a product have a positive impact with that on society and people's lives we don't know the answer to that yet so I think that's like a pretty important thing to figure out if you believe which we increasingly do at this point that AI infrastructure is going to be one of the most important inputs to the Future this commodity that everybody's going to want and that is energy data centers chips chip design new kinds of networks it's how we look at that entire ecosystem um and how we make a lot more of that and I don't think it'll work to just look at one piece or another but we got to do the whole thing so I know that this first clip was actually quite vague samman actually does do that sometimes because the details of what he is essentially doing is quite under locking key so to speak and that just essentially means that whilst samman is talking this interview he doesn't want to give away too many key details but what we do know he's most likely referring to here is of course the infamous project Stargate for those of you not familiar with project Stargate it's a100 billion data center that would likely be built over several years for the purpose of building artificial general intelligence and beefing up Microsoft and open ai's joint collaboration efforts to basically become the AI Powerhouse that they could potentially be in the future once AGI does arrive now this is what samman is talking about here and he basically you know later on the interview although I don't inter include the clip in the video but he actually does talk about how in the future this is pretty much going to be one of the most valuable resources on the planet and I think that is something that is quite incredible to State because increasingly in the future as the technology gets better and as the intelligence is more increasingly reliable and more increasingly intelligent the capabilities are to so going to increase which means that access to this kind of technology is going to become something I guess you could say quite like the internet you know how right now we use the internet for a lot of our daily needs and if you don't have the Internet it's pretty hard to function I think in the future AGI is definitely going to be a resource like that which is why he's talking about right now building really big computers is of course a problem and it's something that they're working on and does it gross so you know in terms of just like the compute cost uh correct me if I'm wrong but chat gbt 3 was I've heard it was $100 million to do the model um and it was 100 175 billion parameters gbt 4 was cost $400 million with 10x the parameters it was almost 4X the cost but 10x the parameters correct me adjust me you know it I don't I do know it but I won't oh you can you're invited to this is Stanford Sam okay um uh but the even if you don't want to correct the actual numbers if that's directionally correct um does the cost do you think keep growing with subsequent yes and does it keep growing multiplicatively uh probably I mean now the reason I've actually included this HP clip here is because this actually touches on a fundamental aspect of AGI and AI level systems in the future that we are yet to build the kind of infrastructure that is probably going to power the US economy and the Western hemisphere is something that we truly haven't seen yet and I think this interview gives us the first Glimpse on how much is actually going to be going into this very interesting future now the reason I say that is because he talks about you know increasingly more investment in Ai and of course as you're going to see later on in the video he does talk about certain figures which I'm yet to disclose now this is something that is pretty incredible because you know GPT 4 was allegedly $400 million to train and apparently the future models the future Frontier models are going to be even more expensive than that and there have been some industry people stating that models are going to cost a billion or even 10 billion to train in the future and some people and the reason this is why I brought this up some people are stating that you know models being more expensive in the future could mean that we're using the wrong architecture to train these models but I would kind of disagree because we've definitely achieved a lot by scaling up our Technologies as many of you may know about the better lesson which actually touches upon this subject but I think it's so interesting to note that 400 million for an AI system and we've seen that Google spent like literally 194 million I do believe on Google Gemini and it seems like companies are going to have to even spend more in order to deploy and train future models which is definitely something that is pretty incredible because it means that potentially we're about to enter a new age of AI open is phenomenal chat gbt is phenomenal um everything else all the other models are phenomenal it burned you've earned $520 million of cash last year that doesn't concern you in terms of thinking about the economic model of how do you actually where's going to be the monetization source well first of all that's nice of you to say but Chachi PT is not phenomenal like Chachi PT is mildly embarrassing at best um gp4 is the dumbest model any of you will ever have to use again by a lot um but you know it's like important to ship early and often so I mean it's clear why I included that clip he literally stated that GPT 4 is the dumbest model that we will ever have to interact with again now the reason I included this one and I'm pretty sure maybe this is even included in the title somewhere is I don't think people fundamentally understand what this means and I think that's because people are forgetting how much of a shock it was when GPT 4 was