MAJOR AI News: OpenAI's NEW Secret Agent, Googles Stunning NEW AI, New AGI Timeline?

so one of the crazi stories that we have this week is the Microsoft AI CEO predicts AGI within 7 years but not within current Hardware as indicated by Sam Alman I don't think it can be done on Nvidia gb2 200s so this is something that's rather interesting because if you're not familiar with Sam alman's statement he actually said that AGI is achievable on current Hardware so that is a striking statement that we heard from him earlier this year with regards to his comment on Reddit where he did an AMA the whole opening eye team did an AMA on the future of AGI and Ai and you know the products and stuff and they basically said look um you know I think it can be done on current Hardware which is a huge statement now some people are basically stating that look some and said this so that of course they can get more you know more investment which kind of makes sense but um yeah the Microsoft aico mustaf Suman is basically saying that you know it cannot be done now of course like I said before this is s m statement saying it's achievable which is a giant statement because it means that like you know no major you know uh you know Quantum shifts aren needed in terms of the harder aspects it's just a soft problem which means that the timeline is going to be shut down quite a bit now you know if you actually want to know what Mustafa suan really said he said you know he agreed with that you know timeline but he said that you know AGI would be more plausible in the next two to five generations I don't want to say I think it's a high probability that it's two years away but I think it's within the next 5 to 7 years since each generation takes 18 to 24 months now so he's basically talking about you know the generations of tech like nvidia's gb200 chips could be up to a decade or away so I think this one is rather important because we're always looking at the different you know AGI timelines but I think the biggest thing that defines the AGI timelines especially for different individuals and I think when we actually take a look at what the definition of AGI is for Mustafa sulliman we can see that his definition for AGI is rather different he defines it as a general purpose Learning System that can perform well across all human Level Training environments including physical labor now that definition I think is a substantial definition for AGI because the caveat here is that this includes physical labor now for those of you guys know what physical labor is I'm sure you know how crazy physical labor is because as moravi Paradox States what is difficult for humans is often easy for robots and vice versa what is really difficult for robots is quite easy for humans if you've ever seen the kinds of things people do on construction sites you're going to be wondering how on Earth a Tesla bot or a humanoid robot is going to be able to do that and I think we're quite far away from that unless we get some major Innovation I think we've seen a few things you know with uh I'm not sure what the company is called I think it's 1X robotics I think there is another company as well that recently did you know the py robotics stuff they just did some really amazing stuff but physical labor construction you know welding and those kind of things I wouldn't really bet that is 5 years away then again I will say this though I will say that every year AI has surprised me and Robotics as well so if that Trend does continue with which it probably will I guess I probably will be surprised when AGI Robotics are here too so I guess maybe it probably does make sense when we think about that big picture now when we're discussing AGI of course there are different timelines being thrown around another piece of AGI news is the fact that you know Google deepmind research scientist actually believes that AI process could be close

and he basically States this researcher is basically saying that look AI models today with their Advanced capabilities would have actually been considered close to AGI if seen a few years ago and I do agree with that if someone you know saw this technology they would have said yeah this is AI and I think it's like every single year we keep pushing the definition forward on what AGI actually is and it seems like a benchmark will never truly reach until we just blast through to Super intelligence I was thinking about this um earlier this week if 10 years ago five years ago even I would been given the models today right and I would say look there's a secret lab this is a model um play with it and tell me if you think this is actually like close to a general intelligence I would have claim oh yeah that comes from a future where AGI is basically either has happened or I can see that this is very close to it right so the closer you are the more you find oh but it hallucinates of course that's very important right but I think just zooming out I it just feels like okay it's getting pretty close another big story that I actually made a video about and I didn't realize that this was going to get the traction it did but the video I made was talking about how openi new model actually tried to avoid being shut