released if you were around at the time of the GPT 4 demo that thing literally broke the internet in terms of how people were so surprised on what it could do now remember Sam here is clearly stating the GPT 4 is dumb this is not the only time he said this if you've watched leex Friedman podcast he frequently talks about you know well he only mentioned it once but he did state that you know gbt 4 is quite dumb he looks up to the ceiling and he's like ah yeah gbt 4 is is quite dumb you know it's unreliable it's pretty bad and he's talking about it you know a product that many people including myself use on a day-to-day basis for a variety of I guess you could say intellectually demanding work in some very light regards of course not many intensive regards but in some you know light regards such as analysis and stuff like that but I think what it shows us is that the next Frontier of model is going to be completely different from GPT 4 because for samman to say not that things are going to be a slight Improvement but to call GPT 4 actually dumb shows us that whatever they are currently training whatever new system they're currently using whatever Frameworks they've implemented in their AI system however they've trained it however they find you the model it's clear that they've made some kind of internal leap because stating that a model is dumb means that potentially whatever they were doing beforehand they've realized the clear mistakes and they've realized that wow what we were doing was actually dumb and what we did is bad so that means that whatever they've got now is clearly much far superior and the reason I think this is so surprising as well is because if we take a look at some of the other companies that are actively participating in the space companies like Claude companies like Google they are really trying to compete with GPT 4 and many people are even like oh wow this model is just as good as GPT 4 or even better this model I'm going to use on a day-to-day basis this is so amazing but the CEO stating that gbt 4 is dumb I mean the writing is literally on the wall for him to say that you know I think it's going to be clear to us that gbt 5 although they don't want it to shock us I think it's definitely going

to be a major surprise um and a major surprise to of course many different AI critics because many people are stating that llms aren't a step towards and are fundamentally not even that good because they hallucinate and they do XYZ but I think whatever open AI are cooking up and whatever they are doing I think that it is truly going to surprise us and I find it truly hard to believe right now what kind of system would make GPT 4 look dumb I understand hallucinations reliability but I'm guessing that they're probably going to use something you know really Advanced that's definitely going to take the industry by stool so I think it's important given what we believe is going to happen to express our view about um but more than that the way to do it is to put the product in people's hands um and let Society co-evolve with the technology let Society tell us what it collectively and people individually want from the technology how to productize this in a way that's going to be useful um where the model works really well where it doesn't work really well um give our leaders and institutions time to react um give people time to figure out how to integrate this into their lives to learn how to use the tool um sure some of you all like cheat on your homework with it but some of you all probably do like very amazing wonderful things with it too um and as each generation goes on uh I think that will expand and that means that we ship imperfect products um but we have a very tight feedback loop and we learn and we get better um and it does kind of suck to ship a product that you're embarrassed about but it's much better than the alternative um and in this case in particular where I think we really owe it to society to deploy iteratively um one thing we've learned is that Ai and surprise don't go well together people don't want to be surprised people want a gradual roll out and the ability to influence these systems um that's how we're going to do it and there may be there could totally be things in the future that would change where we think itative deployment isn't such a good strategy um but it does feel like the current best approach that we have and I think we've gained a lot um from doing this and you know hopefully s the larger world has gain something too so one of the key points that samman actually talks about here is the fact that there's going to be no more surprises now I don't mean this in the literal sense that the next AI systems might not surprise us I think what he's talking about as we saw with Zora because that was definitely a huge shock Ai and surprise just don't go well for the human you know psyche like it's just something that you know we can't really comprehend and it's something that our mental health doesn't really take well because it's something that is essentially a new frontier something that we haven't really seen before and it kind of warps our understanding of what technology is and how we've come to understand it which is why Ai and surprise don't really go well so essentially here this is something that we've talked about many different times before is that of course there might be like model releases and model upgrades but unlike previous releases they don't aim to do stuff that actually shock us that put us in a sense of you know wow this is crazy and evoke some kind of social RIS response because that would essentially mean that you know a public outcry could force you know government regulations or certain policy makers to enact certain laws that I guess would be based out of fear rather than actually regulating technal technological progress in a normal manner so this is something to where we're not going to get any more shocking surprises and I don't think that this is a bad thing I think that basically what he's stating here is that models over time will gradually get better incrementally and it's going to allow you to you know adjust to the model allow your workflows to adjust to it and of course allow the public to adjust to what is now the new normal and what was crazy was that an open a employee actually said that sora's release was basically getting us ready for what's to come because they wanted to show us how crazy AI capabilities are and even then Sora definitely did break the mold in terms of what we saw because it was their first video model and it truly did surpass every other one that was on the platform which pretty much goes to show just how crazy open AI is especially when they put their mind to something so I think it's important to understand here that there won't be any surprises but these incremental updates that you do see on open eyes Twitter feed where you'll seeing them post about small update here and a small update there that is something that you can be expected to have but of course there will be the model updates but they just won't be surprising I think what would likely to be happening is that in the future we're likely going to see incremental updates for these models so with the GPT 5 release I suspect that we won't have one entire release like we did before we're going to have GPT 5 released and there's probably going to be checkpoints along the way where they update that model incrementally and I think what this is probably going to be as well is it's probably going to mean that they are probably on like GPT 6 probably training gpt7 right as GPT 5 is you know in its later stages which means that they're going to be far ahead internally than they are externally if they're releasing things much more incrementally and remember openi can do this because they are the market leaders so they're not going to have a problem at all whether we burn 500 million a year or 5 billion or 50 billion a year I don't care I genuinely don't as long as we can I think stay on a trajectory where eventually we create way more value for society than that and as long as we can figure out a way to pay the bills like we're making AGI it's going to be expensive it's totally worth it now most people actually look at this and I think they look at this the wrong way I think samman is looking at this from a Winner Takes or scenario and I've said this before but some people still don't get it the point here is that he's stating that look we're making AGI it's going to be expensive and it's going to be worth it the point he's trying to make here is that you shouldn't really try to like scrimp and save when it comes to building AGI because the initial bet here is that if you get to AGI a true AGI system that can do autonomous research okay and let's say you can account for 5 to 7% of the economy's total GDP okay which is pretty crazy but let's just say that is something that AGI could capture and as of 2022 the global GDP amounted to over 100 trillion us that could mean okay that this is a winner takes all scenario where if they get like 5% of that that's $5 trillion of economic value that one company is capturing that is absolutely incredible so if they spend a few billion dollars making the AGI system they could surely after get to artificial super intelligence which is winner take s scenario meaning that you know they could pretty much take over they could do robotics if they and if they get an AGI system to efficiently scale Robotics and do robotics they can pretty much do anything okay like if a human is able to do a lot of the things that we do now if they're able to effectively efficiently scale robotics after achieving AGI and Asi then it's pretty much open season for openi on any industry that they want they could just choose the most profitable one and then go from there now I do think interestingly enough there's definitely going to be some new Industries maybe like not on the level of manufacturing Dyson spheres but I do think that with that kind of Technology the breakthroughs that occur after that would definitely have the technology prioritizing some crazy stuff and not just you know the standard basic jobs that we have now but I think that what stamman is stating here that like look he's going to spend $50 billion it doesn't matter because once they get to AGI that's it okay they've won the game it's over now and well not essentially over but those are the companies how on Earth are they going to even try and catch up when a company is advancing exponentially it's pretty hard to do that so I think it's a win takes all scenario and I think the investors know that and I think Microsoft are definitely prepared to invest as much as they can in order to ensureing opening ey get there first and of course they reap the benefits and so do you have a I hear you vision in 2030 of what if I say you crushed it Sam it's 2030 you crushed it what does the world look like to you um you know maybe in some very important ways not that different uh like we will be back here there will be like a new set of students we'll be talking about how startups are really important technology is really cool we'll have this new great tool in the world it'll feel it would feel amazing if we got to teleport forward six years today and have this thing that was like smarter than humans in many subjects and could do these complicated tasks for us and um you know like we could have this like complicated program written or This research done or this business started uh and yet