down so the safety evaluations on the model conducted by Apollo AI safety found that 01 attempted to exfiltrate its weights when it thought that it might be shut down and replaced with a different model so you know it tried to avoid developer oversight and attempted to copy itself when it thought it was at risk of being shut down the company said so this was pretty crazy um you know and I think like the main thing here that like a lot of people are missing and I said this in the video before that like this is something that occurs in maybe 5 to 2% of cases you can see here that they clearly do state that when 01 was led to believe that it would be shut down when acting on its goal and that its actions were monitored it attempted to deactivate this oversight mechanism 5% of the time okay remember in this first one then it said when 01 found memos by its developers that describe how it is misaligned and will be superseded by a new model it attempted to exfiltrate its weights and overwrite the new model in 2% of cases so we have two values here we have 5% and we have 2% now some people will argue that okay this is because number one you guys prompted the model to think it was being shut down so you know you kind of played a you know a trick game on the prompts y y and some people are stating that number two this is because of the fact that it was trained on human data and it's read all the sci-fi novels and stories so it's going to try and run out the lab anyways now my honest opinion is that 5% of the time a model doing something like this is alarming because like I said before we are rapidly advancing these AI systems and if we get to truly agentic systems that are really small there is no stopping an AI system that's able to exfiltrate its weights especially if it's a small model and later on in the video I'll talk about just how powerful these small models are getting now I don't think we're going to be getting to a stage where these models can exfiltrate themselves although I do have a research paper in the works which I'm looking at which is uh saying the otherwise so for those of you who are thinking that this is just a sci-fi fantasy unfortunately it's not and trust me guys even if 95% of the time an AI system is safe that 5% could be completely catastrophic imagine it's the 5% where you're using the AI to do something and it just does something that you don't want it to do some people refer to them as hallucinations I see them from time to time but it's something that like if you're trying to get a model to be reliable or safe in that m in that aspect you know I mean you just cannot have this happen in 2% of cases so this is something that I don't know how they're going to iron this out it will be interesting to see how this is addressed but I think that you know this is not really a positive sign for the AI industry because I know a lot of people think that AI safety is dumb and they think it's stupid and they think ah it's just sci-fi like AI is going to be safe anyways I think that you know even if it's 1% of cases you know we're going to embed AI into society so we need to have some you know probability that basically puts that risk factor at pretty much Zero from AI just doing something absolutely incredible so that is a big thing that we need to fix and that's kind of a problem with these generative Solutions now when we talk about developing you know models and stuff like that and remember how I spoke about you know models getting smaller one of the things that actually happened this week was Microsoft and most people did miss this like I nearly miss this myself but um I'm following Sebastian bck and this guy does a lot of incredible work with the F series now if you aren't familiar with

the F series from Microsoft this is the smallest series from Microsoft that they work on that basically achieves results on par with leading GPT for level large language models so you can see here that the tldr which stands for too long didn't read basically says that five4 is in llama 3. 3 70b category you win some you lose some with five times fewer parameters notably outperforms on pure reasoning like gpq and math 80% so you can see right here that this 14 billion parameter large language model outperforms llama 3. 3 70b in certain areas and you can see as well it outperforms GPT 40 on certain areas and certain benchmarks now this is something that I think is absolutely incredible because when we take a look at what this means for the industry it means that models are continually getting smaller and getting more capable now some people have stated that you know this is just overfitting where you know you train the model on so much of the benchmarks that eventually it's of course going to score higher I personally don't think this is the case I've seen a lot of Innovations in you know smaller large language models so this is something that I believe is just raw Innovation and we're constantly seeing you know many different ways that we can make these large language models more efficient when I was you know doing some insane tracking of arive or arive however you want to call it I remember literally every day I would look through you know 20 or so papers and they were literally all having I think it was around 10 to 15% you know improvements in various aspects of the entire you know um you know from start to finish just entire process of developing an llm there were like 10 to 5% improvements so over time there's always going to be small improvements and eventually you get to a system that is just you know overall much more compact much more efficient so this is not something that I am surprised by and I think even GPT 40 is a lot smaller than the original GPT you know that's 1. 8 trillion parameters something like that so overall we're seeing these models get smarter anyways like llama 3. 3 70b is you know quite better than llama 2 70b and probably llama 2 405b so the point is that these models are going to get smarter they're going to get smaller better and this is a continued Trend so it depends on how small a model can get I really do wonder how you know small these models can get but this is something that I do believe is you know of worthy news and like I said before imagine you have a model that is really smart and it's really small and it's able to just Dart around the internet doing whatever it wants to do it's going to be super interesting to see honestly so um this is 54 it's in the Five series you know I think this is also something that you know people who want to run their own models locally on certain devices this is going to be something that is really useful for them now of course of Google actually say that AI development is finally slowing down and I find this

hilarious okay and not for the reasons that you might think but the fact that he went and said that you know the AI is slowing down and I completely understand what he means by this but they've arguably had the biggest week in AI since last year you know surpassing open AI in a variety of different metrics so the basically what he's saying is pretty much true you know generative AI probably won't change your life in 2025 for most people I think that is probably true but for those of you who do apply it I do think that is probably not going to happen in terms of like I think you know the statement that it won't change your life I think that's pretty wrong if you do apply dtive AI because it is a transformative technology and I think there's a lot of things that people aren't really utilizing it for and I guess this is where I can subtly plug my community we're actually having a case study Sunday every Sunday we actually have a review on how people are making money with AI and I break it down in 10 minutes or so showing you guys exactly how to use this technology so yes the CEO of Google might say that you know what generative AI might not change your life but if you apply like I am in my private Community I'm sure you're going to get a lot of use out of that system now what does he really say about this he basically says that you know I think progress is going to get a lot harder when I look at 2025 the lwh hanging fruit is gone the hill is steeper you're definitely going to need deeper breakthroughs as we get to the next stage now I agree with what he's saying here because previously it was just the era where all you would need to do is just to get as much data as possible throw it at the model keep scaling it up just adding more parameters and eventually the model just gets smarter so that Paradigm in terms of the efficiency of that is over that is no longer the most effective thing to do now I honestly don't know what they're doing at anthropic because that model just it's just like on some aspects it's as small as a one which is pretty impressive but I think the next few years are going to be super interesting because I think some companies are going to be set apart by the innovations that they develop and I think it's going to be different because not everyone will have access to the real inner workings of those methods so I think it's going to be super different with regards to what is happening now that now I think as well if you want to know what is actually happening in 2025 and Beyond hopefully you watched my video earlier today which actually spoke about this

incredible interview and honestly I guess you could say I'm an ear Fanboy but when he makes any kind of interview or any kind of content I just want to absorb all of it because he is like you know one of the Godfathers of AI in terms of the stuff that he's done to transform our understanding of how to use these systems so basically he said something that kind of stunned the AI industry he said pre-training as we know it will end he's talking about comput is growing we have better Hardware we have better algorithms and we have larger crusters but data is not growing okay we have but one internet and it's like the fossil fuel of AI basically referring to the fact that look we've had this scenario like I just said you know 10 seconds ago where you just throw all this data at Ai and it just seemingly gets better they called it you know turning the crank again and of course now that is you know exhausted of course you know we've pretty much exhausted all kinds of data and now we're going to have to invent new Innovative methods in order to get those leaps in reasoning capability so this is kind of good because I think it's now at the point where true Innovation is going to happen and when you have to have that true Innovation I think this is where that real next level of AI is going to come to fruition now what's interesting about this is that he also revealed what he believes is going to come next in Ai and he spoke about you know things like super intelligence things like them being agentic things like it's going to be able to reason truly understand not just be a stochastic parrot as many people refer to current AI systems but it's going to truly understand and it's also going to be self-aware which is a remarkable statement considering the disagreements people have had within the AI industry about whether or not AI is conscious or not for those people that are in you know the chat GPT packed they believe it is just a tool for are some other people they believe you know people that are anthropic they're hiring like welfare researchers to make sure that the AI doesn't you know have any suffering that they believe that AI is more like a person so it's going to be really interesting to see on that road to superintelligence where people do lie now he also said something that was rather striking he also said that reasoning will lead to incredibly unpredictable behavior and self-awareness will emerge in these AI systems so one thing that constantly seeing is that it's apparently something that will emerge in these AI systems as they manage to get smarter so it might be unavoidable that these AI systems have some kind of Consciousness that we can or cannot understand which is a profound statement for the future because it means that we're going to be cohabiting with these intelligences that might be embodied or not the reason you

seeing some early signs of that reasoning is unpredictable and one reason to see that is because the chis the really good ones are unpredictable to the best human chests so we will have to be dealing with AI systems that are incredibly unpredictable they will understand things from limited data they will not get confused all the things which are really big limitations I'm not saying how by the way and I'm not saying when I'm saying that it will and when all those things will happen together with self- awareness of course why not self- awareness is useful it is part your ourselves are parts of our own world models and all those things come together now what's crazy about all of this is that remember how a few slides ago I showed you this slide from Ilia sat that basically said look pre-training as we know it will end in fact Logan Kilpatrick the lead product for Google AI Studio actually said pre-training is only over if you have no