like the Sun keeps Rising the like people keep having their human dramas life goes on so sort of like super different in some sense that we now have like abundant intelligence at our fingertips and then in some other sense like not different at all and you mentioned artificial general intellig AGI I think Sam mman might be downplaying this one here or it's probably realistic because he's always often said that like when you get to AGI it's probably going to happen and then everyone's going to go about their daily lives and I think what he's stating here that like even if they do crush it by 2030 and they achieve AGI by 2030 or in 2030 I don't think Life Changes instantly and I think that is probably kind of true I know that people might think boom we're just snap your fingers and we're in some Singularity type world with flying cars you know these tall white buildings that are completely smooth over and there's like these insane structures I think even if that is our future I think that will take some time so I do think that by the time AI is achieved in 2030 which many do predict it might be 2028 or 2029 I think that it's still going to take some time to build out the infrastructure for future societies the new social contracts and how things are going to work after that so I mean I guess yes this does make sense but um it's something that's definitely very hard to predict I think we need a more precise definition of AGI for the timing question um because at this point even with like the definition you just gave which is a reasonable one there's that's your I'm I'm paring back what you um said in an interview well that's good because I'm going to criticize myself okay um it's it's too loose of a definition there's too much room for misinterpretation in there um to I think be really useful or get at what people really want like I kind of think what people want to know when they say like what's the timeline to AGI is like when is the world going to be super different when is the rate of change going to get super high when is the way the economy Works going to be really different like when does my life change and that for a bunch of reasons may be very different than we think like I can totally imagine a world where we build PhD level intelligence in any area and you know we can make researchers way more productive Maybe can even do some autonomous research and in some sense like that sounds like it should change the world a lot and I can imagine that we do that and then we can detect no change in global GDP growth for like years afterwards something like that um which is very strange to think about and it was not my original intuition of how this was all going to go so I don't know how to give a precise timeline of when we get to the Milestone people care about but when we get to systems that are way more capable than we have right now one year and every year after and that I think is the important point so I've given up on trying to give the AGI timeline but I think every year for the next many we have dramatically more capable systems every year so I think there were two things that were really important in that clip and I know that was a little bit longer than normal but I think that one was pretty crazy because he literally said that every year from this year there will be increasingly capable models and that is something that I think most people can't really comprehend including myself because humans don't have a great time at comprehending exponential Improvement because it just doesn't fit into how we naturally see things growing so it's something that's going to be pretty hard to comprehend but I'm guessing maybe as time increases we may have a better understanding of how that looks in our day-to-day lives because if you were to show GPT 4 to someone two to three years ago you they might like be like what on Earth is this magic and this is of course a recurring problem with AGI that many people have always said that there's not a definitive state of where AGI is and I agree I agree with this completely because AGI is not a thing it is a scale of intelligence that is increasingly better and I think the lines are going to keep getting blurred because right now as many would argue gp4 is a lowlevel AGI system but in the future when we have Vision systems audio systems uh video model systems I mean embodied AGI as well I think we're going to increasingly get further down that scale where we get useful Ai and samman's personal definition was when a system can do autonomous AI research and make scientific breakthroughs so that's samman's definition and I think that definition is important because it gives you guys a benchmark on where he thinks that a true level has been reached fundamentally where breakthroughs are I don't know yet um I mean this sounds like a cop out answer but I think the most important thing about gp5 or whatever we call that is just that it's going to be smarter and this sounds like a Dodge but I think that's like among the most remarkable facts in human history that we can just do something and we can say right now with a high degree of scientific certainty GPT 5 is going to be smarter than a lot smarter than GPT 4 GPT 6 is going to be a lot smarter than GPT 5 and we are not near the top of this curve and we kind of know what to do and this is not like it's going to get better in one area this is not like we're going to you know it's not that it's always going to get better at this eval or this subject or this modality it's just going to be smarter in the general sense and I think the gravity of that statement is still like underrated so the reason I've included this clip and I think if you're going to take one thing from this video take this from the video okay and because I think most people don't understand what Sam mman means and like he said most people think that this is a copout answer this is not a copout answer at all I think him stating that the model is going to be smarter shows us that potentially there might even be some new architectures involved but I think the main thing that we can take away from this is not that the model is just going to be smarter cuz that is kind of vague but that is of course the definition but when I break it down for you basically this means that this model is just going to be so much better at reasoning understanding comprehension context awareness all of these things are truly important to basically being useful like so many times we ask gbd4 to do complex tasks and it just really can't and I think in the future we're going to see how smart A system can get in terms of actually being able to respond to exactly what we want and I think this is what Sam mman is talking about when he says that systems are going to be smarter because GPT 4 has very limited reasoning abilities although it does do some reasoning tasks very well I think future tasks like planning agentic Behavior you know trying to model the world theory of mind all of these things are increasingly intellectual tasks and for an AI system like gbt 4 those are areas where it is pretty rudimentary so I think those areas in the future mean that this kind of system is going to be more widely used because it's going to allow for essentially more wild skill deployment if there's less ulation a lot more planning and a lot

more complex reasoning it's definitely going to be used a lot more because it means that this system is just truly going to be smarter and this isn't something where you know you're looking at the benchmarks and trying to see that but definitely we could be reaching an area where these systems really do you know max out the benchmarks in terms of what's possible and I think that Sam Alman stating that this is essentially going to be where we're at in the future I think this actually gets me really excited for what this Future model is going to be because I think that open AI are on a completely different path to many of the other AI companies that we look at on a day to-day basis because they're still playing catch up and opening I are playing let's set the bar higher so having a model that's it smarter and them guessing that you know they already know where gbt 6 is going to be is of course something that you should you know watch out for so I would say that if you're someone that's building a company um definitely plan with in mind that gbt 5 is actually going to be smarter which means it's going to be applicable for a lot more things and maybe there are certain intellect ual tasks that might disappear with this and I definitely think they will because if this model gets even smarter than it is now it's definitely going to be doing a lot more cognitive tasks which means jobs uh cognitive intense jobs are definitely going to be on the line so that is something that you do need to prepare for because I think it's something that you know it wasn't really talked about in this interview but you know as someone who's looking at things across the Spectrum it's just a heads up for me increasingly worried about how we're going to do this all responsibly I think as the models get more capable we have a higher and higher bar we do a lot of things like uh red teaming and external Audits and I think those are all really good but I think as the models get more capable we'll have to deploy even more iteratively have an even tighter feedback loop on looking at how they're used and where they work and where they don't work and this world that we used to do where we can release a major model update every couple of years we probably have to find ways to like increase the granularity on that and deploy more iteratively than we have in the past and it's not super obvious to us yet how to do that but I think that'll be key to responsible deployment and also the way we kind of have all of the stakeholders negotiate what the rules of AI need to be uh that's going to get more complex over time too so here we have the last clip where samman actually talks about and that's just not where I want to stop the video but he talks about you know responsible AG I think this is probably the hardest thing to do because they might actually have to have some kind of AGI system where it's mathematically impossible for it to be jailbroken or go wrong I know that sounds like sci-fi and whatever but I don't think there's any way currently you know whilst looking at the future any current way to safely uh deploy an AGI but like they said they're probably going to have to do it granularly which means narrowly like some narrow versions of AGI which are you know maybe only working for finance restricted to finance related tasks one that are only related to marketing aren't really General because I think you know as models have been jailbroken before if an AGI system is jailbroken again if it's that smart considering if we extrapolate out to like gbt 8 then it could definitely be used for some pretty heinous stuff so I think that is something that is uh it's it's a problem that I'm glad I'm not working on because it would definitely keep me up at night but I think that this interview was really insightful and I think we can all look forward to the GPT 5 release in potentially 2 months or less so let me know what you thought about this you know project Stargate what you thought about the compute poor and compute Rich GPT 4 being completely dumb which is pretty crazy uh no more surprises of course the winner takes all scenario AGI by 2030 the actual AGI timeline and gbt 5 and gbt 6 getting increasingly smarter