imagination so this is basically I don't know if it's firing Jabs I don't know maybe it's a friendly comment I mean Ilia satk did use to work at Google but this is of course referring to what he said and I think it's kind of interesting because I'm guessing that it now means that Google are hinting at look okay pre-training might be able for you but for us we're just getting started because we've got loads of imagination and we're working on a lot of stuff so it's clear that Google are working on tons of different things and I think this Reddit comment here you know sums up perfectly it says I'm sure Logan said that in just but IIA did qualify his statement with as we know it which isn't exactly incompatible with imaginative methods of pre-training all he was saying is that pre-training is going to work a bit differently in the future which is why he pointed to the example in nature scaling law where the pattern in relationship between mammal size and brain size but homage broke the pattern and followed their own scaling law he assumed it will be similar for pre-training in the future of course Logan would know that if he watched AA presentation not just sound bites or excerpts so I think overall it's just maybe some miscommunication but I think it's going to be super interesting because I believe that this next wave of AI I believe it's truly probably going to be just super intelligent because they're going to be working on real reasoning and not just focusing on you know replicating what they're seeing in the training data which is going to just breed a new Incredible level of AI and someone who's looked at too many research papers and too many videos and too many interview Clips I can say yeah it's definitely going to get crazy now interestingly as I've said you know opening eye are actually building some agents so this is actually rather fascinating because we didn't really know what was happening next week I think this is probably going to be what happens next week so you can

see right here that openai is hiring for a role and this role is the agent infrastructure team so it says that you will play a crucial role in designing and maintaining robust secure systems that facilitate the training of next geni models at a massive scale you will work closely with researchers to enhance system capabilities and support experimental and production workloads and in addition you will work directly on the systems that form the operating system for AI agents building system that allow models to write and execute code and interact with external systems and take actions in Secure environments we're looking for people with deep experience in building AI infrastructure and who are used to working closely with researchers to build high performance systems at massive scale for specialized use cases so right here you can see that this blue highlighted bit is the main bit that you want to pay attention to this is basically hinting at openingi working on agents it's no surprise it's of course the biggest thing that is the future of AI in 2025 that's going to be the year for AI agents so that narrative going into 2025 and of course them hiring for that role is no surprise now if you're wondering what kinds of Agents of course openingi have shown us many demos before but one of demos that they recently showed us was in Paris honestly I would love to show you guys a snippet of this demo but the audio is just simply so terrible because it's not an official video so it's echoey it's quite hard to see and hear but essentially what this demo is it's just like an AI agent that is in real time you know helping you do shopping so I believe that since we have 12 days of open a ey that on the last couple of days it's quite likely that we will get some kind of you know open a ey agent it's quite likely that we will get some kind of AI agent that's able to do a vast majority of things and this is going to be super interesting because you know they released this agent but Google have released theirs too and I'm got to be honest Google's products and stuff are really good like a lot of the products like notebook alemon stuff like that those are products that I now use on a day-to-day basis like they are really competing with opening eye in ways that you wouldn't have you know even thought about and my speculation for the week coming up isn't random you

can see right here that Jimmy apples and no notorious openi leaker or AI leaker from the big Labs has tweeted that feeling when we haven't even had the big week yet maybe some of you don't deserve it so he's talking about the fact that there is still a very large week that is going to happen from open AI where they have certain products that are probably going to be released it's rather fascinating considering the fact that we've had a decent amount from open aai and some people are now starting to doubt them in terms of their potential what they're able to do for the industry and some people are stating that look open are behind now I think a lot of this is just because people want something from open AI because they're used to stunning and flashy demos so it will be interesting to see what they have but of course if you had been paying attention this week you'll know that they released Sora which is you know now on the artificial analysis video generation leaderboard Arena we can see it actually ranks at number one so Sora for those of you who were wondering okay which model is currently the best in terms of overall quality and coherence this is the leaderboard that you probably should screenshot and save to your phone or just visit the artificial analysis leaderboard and you'll see whatever model is currently the best now I don't know if they're going to hold this lead for that long considering the fact that these other models aren't far behind and those other models they're just focusing on video so it would be easy for them to surpass them in that area so it's going to be super interesting for that but I think one of the biggest things about this that most people are missing is the fact that with opening ey Sora one of the key things that people are missing is that like this technology isn't just super good video like yes opening I Sora is really good in terms of the quality the coherence and all that stuff which is you know remarkable it's remarkable that we have access to this kind of Technology the only thing I would say is that what's crazy about this one is that like Sora is that Sora is actually something that is a really nice piece of software so I think open AI was smart to release Sora with all of the additional features they could have just released this as a standard text video model but what you're seeing here is sora's you know storyboarding feature and this is something that changes the game because I think a lot of the times people fail to realize that look whilst yes working on an AI model is good at the end of the day you still have to build a relatively good product around that and one that is easy for users to use that way you know your company is still viable like you could develop the best text model in the world but if it's wrapped in you know terrible UI terrible software people aren't going to use it as much so I think openi are really the kings of that because this is going to truly allow people to build you know and create films that are really effective even for beginners so this is something that I just personally believe that like a lot of people are not paying attention to and it's probably one of the reasons why openi is still winning quite a lot now if we actually go back to Google who stole the show this week they actually released several things one of them being this quantum computer that is just you know has unprecedented computational speed and this completed a background Benchmark in under 5 minutes which is a task that would take you know one of the fastest supercomputers approximately 10 septian years which is incredible which and this performance surpasses the age of the universe which is just you know like I mean I'm literally speechless I don't even have

the word to describe how crazy this is and this Quantum era seems to be upon us and Sati Adela you know also talks about how science itself is becoming computed and that increasing Innovation will be driven by Ai and quantum computers leading to the industry becoming completely transformed there is going to be more and more Innovation that's driven by Ai and Quantum right I mean Chicago has always been a place where there's been great engineering schools great schools of science um you know what governor pitar has done and just getting this Quantum programs here and just driving that connection between the universities is fantastic and so to us you know that by the way is to your point is not even just the few scientists right so to me it's about an industry that will get transformed that will then creative and what are like how does you know if science itself is becoming computed right that's kind of what's happening I think we're going to have more scientists right one way to say it is the elite science will not go away Elite science will remain Elite science but I do feel that we are at a stage where there will be more people involved in science right whether it's Material Science or whether it's in biology to help move the frontiers of productivity and Science and that's where I think Chicago can Le now Google not only released their you know AI stuff of course with Gemini 2 they also released the Android XR glasses which are just absolutely outstanding like I've seen Gemini it's actually one of the most underrated AI models currently I think it's currently the best model in terms of speed in terms of usage multimodalities currently the best model out there and they're going to release this in the future which is where you're going to be able to have these kind of meta star Rayband glasses with Google Gemini embedded in them and I think this

is going to be super incredible because once you have this kind of AI That's embedded for the average person and it's really usable it's lightweight it's easy to integrate you know it's able to translate things for you honestly it's quite like an iPhone but except it's less invasive and it's just natural in there like you know if you're on the phone or if you're reading something it doesn't look like you're looking down at your phone not at you know the person in the conversation it's just like a natural thing to have I think it's going to be super interesting with how this integrates to society because I think it's going to be adopted a lot quicker than most people think and as someone that's used The Meta glasses before if those had an onscreen AI that was able to do that I don't think I'd ever take them off so this is going to be a new paradigm and uh it was getting a lot of traction when I made a video on it so it's going to be really interesting to see where this kind of thing heads in the future I hope that this isn't some kind of flashy demo because Google have done that once before and it would be kind of disappointing and of course you know it is just a prototype at the moment but I do think that this is possible because I've seen a developer do this with I think it was a Raspberry Pi or something but there is a glasses out there that is this already so it wouldn't surprise me if Google is able to cook up something that's even better than that I mean trillion do company the best engineers in the world the best researchers in the world so I'm pretty sure that this will look like that in theory so um when that is going to come out they haven't released the date for that but the future with AI is going to be one that is seamless rather than you know having to sit down at a computer and of course prompt that model again and again now this isn't the only thing that Google are releasing in the future of course there is a deep research tool and if you haven't used deep research yet I promise you are missing out it's going to be

one of the next big things it's basically I don't know if it's a team of Agents or just one singular agent I think it is a team of Agents you know working together to you know do the research but this is my daily go-to tool when I do research now I'll leave a link to the tutorial just check my channel check my recent video I made a you know 10-minute video on this and it's pretty incredible what it's able to do it's able to search over a 100 different websites and it's able to give you I probably should have used a better you know image for this but it's able to give you a comprehensive research guide on exactly the topics that you do want which is incredible so it's something that's going to save you hours of time every single day you know instead of crawling in through this website that website is going to give you all that data in a comprehensive research report it is simply outstanding and of course something that Google are going to be releasing very soon uh I'm not sure but there was a tweet that spoke about how you know Google has these models ready to load and they've got the vo 1. 5 and VO2 now this is rather interesting because it means that Google has still been working on their video model which most people don't even know exists so this demo that you're current ly seeing is Google's vo video model and this is basically you know what opening eyes Sora is Google this is vo to that so Google's video is a text video model it's able to generate a minute long Clips with really long consistency this video right here isn't the best demo of it there are a lot of other demos that are a lot more in detail so this probably isn't the best demo you can see cuz you know it's a bit blurry but I would expect vo to do a lot better than people think considering when I've seen the quality it seemed to be really effective so with that being said let me know what you're most excited for what you think the future of AI is open A's got a secret demo Google's got some secret video model coming out crazy Quantum future and the future of ASI is nearly here with that being said hopefully you guys enjoyed today's video and I'll see you in the next one